| A | B | C | D | E | F | G | H | I | J | K | L | M | N | O | P | Q | R | S | T | U | V | W | X | Y | Z | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1 | Sl No. | Title Of The Paper | Aim | Dataset | Methods Used | Results | Reference | By Whom | ||||||||||||||||||
2 | 1 | Predicting Colorectal Cancer Using Machine and Deep Learning Algorithms: Challenges and Opportunities | To systematically review state-of-the-art ML and DL techniques for predicting colorectal cancer, analyzing their aims, methods, datasets, and identifying challenges and opportunities. | Multiple datasets: 307 colorectal cancer slides, gene expression datasets (e.g., GEO, TCGA, HTSeq-FPKM-UQ), and image datasets (e.g., LC25000, NCT-CRC-HE-100K, CVC-ClinicDB). | Machine Learning: RF, SVM, KNN, DT, AdaBoost, XGB. Deep Learning: CNN, UNET, ResNet, YOLOv3, VGG16, Autoencoders. | Reported high accuracies (up to 100% in some cases). DL models performed better on image datasets. Challenges include data imbalance, small sample size, and lack of standardization. | Alboaneen et al., Big Data Cogn. Comput. 2023, 7(2), 74. https://doi.org/10.3390/bdcc7020074 | Abhay | ||||||||||||||||||
3 | 2 | Colorectal Cancer Detection Based on Deep Learning | Colorectal Cancer detection and segmentation using DL methods from H&E stained histology slides | 307 colorectal cancer related slides (85 normal 222 colorectal cancer) from St Paul's Hospital then random 275 slides(76 normal and 199 colorectal cancer) for training and remaining(9 normal and 23 colorectal cancer) for testing.for better estimation of the model, its performance was tested on a completely independent dataset of 50 CRCslides obtained from Kaohsiung Medical University Chung-Ho Memorial Hospital. E. | CNN method was used.During segmentation,the slides were divided using the classificated based approach into multiple patches and predicted independently and then the final segmentation is done by obtaining all patch predictions.Patches with maximum cancer compromised cells are labelled as tumor positive and with no cancer as tumor negative. | The median prediction for normal slides was 99.9% and cancer slides was 94.8%.It could be improved by using more digital slides for training.For cancer patches,the labels are determined by the percentage of cancer regions in patches. Therefore, small variations may cause the label to flip from tumor positive to tumor negative or vice versa. This ambiguity may cause the model to be less accurate at the edge of a tumor region. | Xu L, Walker B, Liang PI, Tong Y, Xu C, Su YC, et al. Colorectal cancer detection based on deep learning. J Pathol Inform 2020;11:28. | Ankita | ||||||||||||||||||
4 | 3 | Classification of Colorectal Cancer using ResNet and EfficientNet Models | To develop an efficient classification system for colorecteral cancer classification ussing CNNs on histological images | The study utilized the “Kather-texture-2016-image” histology dataset consisting of 5000 histological images consisting of human colorectal cancer tissue and healty normal tissue images.Each image was of 150*150 pixels distributed across 8 distinct tissue classes based on their texture characteristics. | The research uses the CNN models(RESNet34 and EfficientNetB4) such as ADAM optimizer for RESNet34 and ADAMmax for EfficientNetB4, demonstrating a sophisticated understanding of the optimization environment for each model.ResNet34 and EfficientNetB4 improve image classification methods for colorectal cancer by mitigating the vanishing gradient problem and training deeper networks. | According to the findings, both the ResNet34 and EfficientNetB4 models perform well in classifying colorectal cancer histological tiles. However, ResNet34 performs better in terms of total accuracy. On testing over 50 epochs at a learning rate set to 0.005, ResNet34 got an accuracy of 99.976%. The EfficientNetB4 has been halted at epoch 34 after 3 adjustments of the learning rate with no improvement. In this experiment, we got an accuracy of 99.886%. | Abhishek, Ranjan, A., Srivastva, P., Prabadevi, B., Rajagopal, S., Soangra, R., & Subramaniam, S. K. (2024). Classification of Colorectal Cancer using ResNet and EfficientNet Models. The Open Biomedical Engineering Journal, 18(1). https://doi.org/10.2174/0118741207280703240111075752 | Ankita | ||||||||||||||||||
5 | 4 | Artificial Intelligence Driven Colorectal Cancer Classification Via Deep Learning Technique | To improve early detection of colorectal cancer (CRC) by classifying colonoscopy images for polyps using CNNs and comparing the performance with other neural network models. | Kaggle curated colon image dataset | Artificial Neural Network (ANN), Backpropagation Neural Network (BPNN), Convolutional Neural Network (CNN) | CNN achieved 80% accuracy (better than ANN: 50%, BPNN: 65%). Demonstrated CNN's superior performance for image-based colorectal classification. | Ananthi S et al., Proceedings of the 4th International Conference on Ubiquitous Computing and Intelligent Information Systems (ICUIS-2024), IEEE, ISBN: 979-8-3315-2963-5 | Abhay | ||||||||||||||||||
6 | 5 | Early-Stage Detection of Colorectal Cancer using Image Classification | To develop a deep learning-based system using EfficientNetB0 for early-stage detection and classification of colorectal cancer from histopathological images, including a user-friendly interface for prediction and education. | Kaggle curated colon image dataset | EfficientNetB0 architecture with transfer learning, image preprocessing, Streamlit-based interface for prediction and confidence analysis | Achieved 92% accuracy; Precision 90.5%, Recall 91.7%, F1-Score 91.1%, AUC-ROC 0.94. High performance across six CRC classes (e.g., adenocarcinoma, polyps, normal tissue). | Ashlin Santhosh et al., Proceedings of the 4th International Conference on Sentiment Analysis and Deep Learning (ICSADL-2025), IEEE, ISBN: 979-8-3315-2392-3 | Abhay | ||||||||||||||||||
7 | 6 | Transformer-Based Self-Supervised Learning and Distillation for Medical Image Classification | To enhance colorectal cancer tissue classification using Swin-Transformer V2, combining self-supervised learning and progressive distillation for improved model accuracy and efficiency. | NCT-CRC-HE-100K, DigestPath, CAMELYON16, TCGA, FracAtlas, INSPECT, MedMNIST | Swin-Transformer V2, Self-supervised learning (e.g., MoBY), Layer-wise knowledge distillation, Transfer learning, Data augmentation, Temperature scaling | Achieved top-1 accuracy of 95.0% with Swin-B (ours), outperforming ViT, EfficientNet, and RegNet variants; showed +1.7% accuracy improvement via distillation. Demonstrated robustness with 94–96% accuracy across model sizes. | Meng Li, 2024 3rd International Conference on Cloud Computing, Big Data Application and Software Engineering (CBASE), IEEE, ISBN: 979-8-3315-0658-2 | Abhay | ||||||||||||||||||
8 | 7 | Colorectal Cancer Classification using Deep Convolutional Networks An Experimental Study | To propose a deep learning technique based on Convolutional Neural Networks (CNNs) to differentiate adenocarcinomas from healthy tissues and benign lesions. | The dataset used in this study was extracted from a public repository of H&E stained whole-slide images (WSIs) of colorectal tissues, available on line at http://www.virtualpathology.leeds.ac.uk/. | Convolutional Neural Network: Architecture and Training Paradigm and Transfer Learning From Pre-trained CNN that consisted of two steps i)Pre-trained CNN as A Fixed Feature Generator and ii)Fine-tuning of Pre-trained CNN. | All the proposed classification frameworks obtained accuracy (both patch and patient-wise) above 90%. In our experiments the transfer learning techniques outperform the full training approach both in terms of classification accuracy (above 96%) as well as in terms of training time. Hence, they demonstrate that low-level features learnt by the CNN in a very different context (the ImageNet, in this case) can be successfully transferred to the classification of colorectal images. | Ponzio, F., Macii, E., Ficarra, E., & Di Cataldo, S. (2018). Colorectal Cancer Classification using Deep Convolutional Networks - An Experimental Study. Proceedings of the 15th International Joint Conference on Biomedical Engineering Systems and Technologies. https://doi.org/10.5220/0006643100580066 | Ankita | ||||||||||||||||||
9 | 8 | Machine Learning Techniques for Medical Image Processing | This paper introduces a survey on ML algorithms used in medical image processing and it focuses on two main types (supervised and unsupervised learning) and their importance in medical image processing. | An open access brain tumors dataset has been used. This dataset consists of 150 magnetic resonance images used for training (84 images belong to the benign class and 66 images belong to the malignant class) and 50 resonance images used for testing ( 16 images are classified as “benign” and 34 images are classified as “malignant”). | K Nearest Neighbour,Decision Tree,Logistic Regression,Support Vector Machines(SVM) and Random Forest. | K-Nearest Neighbors algorithm gave the best outcome in contrast with the others, followed by the Decision Trees algorithm which got the accuracy of 99.429%. SVM algorithm proved an accuracy of 97.287% followed by Logistic Regression algorithm (94.246%). Random Forest algorithm had an accuracy of 90.128%. | Rashed, B. M., & Popescu, N. (2021). Machine learning techniques for medical image processing. 2021 International Conference on e-Health and Bioengineering (EHB), 1–4. https://doi.org/10.1109/ehb52898.2021.9657673 | Ankita | ||||||||||||||||||
10 | 9 | Colon Cancer Diagnosis Based on Machine Learning and Deep Learning: Modalities and Analysis Techniques | Provides a review on the study of colon cancer classified into machine learning(ML) and Deep Learning(DL) techniques. | ------ | ------- | ----- | Tharwat, M., Sakr, N. A., El-Sappagh, S., Soliman, H., Kwak, K., & Elmogy, M. (2022). Colon cancer diagnosis based on Machine learning and Deep Learning: Modalities and analysis techniques. Sensors, 22(23), 9250. https://doi.org/10.3390/s22239250 | Ankita | ||||||||||||||||||
11 | 10 | Detection and identification of Colon Cancer and Rectum Cancer Using Fluorescence and Raman Spectrum | To investigate the diagnostic potential of Raman and laser-induced fluorescence (LIF) spectroscopy in detecting metabolic differences in serum samples of colon and rectum cancer patients. | 82 colon cancer, 69 rectum cancer, and control serum samples | Laser-induced fluorescence (LIF), Raman spectroscopy, spectral analysis (488.0nm and 514.5nm excitation), red shift, intensity ratios | Achieved 80.7% (colon) and 82.5% (rectum) accuracy compared to clinical diagnosis. Used parameters like red shift, fluorescence ratios, and Raman peak intensity to distinguish between normal, colon, and rectum cancer. | Xiaozhou Li et al., Proceedings of the 27th Annual IEEE EMBS Conference, Shanghai, China, 2005. ISBN: 0-7803-8740-6 | Abhay | ||||||||||||||||||
12 | 11 | An Integrated Microfluidic System for Screening of Peptides Specific to Colon Cancer Cells and Colon Cancer Stem Cells Using the Phage Display Technology | To automate and accelerate the selection of colon cancer and cancer stem cell-specific peptides using a microfluidic chip integrated with phage display technology. | Colon cancer cell line (HCT-8), colon cancer stem cells derived from HCT-8 | Phage display (M13 library), microfluidics, magnetic bead-based biopanning, multi-round selection with variable shear force, PCR and electrophoresis analysis | Completed peptide screening in 36 hours vs. ~1 month traditionally. Achieved high specificity for CRC and CSCs. Identified 3 peptide candidates for HCT-8; process is automated and highly efficient. | Yu-Jui Che et al., Proceedings of the 9th IEEE International Conference on Nano/Micro Engineered and Molecular Systems (NEMS), 2014, Hawaii, USA. ISBN: 978-1-4799-4726-3 | Abhay | ||||||||||||||||||
13 | 12 | Gastric and Colon Cancer Imaging with Swept Source Optical Coherence Tomography | To investigate the use of Swept Source Optical Coherence Tomography (SS-OCT) for imaging gastric and colon cancer tissues for diagnosis and surgical guidance. | Clinical samples from 12 patients (5 gastric, 7 colon cancer) | SS-OCT imaging, histopathology comparison, real-time intraoperative imaging | OCT images showed clear structural differences between normal and cancer tissues. Cancerous tissue presented disordered and heterogeneous layers compared to normal tissue. SS-OCT provides micron-scale resolution suitable for surgery guidance. | Site Luo et al., Presented at conference via Tsinghua University & collaborators. | Abhay | ||||||||||||||||||
14 | 13 | Colon Cancer Detection by Designing and Analytical Evaluation of a Water-Based THz Metamaterial Perfect Absorber | To develop a non-invasive colon cancer detection method using a water-based terahertz (THz) metamaterial perfect absorber that differentiates healthy and cancerous tissues based on spectral response. | Theoretical/simulated samples (healthy vs. cancerous colon tissue); CST Microwave Studio simulation | THz Metamaterial design, CST-based modeling, electromagnetic simulation, SPP resonance analysis | 99.82% absorption at 707.5 μm; spectral shift observed between healthy (719.4 μm) and cancerous (709.2 μm) tissues. Device shows potential for optical cancer diagnosis based on water content variation. | Zohreh Vafapour et al., IEEE Sensors Journal, Vol. 21, No. 17, 2021. DOI:https://ieeexplore.ieee.org/document/9449873 | Abhay | ||||||||||||||||||
15 | 14 | A MachineLearning Approach to Diagnosing Lung and Colon Cancer Using a Deep Learning-Based Classification Framework | Provides a classification framework using Deep Learning(DL) and Digital Image Processsing(DIP) to differentiate five types of lung and colon tissues(two benign and three malignant) using histopathological images. | This research worked with lung and colon cancer histopathological image dataset known as the LC25000 dataset .This dataset contains 25,000 color-images of five types of lung and colon tissues . They used 70% of the images (randomly chosen) to train this super vised learning model and the remaining 30% image to test it. | After choosing the histopathological image dataset, features were extracted from them using two separate algorithms(Extraction of 2D Fourier Features and Extraction of 2D Wavelet Features). The collected features were concatenated to formulate a combined feature set, and classification was performed based on it using a multi-channel CNN. | On the testing subse(500 epochs)t, the classification accuracy at the last epoch was 95.11%; however, the best outcome was achieved at the 392nd and 488th epoch, both of which yielded an accuracy of 96.33%. The highest training accuracy was 98.91% (493rd epoch), which is very close to the accuracy of the last epoch (98.87%). The testing accuracy curve was not as steady as the training accuracy curve, which indicates the occasional decline in performance. However, the result improved as the training process continued. After the 100th epoch, almost 55% of the testing subset’s recorded accuracy values were over 95%. The curve fell below 90% only four times, which assures that the model can provide a good classification outcome even if it is built with fewer epochs. | Masud, M.; Sikder, N.; Nahid, A.-A.; Bairagi, A.K.; AlZain, M.A. AMachine Learning Approach to Diagnosing Lung and Colon Cancer Using a Deep Learning-Based Classification Framework. Sensors 2021, 21, 748. https://doi.org/ 10.3390/s21030748 | Ankita | ||||||||||||||||||
16 | 15 | Deep Learning on Histopathological Images for Colorectal Cancer Diagnosis: A Systematic Review | Provides a review of colorectal cancer using Deep Learning methods on Histopathological Images. | ------ | ------ | ------ | Davri, A.; Birbas, E.; Kanavos, T.; Ntritsos, G.; Giannakeas, N.; Tzallas, A.T.; Batistatou, A. Deep Learning on Histopathological Images for Colorectal Cancer Diagnosis: A Systematic Review. Diagnostics 2022, 12, 837. https://doi.org/10.3390/ diagnostics12040837 | Ankita | ||||||||||||||||||
17 | 16 | Locality Sensitive Deep Learning for Detection and Classification of Nuclei in Routine Colon Cancer Histology Images | Proposes a Spatially Constrained Convolutional Neural Network (SC-CNN) to perform detection and classification of nuclei in routine hematoxylin and eosin (H&E) stained histopathology images of colorectal adenocarcinoma. | 100 H&E stained histology images of colorectal adenocarcinomas, consisting of more than 20,000 annotated nuclei belonging to four different classes. | Nucleus detection was done using Spatially Constrained Convolutional Network(SC-CNN) and nucleus classification was done using softmax CNN. | Results show that the joint detection and classification of the proposed SC-CNN and NEP produces the highest average F1 score as compared to other approaches. | Sirinukunwattana, K., Raza, S. E. A., Tsang, Y., Snead, D. R. J., Cree, I. A., & Rajpoot, N. M. (2016). Locality sensitive deep learning for detection and classification of nuclei in routine colon cancer histology images. IEEE Transactions on Medical Imaging, 35(5), 1196–1206. https://doi.org/10.1109/tmi.2016.2525803 | Ankita | ||||||||||||||||||
18 | 17 | Automated Detection and Characterization of Colon Cancer with Deep Convolutional Neural Networks | Promotes a system for detecting and classifying colon adenocarcinomas by applying a deep convolutional neural network (DCNN) model with some preprocessing techniques on digital histopathology images. | From the LC25000 datasets, a total of 10 thousand digital photographs of histopathology slides were used. It contains 500 images of colon tissue in total (250 images of benign colonic tissue and 250 images of colon adenocarcinomas), which have been augmented to 10,000 images using the Augmentor program. | The proposed method uses transfer learning with pre-trained models such as ResNet50, DenseNet121, EfficientNetB0, VGG-16, and MobileNetV2, alongside a custom DCNN. The approach includes fine-tuning on a medical colon cancer dataset, applying data augmentation to address limited labeled data, and optimizing using the Adam optimizer. Backpropagation is used for weight updates to distinguish between benign and cancerous tissues. | Compared to the other transfer learning pretrained models the proposed DCNN model showed i)Precision-100% ii)Recall-99.59% iii)F1 score-99.80% iv)Training Accuracy-99.87% v)Testing Accuracy-99.80%. | Hasan, M. I., Ali, M. S., Rahman, M. H., & Islam, M. K. (2022). Automated Detection and Characterization of Colon Cancer with Deep Convolutional Neural Networks. Journal of Healthcare Engineering, 2022, 1–12. https://doi.org/10.1155/2022/5269913 | Ankita | ||||||||||||||||||
19 | 18 | Colon Cancer Detection Using YOLOv5 Architecture | To develop an AI-based colon cancer detection system using YOLOv5 for rapid and accurate tumor classification into benign and adenocarcinomas. | Custom dataset of 2,000 colon images (1,000 benign and 1,000 adenocarcinomas) from Kaggle | YOLOv5 architecture, CNN, PyTorch, Roboflow for preprocessing, data augmentation | Achieved 99.28% accuracy; specificity 100%, sensitivity 98.63%, F1-score 99.03%; latency per image: 0.031 sec | Reshmaja K Ramesh et al., 2022 International Conference on Communication, Computing and Internet of Things (IC3IoT), IEEE. DOI: 10.1109/IC3IoT53935.2022.9768016 | Abhay | ||||||||||||||||||
20 | 19 | EfCNN-Net: Smart Detection of Colon and Lung Cancer using Histopathological Images | To propose a deep learning framework using CNN-based architectures for accurate detection of colon and lung cancer using histopathological images. | LC25000 dataset (25,000 histopathological images after augmentation, 5 classes: benign colon, benign lung, lung squamous, lung metastatic, lung adenocarcinoma) | VGG16, ResNet50, EfficientNet B1, EfficientNet B3, image augmentation (Augmentor), CNN-based classifiers | Achieved accuracy: VGG16 - 100%, ResNet50 - 99.76%, EfficientNet B1 - 99.45%, EfficientNet B3 - 100%; highest reported accuracy in the study: 99.7% | Nandini Kapoor et al., 2023 3rd International Conference on Intelligent Technologies (CONIT), IEEE. | Abhay | ||||||||||||||||||
21 | 20 | Colorectal cancer detection based on convolutional neural networks (CNN) and ranking algorithm | The model proposed in this work is an application to detect colorectal cancer based on the Convolutional Neural Network and Ranking algorithm. | Dataset taken from cancer.net classified into three types as Tumor(T),Node(N),Metastasis(M).The dataset is a 4-year follow-up data from 334 patients treated from colorectal cancer.From the dataset 284 images were used for training and the test was of size 50.Thre dataset had been augmented to 5 times using simple image processing mechanisms such as image enhancement,image quality adjustment etc. | The proposed model fuses CNN and LSTM simulator models to identify the affected tumor tissue.CNN trains the model to predict the tumor similaarity and LSTM helps in identifying. | The proposed model showed :- i)Precision-92% ii)Recall-93% iii)Accuracy-91% which was better than the performance of other classifiers such as ANN,BPNN. | Karthikeyan, A., Jothilakshmi, S., & Suthir, S. (2023). Colorectal cancer detection based on convolutional neural networks (CNN) and ranking algorithm. Measurement Sensors, 31, 100976. https://doi.org/10.1016/j.measen.2023.100976 | Ankita | ||||||||||||||||||
22 | 21 | Colon cancer survival prediction using ensemble data mining on SEER data | To develop accurate colon cancer survival prediction models using ensemble classifiers and class balancing with SMOTE on SEER data. | SEER Colon and Rectum Cancer Data (1973-2009), 105,133 colon cancer patient records after preprocessing | Decision Trees (J48, REPTree), Random Forest, Alternating Decision Tree, Logistic Regression, Ensemble Voting, Bagging, AdaBoost, Random Subspace, SMOTE for class balancing | Ensemble voting achieved 90.38% (1-year), 88.01% (2-year), and 85.13% (5-year) survival prediction accuracy with AUC values of 0.96, 0.95, and 0.92 respectively. Balanced datasets significantly improved performance. | Reda Al-Bahrani et al., 2013 IEEE International Conference on Big Data. | Abhay | ||||||||||||||||||
23 | 22 | A Robust Colon Cancer Detection Model Using Deep-Learning | To develop a noise-robust colon cancer detection model using a two-phase architecture with deep learning to handle noisy histopathology image labels. | Chaoyang Dataset: 512×512 colon slide patches (training: 4021, testing: 2139) from Chaoyang Hospital, China | Two-phase Hard Sample Aware Noise Robust Learning (HSA-NRL) algorithm, Bayesian classifier, CNN, MobileNet, ResNet-34, XCiT, SqueezeNet, co-learning strategy | MobileNet backbone achieved highest test accuracy of 84.39%, outperforming ResNet-34 (83.40%), XCiT (80.83%), and SqueezeNet (78.07%). Demonstrated improved robustness to noisy labels. | Vanishka Kadian et al., 2023 Third International Conference on Secure Cyber Computing and Communication (ICSCCC), IEEE. | Abhay | ||||||||||||||||||
24 | 23 | The development of an efficient artificial intelligence-based classification approach for colorectal cancer response to radiochemotherapy: deep learning vs. machine learning | In this study, seven artificial intelligence models including decision tree, K-nearest neighbors, Adaboost, random forest, Gradient Boosting, multi-layer perceptron, and convolutional neural network were implemented to detect patients responder and non-responder to radiochemotherapy. | For classifying CRC response to radiochemotherapy, the data (GSE45404) was extracted from the NCBI Gene Expression Comprehensive (GEO) web resource (https://www.ncbi.nlm.nih.gov/geo/) based on different ML models by considering the most applicable features that were obtained from feature selection tools. In this study, the data was split into a training set (80%) and a testing set (20%). | Methods In this study, seven machine learning and deep learning models including DT, KNN, AdaBoost, RF, Gradient Boosting, MLP, and CNN were used for predicting CRC based on outputs of three feature selection strategies including MI, F-classif, and Chi-Square. | The results of this study confirm that random forest, Gradient Boosting, decision tree, and K-nearest neighbors provided more accurate results in terms of accuracy, by 93.8%. Moreover, Among the feature selection methods, mutual information and F-classif showed the best results, while Chi-Square produced the worst results. | Bahrambanan F, Alizamir M, Moradveisi K, Heddam S, Kim S, Kim S, Soleimani M, Afshar S, Taherkhani A. The development of an efficient artificial intelligence-based classification approach for colorectal cancer response to radiochemotherapy: deep learning vs. machine learning. Sci Rep. 2025 Jan 2;15(1):62. doi: 10.1038/s41598-024-84023-w. PMID: 39748016; PMCID: PMC11696929. | Ankita | ||||||||||||||||||
25 | 24 | Transfer Learning Based Model for Colon Cancer Prediction Using VGG16 | This research work suggests a transfer learning based framework for the colon cancer prediction. This framework is planned on the basis of VGG16 and CNN in colon cancer prediction. | The dataset comprises 20,280 images, with 10,284 of them having labels for specific cell types such as fibroblast, inflammatory, epithelial, and others. The remaining 9,996 images lack cell type labels. The distribution of cell types in the labelled subset shows that epithelial constitutes 42%, inflammatory comprises 26%, fibroblast represents 18%, and others account for 14% of the labelled images. The images are categorized into two classes: "Cancerous" and "Non- Cancerous." Approximately 33% of the images belong to the "Cancerous" category, while the remaining 67% are labelled as "Non-Cancerous. | The transfer learning model is applied for diagnosing colon cancer. TL is the combination of VGG16 and CNN. The VGG16 is used as the base model over which CNN model is used for the training. | The colon cancer prediction model is the combination of CNN and VGG16. The VGG16 is the based model and convolution layer will extract features for the classification. The proposed model is implemented in python and results is compared with CNN. It is analysed that proposed model achieve approx. 6 percent high accuracy in comparison with others. | S. Koppad, A. B and A. Acharjee, "Transfer Learning Based Model for Colon Cancer Prediction Using VGG16," 2023 IEEE 5th International Conference on Cybernetics, Cognition and Machine Learning Applications (ICCCMLA), Hamburg, Germany, 2023, pp. 615-620, doi: 10.1109/ICCCMLA58983.2023.10346705. | Ankita | ||||||||||||||||||
26 | 25 | SIGMO: Sigmoid Colon Cancer Prediction Using Machine Learning and Deep Learning | To develop SIGMO, a hybrid ML and DL-based system for accurate and real-time sigmoid colon cancer prediction using integrated clinical and medical imaging data. | Structured clinical data (age, lab results, genetic data) and unstructured imaging data (CT scans, histopathological images) | CNN (ResNet50, VGG16) for feature extraction, Autoencoders for feature selection and dimensionality reduction, XGBoost for classification, SMOTE for class balancing, Explainable AI (Grad-CAM, SHAP), Kubernetes-based real-time deployment | Achieved 92.8% accuracy, 89.6% precision, 94.2% recall, 91.9% F1-score, and 95% AUC-ROC. Outperformed existing systems significantly. Combines clinical and imaging data for best performance. | S. Dhivya Dharshini and K. Anand, 2025 International Conference on Frontier Technologies and Solutions (ICFTS). | Abhay | ||||||||||||||||||
27 | 26 | Colorectal Cancer Detected by Machine Learning Models Using Conventional Laboratory Test Data | This study aimed to determine models for CRC identification that involve minimally invasive, affordable, portable, and accurate screening variables. | This retrospective study used data from patients with CRC and healthy individuals who visited hospital between July 2017 and June 2018. All data were retrieved from the electronic medical record (EMR) system. | Five machine learning models (eg, logistic regression, random forest, k-nearest neighbors, SVM, and naïve Bayes) were used to identify CRC. | The logistic regression model and the SVM model had the highest diagnostic values among the five machine learning models to discriminate patients with CRC from healthy people, with AUCs of 0.865 (95% confidence interval [CI]: 0.857-0.877) and 0.865 (95% CI: 0.857-0.874), respectively. The logistic regression model had a mean sensitivity of 89.5% (95% CI: 88.0%-90.3%), slightly lower than that of the SVM model, whose mean sensitivity was 90.1% (95% CI: 88.0%-91.4%).The proposed model discriminated CRC patients from healthy individuals with an AUC of 0.849 (0.840 to 0.860), a sensitivity of 88.3% (87.4% to 90.3%), and a specificity of 81.5% (79.4%-83.4%). This model had better performance for diagnosing colon cancer than for rectal cancer, and better performance for diagnosing late-stage than for early-stage CRC. This model had the best AUC (0.905 [0.889-0.929]) in discriminating patients with late colon cancer from healthy individuals among CRC patients with different disease sites and stages | Li, H., Lin, J., Xiao, Y., Zheng, W., Zhao, L., Yang, X., Zhong, M., & Liu, H. (2021). Colorectal cancer detected by machine learning models using conventional laboratory test data. Technology in Cancer Research & Treatment, 20. https://doi.org/10.1177/15330338211058352 | Ankita | ||||||||||||||||||
28 | 27 | Colon Cancer Detection Based on Structural and Statistical Pattern Recognition | To develop a novel hybrid model combining structural and statistical pattern recognition techniques for automatic colon cancer detection using histopathological images. | 113 colon tissue images (64 cancerous, 49 normal) | Structural pattern recognition (tissue graph, query graph), statistical feature extraction (GLCM features), classifiers: Multilayer Perceptron (MLP), Sequential Minimal Optimization (SMO), Bayesian Logistic Regression (BLR), K-star using WEKA | MLP achieved highest accuracy of 83.33% with kappa statistic of 0.625, best performance compared to SMO, K-star, and BLR. MLP also showed lowest error rates in all metrics. | Beema Akbar et al., 2015 2nd International Conference on Electronics and Communication Systems (ICECS), IEEE. | Abhay | ||||||||||||||||||
29 | 28 | Lung and Colon Cancer Detection Using a Deep AI Model | This research introduces a novel deep learning model for efficient lung and colon cancer detection, aiming to address the limitations of existing computationally intensive models. | Dataset Description This research uses the lung and colon cancer histopathological images LC25000 dataset . This dataset comprises two main categories of cancer cells: colon adenocarcinoma and benign colon tissue, and lung adenocarcinoma, lung squamous cell carcinoma, and benign lung tissue. The original LC25000 dataset includes 750 lung tissue samples, comprising 250 adenocarcinoma, 250 squamous cell carcinomas,and 250 benign tissue samples. The dataset includes 500 colon tissue samples, with 250 adenocarcinoma and 250 benign tissue samples. These photos were then augmented with techniques like rotation and flipping, resulting in a collection of 5000 images per class and totaling 25000 images for lung and colon cancers. | Utillizes a 1D convolutional neural network enhanced with Squeeze-and-Excitation layers | This method achieves perfect scores in all evaluation metrics, i.e., 100% accuracy rate, precision, sensitivity, and F1-score, showcasing its versatility in detecting lung and colon cancer | Shahadat, N., Lama, R., & Nguyen, A. (2024). Lung and colon cancer detection using a deep AI model. Cancers, 16(22), 3879. https://doi.org/10.3390/cancers16223879 | Ankita | ||||||||||||||||||
30 | 29 | Deep Learning Predictive Model for Colon Cancer Patient using CNN-based Classification | In this research, CNN models are employed to analyse imaging data of colon cells | Kaggle.com was used to gather the dataset. There are 25000 images in the dataset. The images are 768 x 768 pixels in resolution and JPEG format. In the dataset, there are two classes, i.e. 1) Colon adenocarcinoma (cancerous). 2) Colon benign tissue (not cancerous). Of all the images in the dataset, 12,500 images are of colon cancer cells | They are using the Transfer learning model MobileNetV2. The process contains two CNN layers, Max Pooling, and average pooling. The image data goes through a number of preprocessing steps to give a better classification outcome. The performance of the model is evaluated based on the confusion matrix. on its features. In this system, two convolutional layersin the CNN model are used where each convolutional layerused convolutional 2D. In both convolutional 2D layers, 'Reluactivation' is utilized. For complete connectivity, two DenseLayers are used. 'Relu activation' for the first dense layer and 'Sigmoid activation' for the second dense layer is used. Asidefrom these layers, there are several hidden layers, as well as aninput layer. In this study, two pooling layers: Max Pooling 2Dand Average Pooling 2D, are implemented.Finally, forthe classification of image data MobileNetV2 classifier is used. | It's found that the accuracy of the max pooling and average pooling layers is 97.49% and 95.48%, respectively. And MobileNetV2 outperforms the other two models with the most remarkable accuracy of 99.67% with a data loss rate of 1.24. | Tasnim, Z., Chakraborty, S., Shamrat, F. M. J. M., Chowdhury, A. N., Nuha, H. A., Karim, A., Zahir, S. B., & Billah, M. M. (2021). Deep Learning Predictive Model for Colon Cancer Patient using CNN-based Classification. International Journal of Advanced Computer Science and Applications, 12(8). https://doi.org/10.14569/ijacsa.2021.0120880 | Ankita | ||||||||||||||||||
31 | 30 | An Efficient Deep Learning Approach for Colon Cancer Detection | Proposes a new lightweight deep learning approach based on a Convolutional Neural Network (CNN) for efficient colon cancer detection. | Worked on the LC25000 Lung and colon histopathological image dataset [21]. This dataset contains 10,000 histopathological images of colon tissue classified into two classes such as benign colon (5000 images) and colon adenocarcinomas (5000 images). The size of each image is 768 × 768 pixels in JPEG format. We divided the data into 6400 for train, 1600 for validation and 2000 for test. | In our method, the input histopathological images are normalized before feeding them into our CNN model, and then colon cancer detection is performed. The efficiency of the proposed system is analyzed with publicly available histopathological images database and compared with the state-of-the-art existing methods for colon cancer detection | The result analysis demonstrates that the proposed deep model for colon cancer detection provides a higher accuracy of 99.50%, which is considered the best accuracy compared with the majority of other deep learning approaches. Because of this high result, the proposed approach is computationally efficient. | Sakr, A. S., Soliman, N. F., Al-Gaashani, M. S., Pławiak, P., Ateya, A. A., & Hammad, M. (2022). An efficient deep learning approach for colon cancer detection. Applied Sciences, 12(17), 8450. https://doi.org/10.3390/app12178450 | Ankita | ||||||||||||||||||
32 | 31 | LaCK: Lung Cancer Classification and Detection using Convolutional Neural Network-based Gated Recurrent Unit Neural Network Model | To develop the LaCK model combining CNN and GRU for accurate lung cancer classification, focusing on Non-Small Cell Lung Cancer (NSCLC). | Lung Cancer Dataset from Kaggle, 284 patient records with 16 clinical attributes | CNN with 5 convolutional and pooling layers, 5 GRU units, Adam optimizer, Sigmoid activation, comparison with Logistic Regression and SVC models | LaCK model achieved 98% accuracy, outperforming SVC (72%) and Logistic Regression (68%) in precision, recall, and F1-score. | Machakanti Navya Thara et al., 2024 Asia Pacific Conference on Innovation in Technology (APCIT). | Abhay | ||||||||||||||||||
33 | 32 | Biomedical Image Analysis for Colon and Lung Cancer Detection Using Tuna Swarm Algorithm With Deep Learning Model | To develop the BICLCD-TSADL system combining Gabor filtering, GhostNet, and Echo State Network (ESN) for accurate lung and colon cancer detection using biomedical images. | LC25000 dataset (colon and lung cancer histopathological images) | Gabor filtering, GhostNet feature extraction, AFAO-based hyperparameter tuning, Tuna Swarm Algorithm (TSA), ESN classifier, extensive comparative analysis | Achieved 99.33% accuracy, 98.31% precision, 98.31% recall, and F1-score of 98.31%. Outperformed state-of-the-art models like ResNet50 and Faster R-CNN. | Marwa Obayya et al., IEEE Access, 2023. | Abhay | ||||||||||||||||||
34 | 33 | Methodology for Automatic Diagnosing Colon and Lung Carcinoma Using TSA-PSODL Algorithm | To propose an automatic colon and lung cancer diagnosis system using TSA-PSODL algorithm integrating LoG preprocessing, InceptionNetV5, and ESN for early and accurate cancer detection. | LC25000 dataset (colon and lung cancer histopathological images) | Laplacian of Gaussian (LoG) filtering, InceptionNetV5 feature extractor, Adaptive Fuzzy Adam Optimizer (AFAO), Tuna Swarm Algorithm (TSA), Particle Swarm Optimization (PSO), Echo State Network (ESN) classifier | Achieved 99.44% accuracy, 98.41% precision, recall, and F1-score. Outperformed ResNet50, CNN, Faster CNN, and other contemporary models. | Janani V G et al., 2025 2nd International Conference on Trends in Engineering Systems and Technologies (ICTEST). | Abhay | ||||||||||||||||||
35 | 34 | Evaluation of a Deep Neural Network for Automated Classification of Colorectal Polypson Histopathologic Slides | To evaluate the performance and generalizability of a deep neural network for colorectal polyp classification on histopathologic slide images using a multi-institutional data set. | To evaluate the performance and generalizability of a deep neural network for colorectal polyp classification on histopathologic slide images using a multi-institutional data set. | In this study, we implementedthedeepresidualnetwork(ResNet),aneuralnetworkarchitecture that significantly outperformed all other models on the ImageNet andCommonObjectsinContext image recognition benchmarks.24 For model training, we applied a sliding window method to the 3848variable-size regions of interest labeled by pathologists in the training set, extracting approximately 7000fixed-size 224 × 224-pixelpatchesperpolyptype.Then,weinitialized ResNet with the MSRA(Microsoft Research Asia) weight initialization11 and trained the neural network for 200epochswithaninitial learning rate of 0.001, which decayed by a factor of 0.9 every epoch. | For the internal evaluation on the157slides with ground truth labels from 5 pathologists, the deep neural network had a mean accuracy of 93.5%(95%CI,89.6%-97.4%)comparedwithlocal pathologists’ accuracy of 91.4% (95% CI, 87.0%-95.8%). Ontheexternaltestsetof238slideswith ground truth labels from 5 pathologists, the deep neural network achieved an accuracy of 87.0% (95%CI,82.7%-91.3%),whichwascomparablewithlocalpathologists’accuracy of86.6%(95%CI, 82.3%-90.9%). | Wei JW, Suriawinata AA, Vaickus LJ, et al. Evaluation of a Deep Neural Network for Automated Classification of Colorectal Polyps on Histopathologic Slides. JAMA Netw Open. 2020;3(4):e203398. doi:10.1001/jamanetworkopen.2020.3398 | Ankita | ||||||||||||||||||
36 | 35 | Deep Learning Based Colorectal Cancer (CRC) Tumors Prediction | The expected outcome of the research is to perfectly classify the tumors from histopathological images and predict CRC from affected tumors images. | Dataset was collected from Borkowski et al. which contains 25000 histopathological image dataset of Lung cancer and Colon cancer.a. In the colon cancer image dataset colon adenocarcinoma images was 5,000 and benign colonic tissue images was 5,000. The dimension (height and width) of each image dataset was 768px (Figure 4). Our selected two types of images were: a)Adenocarcinoma b) Benign colonic tissue | Model was classified using Support Vector Machine (SVM), Decision Tree, Random Forest, Ensemble Method, CNN and Transfer Learning methods. SVM is a supervised machine learning algorithm. For classifying tumor histopathological images we used SVM. Decision Tree is a simple decision making diagram. But for prediction or classification task decision trees are more powerful tools. Applying decision tree technique on tumor images we classified tumors and predict. Using a number of Decision Trees the Random Forest model is created. Using Random Forest model we can classify our data (images) more specifically. To better understand the tumor image we used Ensemble methods, Ensemble methods technique is the combination of several base model of machine learning technique. And this method gave us optimal and best predictive solution. Convolutional Neural Network (CNN) is the best neural network technique for Deep Learning based work. CNN technique was used for analyzing complex data | Got the best classification accuracy from the XGBoost classifier and its accuracy was 99%. And others classifiers accuracy was within 98% to 89%. | Mohalder, R. D., & Talukder, K. H. (2021). Deep Learning Based Colorectal Cancer (CRC) Tumors Prediction. 2022 13th International Conference on Computing Communication and Networking Technologies (ICCCNT). https://doi.org/10.1109/icccnt51525.2021.9579847 | Ankita | ||||||||||||||||||
37 | 36 | Li-SegPNet: Encoder-Decoder Mode Lightweight Segmentation Network for Colorectal Polyps Analysis | To develop Li-SegPNet, a lightweight encoder-decoder deep learning model for efficient colorectal polyp segmentation with cross-dimensional attention and improved skip connections. | Kvasir-SEG, CVC-ClinicDB (training/testing); Hyper-Kvasir, EndoTect 2020 (cross-dataset testing) | Li-SegPNet architecture with modified triplet attention (MTA), pre-trained ResNet-50 backbone, attention-skip connections, data augmentation, Dice loss function | Dice Score: 0.9058 (Kvasir-SEG), 0.9372 (CVC-ClinicDB); mIoU: 0.8800 (Kvasir-SEG), 0.8969 (CVC-ClinicDB); Dice Score: 0.9652 (Hyper-Kvasir), 0.8620 (EndoTect 2020); Outperformed 9 SOTA models with fewer parameters and better generalization. | Pallabi Sharma et al., IEEE Transactions on Biomedical Engineering, Vol. 70, No. 4, April 2022. | karabi Maam | ||||||||||||||||||
38 | 37 | The Role of Deep Learning in Diagnosing Colorectal Cancer | To review the applications of deep learning in diagnosing colorectal cancer across histopathology, endoscopy, CT colonography, and serologic screening. | Various datasets including whole slide images (WSI), colonoscopy images, CT colonography scans, capsule endoscopy, and public clinical datasets | Deep Learning (CNN, semi-supervised and supervised learning), endoscopic real-time analysis, histopathology image classification, CT colonography interpretation, capsule endoscopy, serologic screening | Sensitivity up to 99.12% in histopathology diagnosis, ADR increased by up to 80% using DL in colonoscopy, real-time polyp detection accuracy up to 96%, optical biopsy sensitivity up to 90%, DL-enabled CT colonography AUROC up to 0.945, capsule endoscopy AUROC up to 0.99. | Dimitrios Bousis et al., Gastroenterology Review, July 2023. | karabi maam | ||||||||||||||||||
39 | 38 | Colon cancer diagnosis by means of explainable deep learning | The study aimed to develop a method for the early detection of adenocarcinoma in colon tissue using histological images,leveraging deep learning techniques to improve diagnostic accuracy. | Used the LC25000 dataset from Kaggle,containing 5,000 histological images each of benign colon tissue and adenocarcinoma (binary classification).Applied data augmentation (rotation,scaling, flipping) to increase the dataset size tenfold (total:10,000 images). | Tested seven models: ResNet50, DenseNet, VGG19,Standard_CNN, Inception-V3,EfficientNet, and MobileNet. o Training parameters:50 epochs, batch size of 8, learning rate of 0.0001, and image dimensions of 224×224×3.o Loss function: Binary cross-entropy for binary classification.Grad-CAM, Score-CAM, and FastScore-CAM: Generated heatmaps to visualize regions of interest (ROIs) contributing to predictions. Structural Similarity Index Measure (SSIM): Quantified the similarity between heatmaps to assess model robustness (MR-SSIM). | Quantitative Performance: Top-performing models: MobileNet (99.9% accuracy), Inception-V3 (99.8%), and EfficientNet (99.6%). Poor performers: VGG19 (49.9% accuracy),Standard_CNN (50%), ResNet50 (65.2%), and DenseNet (68.8%). MobileNet achieved near- perfect metrics (precision, recall, F-measure, AUC > 0.99) and minimal loss (0.045). 2. Qualitative Results (Explainability): MobileNet: Produced accurate heatmaps localizing cancerous cell clusters (Fig. 6), validated by high MR-SSIM scores (0.79 for Grad- CAM/Score-CAM). EfficientNet: Generated repetitive, non-specific heatmaps (Fig. 5), failing to adapt to image variations. Inception-V3: Could not generate heatmaps, indicating poor pattern recognition for explainability. | Di Giammarco, M., Martinelli, F., Santone, A. et al. Colon cancer diagnosis by means of explainable deep learning. Sci Rep 14, 15334 (2024). https://doi.org/10.1038/s41598-024-63659-8 | Karabi Ma'am | ||||||||||||||||||
40 | 39 | Prediction of the histology of colorectal neoplasm in white light colonoscopic images using deep learning algorithm | The primary objective of the study was to develop a computer-aided diagnostic (CAD) system using deep learning algorithms to predict the histology of colorectal neoplasms from white light colonoscopic images. The system aimed to classify lesions into four categories: Normal (no adenoma) Adenoma with low-grade dysplasia (A-LGD) Adenoma with high-grade dysplasia (A-HGD) Adenocarcinoma (CA | Retrospectively collected 3,400 colonoscopic images from two tertiary hospitals (KUMC and HYUMC). Images were labeled into four classes based on histopathology reports. Excluded poor-quality images (blurred, out-of-focus, or with artifacts). Standardized images by cropping and resizing to 480×480 pixels. | Employed three Convolutional Neural Network (CNN) architectures: Inception-v3 ResNet-50 DenseNet-161 Used transfer learning (pre-trained on ImageNet) and fine-tuned for colonoscopic images. Applied ensemble learning to combine predictions from the three models for improved accuracy. | Ensemble CNN-CAD achieved: Sensitivity: 90.65% Specificity: 97.55% F1-score: 0.9055 Mean diagnostic time: 0.12 seconds per image (vs. 7.96s for experts). Outperformed trainees and matched/surpassed expert endoscopists. Multi-Center Test (HYUMC Dataset): Performance slightly decreased but remained robust: Sensitivity: 77.25% Specificity: 92.42% F1-score: 0.7681 Still comparable to experts (72.38% sensitivity). Comparison with Endoscopists: Experts: 85.00% sensitivity, 95.00% specificity. Trainees: 77.98% sensitivity, 92.63% specificity. CNN-CAD was faster and more consistent than human observers. Visual Explanations (CAM): Heatmaps highlighted lesion regions, improving interpretability. | Choi, S.J., Kim, E.S. & Choi, K. Prediction of the histology of colorectal neoplasm in white light colonoscopic images using deep learning algorithms. Sci Rep 11, 5311 (2021). https://doi.org/10.1038/s41598-021-84299-2 | Karabi Ma'am | ||||||||||||||||||
41 | 40 | A promising deep learning assistive algorithm for histopathological screening of colorectal cancer | To develop a deep learning-based assistive tool for histopathological screening of colorectal cancer to improve detection and classification efficiency in clinical practice. | Whole Slide Images (WSI) from 294 colorectal specimens (Singapore General Hospital and TCGA) | Faster-RCNN with ResNet-101 backbone for gland segmentation, Gradient Boosted Decision Tree (GBDT) for slide classification, data augmentation | Achieved AUC of 91.7%, sensitivity 97.4%, specificity 60.3%. Designed as a screening tool to minimize false negatives, assisting pathologists in triage and workflow optimization. | Cowan Ho et al., Scientific Reports, 2022. | karabi maam | ||||||||||||||||||
42 | 41 | Deep learning for multi-class semantic segmentation enables colorectal cancer detection and classification in digital pathology images | Develop an AI-based method for multi-class semantic segmentation of 14 tissue compartments in H&E-stained whole-slide images (WSIs) of colorectal cancer (CRC). Compare the performance of state-of-the-art loss functions (Categorical Cross-entropy, Focal, Bi-tempered, and Lovasz-softmax) for histopathology image segmentation. Create a computer-aided diagnosis (CAD) system to classify colon biopsies into four clinically relevant categories: High-risk (tumor and high-grade dysplasia) Low-grade dysplasia Hyperplasia Benign conditions Validate the segmentation model on multi-centric and publicly available datasets (GLAS and CRAG). | Segmentation Dataset (Dseg): 79 WSIs from 5 medical centers (Netherlands and Germany). Annotated with 14 tissue classes (e.g., normal glands, tumor, stroma, lymphocytes). Scanned using three different scanners (Pannoramic, IntelliSite, NanoZoomer). Biopsy Classification Dataset (Dcls): 1,054 colon biopsies from Cannizzaro Hospital (Italy). Labeled into four diagnostic categories. | Segmentation Model Architecture: Modified U-Net with residual skip connections and nearest-neighbor upsampling. Training: Input: 512×512 px patches at 1 μm/px resolution. Augmentation: Flipping, rotation, elastic deformation, stain variation. Optimizer: Adam with adaptive learning rate. Loss Functions Tested: Categorical Cross-entropy (CC) Focal Loss Bi-tempered Loss Lovasz-softmax Loss Biopsy Classification Features Extracted: Normalized histogram of tissue classes. Number, average, min, and max size of tumor clusters. Classifier: Random Forest with 1,000 decision trees (5-fold cross-validation). | Segmentation Performance Best Loss Function: Lovasz-softmax (mean Dice = 0.72), closely followed by Bi-tempered (0.71). Multi-Centric Robustness: Model performed well across different centers but struggled with dark-stained slides. Biopsy Classification AUC Scores: High-grade dysplasia/tumor: 0.87 Low-grade dysplasia: 0.82 Hyperplasia: 0.89 Benign: 0.79 Overall Accuracy: Quadratic weighted kappa = 0.91. | Bokhorst, JM., Nagtegaal, I.D., Fraggetta, F. et al. Deep learning for multi-class semantic segmentation enables colorectal cancer detection and classification in digital pathology images. Sci Rep 13, 8398 (2023). https://doi.org/10.1038/s41598-023-35491-z | Karabi Ma'am | ||||||||||||||||||
43 | 42 | A trial deep learning-based model for four-class histologic classification of colonic tumor from narrow band imaging | The study aimed to: Develop a deep learning (DL)-based model for the four-class histologic classification of colonic tumors using narrow band imaging (NBI). The classes included: Low-grade dysplasia (LGD) High-grade dysplasia or mucosal carcinoma (HGD) Superficially invasive submucosal carcinoma (SMs) Deeply invasive submucosal carcinoma (SMd). | Collected 1,390 NBI images from 210 lesions (53 LGD, 120 HGD, 20 SMs, 17 SMd). Images were manually partitioned into lesion and background regions, and 598,801 patches (128×128 pixels) were extracted. Patches with excessive blackout (>10% dark pixels) or halation (>5% bright pixels) were excluded. Patches were classified into 7 categories: BG (background), LGD, HGD, SMs, SMd, BG-oof (out-of-focus background), and L-oof (out-of-focus lesion). | Used ResNet50 (a residual convolutional neural network) without pretraining. Implemented three-fold cross-validation to reduce bias. Hyperparameters: Adam optimizer, cross-entropy loss, 50 epochs, batch size 256, learning rate 0.00005 | Patch-Level Performance: Validation accuracies: LGD (87.6%), HGD (95.7%), SMs (90.7%), SMd (92.9%). Background and out-of-focus patches also showed high accuracy (>90%). Image-Level Performance: Overall accuracy: 98.6% (1,371/1,390 images correctly classified). Precision/F1-scores: LGD: 97.3% / 97.8% HGD: 99.2% / 99.1% SMs: 100% / 98.2% SMd: 96.7% / 98.0%. Misclassifications were rare but occurred in HGD vs. SMd (e.g., background mislabeled as lesion). | Shimizu, T., Sasaki, Y., Ito, K. et al. A trial deep learning-based model for four-class histologic classification of colonic tumor from narrow band imaging. Sci Rep 13, 7510 (2023). https://doi.org/10.1038/s41598-023-34750-3 | Karabi Ma'am | ||||||||||||||||||
44 | 43 | Predicting Colorectal Cancer Using Machine and Deep Learning Algorithms: Challenges and Opportunities | To systematically review AI-based ML and DL techniques applied to colorectal cancer modeling, prediction, and early diagnosis, including challenges and future opportunities. | Multiple datasets: gene expression profiles, histopathology images, colonoscopy images, public datasets like TCGA, LC25000, NCT-CRC-HE-100K, among others | Machine Learning (SVM, RF, KNN, Logistic Regression, AdaBoost, Decision Trees), Deep Learning (CNN, UNET, ResNet, YOLOv3, VGG16, Autoencoders) | ML and DL algorithms showed high predictive accuracy in multiple studies; DL generally outperformed ML in image-based tasks. Identified gaps include small datasets, class imbalance, and limited generalization across datasets. | Alboaneen et al., Big Data Cogn. Comput. 2023, 7(2), 74 | Karabi maam | ||||||||||||||||||
45 | 44 | Assessment of a Deep-Learning System for Colorectal Cancer Diagnosis Using Histopathology Images | To assess and propose a deep-learning system using histopathology images for colorectal cancer diagnosis, including a new model | NCT-CRC-HE-100K dataset (100,000 image patches), Cancer Grade dataset (139 images) | CNN models (AlexNet, GoogLeNet, ResNet, MobileNet, Xception, DenseNet, ResNeXt), Xception+ (proposed), LIME for explainability, PyTorch framework | achieved 99.37% accuracy for cancer detection and 98.22% for tissue classification, outperforming all other models. Cancer grade classification achieved 94.48% accuracy | Purna Kar & Sareh Rowlands, American Journal of Computer Science and Technology, 2024 | Karabi maam | ||||||||||||||||||
46 | 45 | Machine Learning-based Lung and Colon Cancer Detection using Deep Feature Extraction and Ensemble Learning | To develop a hybrid machine learning model combining deep feature extraction, ensemble learning, and high-performance filtering for lung and colon cancer detection. | LC25000 dataset (lung and colon histopathological images) | Deep feature extraction using VGG16, VGG19, MobileNet, DenseNet169, DenseNet201, high-performance filtering, ML algorithms (SVM, RF, MLP, XGB, LGB), ensemble learning | Achieved 99.05% accuracy for lung cancer, 100% for colon cancer, and 99.30% for combined lung and colon cancer detection. Outperformed prior models significantly. | Md. Alamin Talukder et al., Expert Systems with Applications, 2022. | Karabi Maam | ||||||||||||||||||
47 | 46 | Enhancing Deep Learning-Based Models for Early Detection of Colon Cancer | Develop an accurate medical diagnostic support system for early detection of colon cancer using histopathological images. | Datasets: LC25000: 10,000 images (5,000 adenocarcinoma, 5,000 benign). NCT-CRC-HE-100K: 27,000 images (9 classes, 3,000 per class). | Feature Extraction: Applied 2D-DWT with biorthogonal wavelets to decompose images into sub-bands (LL, LH, HL, HH). Extracted statistical features (mean, standard deviation) from wavelet coefficients. Classification: Multi-layered CNN with convolutional, pooling, dropout, and dense layers. Sigmoid activation for binary classification (LC25000) and softmax for multi-class (NCT-CRC-HE-100K). | 4. Performance Metrics Primary Metric: Accuracy (99.3% for LC25000, 96.33% for NCT-CRC-HE-100K). Secondary Metrics: Precision (99%), Recall (99.9%), F1-score (99%). Sensitivity (99.9%), Specificity (99.9%), G-Mean (99.9%). LC25000 Dataset: With DWT: 99.3% accuracy, 505/508 adenocarcinoma correctly classified. Without DWT: 97% accuracy. NCT-CRC-HE-100K Dataset: With DWT: 96.33% accuracy. Without DWT: 95.54% accuracy. Comparative Advantage: Outperformed prior models (e.g., VGG16: 93.81%, Random Forest: 83%). | Elatrash, E. A., Ghannam, N. E., Zaki, M. S., & El-Sayed, R. S. (2025). Enhancing deep learning-based models for early detection of colon cancer. IAENG International Journal of Computer Science, 52(4), 1148–1158. | Karabi Ma'am | ||||||||||||||||||
48 | 47 | Deep learning ensemble approach with explainable AI for lung and colon cancer classification using advanced hyperparameter tuning | To develop a novel deep learning framework combining Xception and MobileNet architectures for improved classification of lung and colon cancer histopathological images. To enhance feature extraction, model robustness, and reduce overfitting through ensemble learning. To integrate Gradient-weighted Class Activation Mapping (Grad-CAM) for explainable AI, providing visual insights into the model's decision-making process. To achieve high diagnostic accuracy and generalizability for clinical applications. | Dataset: LC25000 dataset with 25,000 histopathological images (5 classes: Colon Adenocarcinoma, Benign Colonic Tissue, Lung Adenocarcinoma, Lung Squamous Cell Carcinoma, Benign Lung Tissue). Split into training (80%), validation (10%), and testing (10%) sets. | Model Architecture: Ensemble Model: Combines Xception and MobileNet for feature extraction. Feature Concatenation: Outputs of both models merged and fed into dense layers (1024, 512, 256 neurons with ReLU activation). Classification Layer: Softmax activation for multi-class prediction. Training: Optimizer: Adamax (learning rate = 0.001, adjusted dynamically). Loss Function: Categorical cross-entropy. Early stopping to prevent overfitting. Explainability: Grad-CAM for visualizing influential regions in images. | Performance Metrics Accuracy: 99.44% on the test set. Precision: Ranged from 0.9559 (Lung SCC) to 1.0000 (Colon ACA, Benign Lung). Recall: Lowest for Lung ACA (0.9508), highest for Benign Lung (1.0000). F1-Score: Consistently high across all classes (balanced precision and recall). AUC-ROC: 1.00 for all classes, indicating excellent discriminative performance. Confusion Matrix: Minimal misclassifications, primarily between Lung SCC and Lung ACA. The ensemble model achieved state-of-the-art performance (99.44% accuracy). Grad-CAM provided interpretable heatmaps, aiding clinical trust and validation. Misclassifications were rare but occurred between visually similar classes (e.g., Lung SCC and Lung ACA). The model outperformed existing methods | Vanitha, K., R, M.T., Sree, S.S. et al. Deep learning ensemble approach with explainable AI for lung and colon cancer classification using advanced hyperparameter tuning. BMC Med Inform Decis Mak 24, 222 (2024). https://doi.org/10.1186/s12911-024-02628-7 | Karabi Ma'am | ||||||||||||||||||
49 | 48 | Application of Artificial Intelligence in diagnosis and treatment of colorectal cancer:A novel prospect | To review the latest advancements in AI for diagnosing and treating colorectal cancer (CRC). To evaluate AI's role in improving early detection, diagnostic accuracy, and treatment efficacy. To explore AI applications in endoscopy, non-invasive screening, histopathology, radiology, surgery, chemoradiotherapy, and targeted therapy. | ------ | ----- | AI improved early CRC detection (e.g., higher ADR/PDR in colonoscopy). AI-assisted pathology reduced misdiagnosis rates (e.g., 94% accuracy in histopathology). Radiomics enhanced tumor segmentation and treatment response prediction. Robotic surgery (e.g., da Vinci) improved precision but had high costs. AI models (e.g., DL) showed promise in predicting chemoradiotherapy outcomes. | Yin Z, Yao C, Zhang L and Qi S (2023) Application of artificial intelligence in diagnosis and treatment of colorectal cancer: A novel Prospect. Front. Med. 10:1128084. doi: 10.3389/fmed.2023.1128084 | Karabi Ma'am | ||||||||||||||||||
50 | 49 | A Deep Learning Approach for Detecting Colorectal Cancer via Raman Spectra | Primary Objective: To develop a deep learning-based method for detecting colorectal cancer (CRC) using Raman spectroscopy, aiming to improve the accuracy and efficiency of intraoperative and pathological diagnoses. Secondary Objectives: Collect and preprocess a large Raman spectroscopy dataset from colorectal cancer patients. Design a one-dimensional residual convolutional neural network (1D-ResNet) to classify tumor and normal tissues. Compare the performance of the proposed deep learning model with traditional machine learning methods. Interpret the Raman spectral peaks identified by the model to understand biochemical differences between tumor and normal tissues. | Data Collection: Raman spectra were collected from 26 CRC patients (6 grade I, 12 grade II, 8 grade III) using a Renishaw inVia Raman spectrometer (385–1545 cm−1−1 range). Dataset included 20,424 spectra (10,866 normal, 9,558 tumor), split into training (80%), validation (10%), and test (10%) sets | Model Architecture (1D-ResNet): Adapted from ResNet-34, with 1D convolutional layers for spectral data. Used residual blocks with batch normalization, ReLU activation, and dropout to prevent overfitting. Ensemble strategy: Combined three variants (original, intensity-enhanced, split spectra) with weighted voting. Comparison Methods: Traditional machine learning models (SVM, Random Forest, XGBoost, LightGBM) were trained and tuned for benchmarking. Segmentation Input Data: 1D Raman spectra vectors (1 × 1024). Feature Extraction: Convolutional kernels (size 1×3) captured local spectral patterns. Residual connections preserved gradient flow in deep layers. Output: Binary classification (tumor vs. normal). | Performance Metrics Metrics: Accuracy: 98.5% (ensemble model). Precision: 0.980, Recall: 0.986, F1-score: 0.983. AUC-ROC: ~0.98 (outperformed SVM: 0.87, Random Forest: 0.88, XGBoost: 0.96). Confusion Matrix: Minimal false positives/negatives. Training Stability: Accuracy and loss curves showed convergence after 200 epochs. High Accuracy: The ensemble 1D-ResNet achieved 98.5% accuracy, surpassing traditional methods (e.g., PCA-LDA: 92.4% in prior work). Key Spectral Peaks: Tumor tissues showed activated regions at 450–1200 cm−1−1 (e.g., DNA, proteins). Normal tissues focused on 800–1000 cm−1−1 (e.g., glucose, polysaccharides). Visualization: Grad-CAM highlighted discriminative Raman shifts, aligning with known biochemical markers | Cao, Z., Pan, X., Yu, H., Hua, S., Wang, D., Chen, D. Z., Zhou, M., & Wu, J. (2022). A deep learning approach for detecting colorectal cancer via Raman spectra. BME Frontiers, *2022*, Article 985356. https://doi.org/10.34133/2022/9872028 | Karabi Ma'am | ||||||||||||||||||
51 | 50 | Detection of Colon Cancer using Deep Learning Techniques / Colon Cancer Detection Using Deep Learning Algorithm | To analyze and improve colon cancer detection using various deep learning algorithms on histopathological images for accurate classification and diagnosis. | Histopathological image data; dataset details not explicitly stated but multiple references cite use of images and microarray data | Deep Learning (DL), CNN, ANN, supervised and unsupervised ML, Feed-forward Neural Networks, Gradient Descent optimization, comparison of literature models (e.g., DNN, ensemble, hybrid methods) | Performance metrics such as accuracy, sensitivity, specificity, and F1-score discussed; no specific numerical results reported in this study—focus is on survey, methodology review, and problem identification | E.N. Srivani & G. Seshikala, 2024 International Conference on Integrated Circuits and Communication Systems (ICICACS), IEEE. | Karabi maam | ||||||||||||||||||
52 | 51 | Colon Cancer Detection of Colonoscopy Images Using CNN | Primary Objective: To develop a reliable framework for early and precise detection of colon cancer using deep learning techniques applied to colonoscopy images. Secondary Objectives: Compare the performance of three CNN architectures (VGG-16, ResNet-50, InceptionV3) for colon cancer classification. Enhance image quality through preprocessing techniques (resizing, contrast enhancement, histogram equalization). Leverage transfer learning to adapt pre-trained models for medical imaging tasks. Improve diagnostic accuracy to aid healthcare professionals in early intervention | ------ | Image Preprocessing: Resizing images to 128×128 pixels. Contrast enhancement via histogram equalization. RGB conversion for standardization. One-hot encoding for multi-class labeling. Deep Learning Models: VGG-16: 16-layer CNN with small (3×3) filters. ResNet-50: 50-layer residual network to address vanishing gradients. InceptionV3: 42-layer model with optimized convolutions. Transfer Learning: Fine-tuning pre-trained models on colonoscopy datasets. Segmentation Approach: The study focuses on image-level classification rather than pixel-wise segmentation. Preprocessing steps (e.g., resizing, contrast adjustment) indirectly aid feature extraction for CNNs. | ResNet-50 outperformed other models due to its residual connections, mitigating gradient issues. Preprocessing (histogram equalization) improved input quality for CNNs. Transfer learning reduced training time and enhanced performance. Comparative Analysis: ResNet-50’s accuracy surpassed prior studies (e.g., 99.8% recall in Hasan et al., 2022). VGG-16 showed moderate performance, while InceptionV3 struggled with the dataset. | Sheth, C. V., & Yadav, R. (2024). Colon cancer detection of colonoscopy images using CNN. International Journal for Research in Applied Science & Engineering Technology (IJRASET), *12*(V), 178-183. https://doi.org/10.22214/ijraset.2024.61465 | Karabi Ma'am | ||||||||||||||||||
53 | 52 | Deep learning for colorectal cancer detection in contrast enhanced CT without bowel preparation: a retrospective, multicentre study | Primary Objective: Develop a deep learning (DL) model to accurately detect colorectal cancer (CRC) in contrast-enhanced CT scans without bowel preparation, addressing the challenge of missed diagnoses by radiologists due to subtle imaging features or lack of bowel preparation. Secondary Objectives: Validate the DL model across multiple datasets (internal, external, and real-world) to ensure generalizability. Compare the DL model’s performance with that of radiologists (both generalists and specialists). Evaluate whether the DL model can improve radiologists’ detection accuracy when used as a computer-aided diagnostic tool. | Datasets: Retrospective multicentre data from 5,220 patients (CRC vs. normal) across four hospitals. Training/Validation: 2,017 patients (1,098 CRC, 919 normal). Test Sets: Internal (213 patients), External 1 (467), External 2 (733), Real-world (1,577) | Modified nnUNet (a self-configuring U-Net for medical segmentation) with an auxiliary classification branch for patch-level CRC/non-CRC classification. Annotations: Colorectum: Semi-automated segmentation (ITK-SNAP) + manual correction. Tumors: Manually annotated by radiologists, reviewed by a CRC specialist. Segmentation Approach: Pixel-level segmentation of CRC regions. Detection threshold: Tumor volume >250 mm³ (determined via Youden Index) | AUC: 0.957–0.994 across all test sets. Sensitivity: 91.7–96.9% (higher than radiologists in standalone mode). Specificity: 95.6–97.4%. Real-world: Detected 2 missed CRC cases by radiologists. Comparison with Radiologists: DL outperformed radiologists in accuracy (97.2% vs. 86.0% in internal test set). DL-assisted radiologists showed improved accuracy (93.4% vs. 86.0%). Subgroup Insights: DL excelled in detecting small tumors (<3 cm) and early-stage (T1) cancers. Highest sensitivity for T4 tumors (100%) and ascending colon (98.1%). | Yao, L., Li, S., Tao, Q., Mao, Y., Dong, J., Lu, C., Han, C., Qiu, B., Huang, Y., Huang, X., Liang, Y., Lin, H., Guo, Y., Liang, Y., Chen, Y., Lin, J., Chen, E., Jia, Y., Chen, Z., . . . Liu, Z. (2024). Deep learning for colorectal cancer detection in contrast-enhanced CT without bowel preparation: A retrospective, multicentre study. The Lancet, *104*(June), 1–11. https://doi.org/10.1016/j.chiom.2024.105183 | Karabi Ma'am | ||||||||||||||||||
54 | 53 | Application of Artificial Intelligence in the diagnosis and treatment of colorectal cancer: a bibliometric analysis, 2004–2023 | Primary Objective: To systematically analyze and visualize the research landscape of AI applications in colorectal cancer (CRC) diagnosis and treatment over two decades (2004–2023). Identify key trends, hotspots, and future directions in AI-driven CRC research. Secondary Objectives: Quantify global contributions (countries, institutions, authors) and collaborations. Highlight research gaps and propose areas for future exploration. | ------ | ------ | Performance Metrics Diagnostic Accuracy: AI-assisted colonoscopy achieved 96% accuracy in polyp detection (Wang et al., 2018). CADx systems matched/exceeded endoscopist performance in polyp characterization. Prognostic Models: Deep learning predicted CRC survival with higher accuracy than traditional method | Sun, L., Zhang, R., Gu, Y., Huang, L., & Jin, C. (2024). Application of Artificial Intelligence in the diagnosis and treatment of colorectal cancer: A bibliometric analysis, 2004–2023. Frontiers in Oncology, 14, 1424044. https://doi.org/10.3389/fonc.2024.1424044 | Karabi Ma'am | ||||||||||||||||||
55 | 54 | Colorectal Cancer Diagnosis with Deep Learning Models | The primary objective of this study is to develop and assess deep learning models for the early and accurate diagnosis of colorectal cancer using histopathological images. The secondary objectives include comparing the performance of five deep learning architectures—CNN, Hybrid CNN-LSTM, AlexNet, VGG-16, and ResNet—to identify the model that achieves the highest accuracy and lowest loss in detecting colorectal cancer. Additionally, the study aims to demonstrate the potential of computational tools in supporting pathologists with histopathological image analysis. | Dataset: Source: Colorectal Histology MNIST dataset from Kaggle, containing 5,000 RGB images of eight colorectal tissue types (e.g., tumor, stroma, lymphoid). | CNN: Custom architecture with ReLU/Softmax activation, Adam optimizer (lr=0.001), and 128 epochs. Hybrid CNN-LSTM: Combined CNN for spatial feature extraction and LSTM for temporal sequence analysis, trained for 72 epochs. Pretrained Models: AlexNet: 8 layers, ReLU activation, 20 epochs. VGG-16: 16 layers (13 convolutional), 20 epochs. ResNet50: 50 layers with identity mapping, 100 epochs. | Accuracy Rates: Hybrid CNN-LSTM: 92.4% (highest). CNN: 90.2%. ResNet: 88%. VGG-16/AlexNet: ~73%. Loss Rates: CNN: 0.0176 (lowest). Hybrid CNN-LSTM: 0.2298. ResNet/AlexNet/VGG-16: 0.23–1.3875. | Taşcı, M. E., Elmi, Z., Albayrak, &. F., Tokat, M. (2024). Colorectal Cancer Diagnosis with Deep Learning Models. In N. Callaos, E. Gaile-Sarkane, N. Lace, B. Sánchez, M. Savoie (Eds.), Proceedings of the 28th World Multi-Conference on Systemics, Cybernetics and Informatics: WMSCI 2024, pp. 92-98. International Institute of Informatics and Cybernetics. https://doi.org/10.54808/WMSCI2024.01.92 | Karabi Ma'am | ||||||||||||||||||
56 | 55 | DEEP LEARNING-BASED COLON CANCER CLASSIFICATION USING PRE-TRAINED CUSTOM CONVOLUTIONAL NEURAL NETWORK WITH HISTOPATHOLOGICAL IMAGES | To develop a custom Convolutional Neural Network (CNN) model for classifying colon cancer using histopathological images. To compare the performance of the proposed custom CNN with pre-trained models (VGG16, ResNet50, InceptionV3) in terms of accuracy, sensitivity, specificity, and F1-score. To improve early detection and staging of colon cancer through deep learning techniques. To address the research gap in applying CNN and transfer learning for predicting colon cancer survival and stage III classification. | Publicly available histopathological image datasets: Colorectal Histology Images Dataset (26,000 images) CRCHistoPhenotypes Dataset (4,000 images) Colorectal Cancer Digital Slide Archive (6,000 whole slide images | Custom CNN: Designed with convolutional, pooling, and fully connected layers, optimized for filter size, stride, and dropout regularization. Pre-trained Models: Fine-tuned versions of VGG16, ResNet50, and InceptionV3. | The custom CNN outperformed pre-trained models, achieving 93.6% accuracy and higher sensitivity/specificity. Demonstrated that custom architectures can be more effective than transfer learning for histopathological image classification. Highlights the potential of deep learning in improving computer-aided diagnosis (CAD) systems for colon cancer. | Sakthipriya, N., Govindasamy, V., Narmadhadevi, R., Prasanth, L., & Kanimozhi, P. (2023). Deep learning-based colon cancer classification using pre-trained custom convolutional neural network with histopathological images. International Journal of Creative Research Thoughts (IJCRT), 11(4), c397-c401. https://www.ijcrt.org. | Karabi Ma'am | ||||||||||||||||||
57 | 56 | Colon Cancer Classification on Histopathological Images using Deep Learning Techniques | Primary Objective: To classify histopathological images of colon tissues into cancerous (malignant) and non-cancerous (benign) categories using deep learning techniques. Secondary Objectives: Improve early detection of colorectal cancer to reduce mortality rates. Compare the performance of Convolutional Neural Networks (CNNs) with traditional methods. Achieve high accuracy in classification to assist pathologists and reduce diagnostic subjectivity. Utilize the LC25000 dataset to train and validate the model. | LC25000 dataset: Contains 25,000 histopathological images (10,000 colon images: 5,000 adenocarcinoma and 5,000 benign). | Convolutional Neural Network (CNN): Used for feature extraction and classification. Training-Validation Split: 80% training, 20% validation. Workflow: Input histopathological images. Preprocess images to remove noise. Train CNN on labeled data. Classify images as malignant or benign | High Accuracy: The CNN model achieved 99.7% accuracy in classifying colon cancer images. Efficiency: The model is resource-efficient and faster than manual diagnosis. Validation: Results were validated using a holdout dataset (20% of images). Graphical Representation: Training vs. validation accuracy/loss curves showed minimal overfitting. | Deiva Nayagam, R., Ramco Institute of Technology, Aarthi, K., Ramco Institute of Technology, Mirra, S., & Ramco Institute of Technology. (2022). Colon Cancer Classification on Histopathological Images using Deep Learning Techniques. International Journal of Engineering Research & Technology (IJERT), 10–10(08), 69–69. https://www.ijert.org/research/colon-cancer-classification-on-histopathological-images-using-deep-learning-techniques-IJERTCONV10IS08017.pdf | Karabi Ma'am | ||||||||||||||||||
58 | 57 | Advanced deep learning for multi-class colorectal cancer histopathology: integrating transfer learning and ensemble methods | Develop an advanced deep learning (DL) framework to improve the classification accuracy of colorectal cancer (CRC) histopathological images. Assist pathologists in early and accurate diagnosis by automating the analysis of multi-class CRC images | A newly released publicly available dataset, named EBHI, contains 5,532 electron microscopy images with a resolution of 2048×1536 pixels, collected from colonoscopic biopsy samples. | Transfer Learning (TL): Fine-tuned pre-trained CNNs (ResNet, Xception, DenseNet, EfficientNet) on ImageNet weights. Ensemble Learning: Combined top-performing models (e.g., EfficientNetB1, EfficientNetV2M) using: Majority Voting Unweighted Averaging Stacking (with Naïve Bayes as meta-learner). Segmentation: Grad-CAM (Gradient-weighted Class Activation Mapping): Visualized regions of interest in histopathology images to interpret model decisions. | Single-Model Performance: EfficientNetB1 achieved the highest accuracy (98.24% at 40x, 97.78% at 400x). EfficientNetV2M performed best at 100x (98.47%). Ensemble Model Performance: EL-Sta5 (stacking with 5 models) achieved the highest accuracy: 99.36% (100x), 99.29% (200x), 98.96% (400x). EL-MV5 (majority voting) and EL-UA5 (unweighted averaging) also outperformed single models. Comparison with State-of-the-Art: Outperformed previous studies (e.g., VGG16: 95.37% vs. EL-Sta5: 99.26% on 200x images). | Ke, Q., Yap, W., Tee, Y. K., Hum, Y. C., Zheng, H., & Gan, Y. (2025). Advanced deep learning for multi-class colorectal cancer histopathology: integrating transfer learning and ensemble methods. Quantitative Imaging in Medicine and Surgery, 15(3), 2329–2346. https://doi.org/10.21037/qims-24-1641 | Karabi Ma'am | ||||||||||||||||||
59 | 58 | A Systematic Review of Deep Learning Techniques in Colon Cancer Screening | To review various deep learning architectures used in colon cancer detection, comparing their performance based on datasets, accuracy, precision, recall, and other diagnostic metrics. | TCGA-COAD, LC25000, Zenodo, Kaggle, Chaoyang Dataset | CNN, VGG16, ResNet50, ResNet101, InceptionV3, Swin Transformer, Vision Transformer, MobileNet, XCiT, SqueezeNet, Federated Learning, Explainable AI, PCA, QDA, LDA, SVM | Reported top accuracy values: ResNeXt50 (99.53%), CNN (99.5%), Linear SVM (99.6% precision), ResNet-101 (98.97% accuracy), MobileNet (84.39% accuracy); ResNeXt50 and CNN-based models consistently performed best across studies. | Sakshi Takkar & Manik Rakhra, 2025 3rd International Conference on Advancement in Computation & Computer Technologies (InCACCT), IEEE. | Abhay | ||||||||||||||||||
60 | 59 | Deep Learning-Based Colon Cancer Tumor Prediction Using Histopathological Images | To develop a deep learning model for accurate prediction of colon cancer tumors using histopathological images and compare it with various ML models. | LC25000 dataset (10,000 colon tissue images: 5,000 benign, 5,000 adenocarcinomas) | CNN (custom 5-layer DNN), ReLU, Softmax, Adam optimizer, 10-fold cross-validation, image preprocessing (resizing, normalization, RGB to HSV), comparison with XGBoost, RF, KNN, SVM, etc. | Proposed DL model achieved 99.7% accuracy, outperforming all other ML models: XGBoost (99.06%), RF (98.54%), KNN (98.54%), SVM (91.13%). Model precision, recall, and F1-score all at 0.99. | Rahul Deb Mohalder et al., 2022 25th International Conference on Computer and Information Technology (ICCIT), IEEE | Abhay | ||||||||||||||||||
61 | 60 | A Robust Colon Cancer Detection Model Using Deep-Learning | To evaluate deep learning models integrated into a noise-robust pipeline (HSA-NRL) for colon cancer histopathology classification using noisy datasets. | Chaoyang Dataset (China) – 512x512 histology patches with noisy labeling | Two-phase HSA-NRL algorithm, models: ResNet-34, XCiT, SqueezeNet, MobileNet; co-learning architecture; focal loss | MobileNet outperformed others: Accuracy 84.39%, Precision 81.79%, Recall 79.67%, F1-score 80.73%; best model for noisy label histology classification | Vanishka Kadian et al., 2023 3rd Int’l Conf. on Secure Cyber Computing and Communication (ICSCCC), IEEE. | Abhay | ||||||||||||||||||
62 | 61 | Prediction of Colon Cancer using DenseNet121, CNN, and REsNET50 Machine Learning Models and using Image Processing Techniques | To predict colon cancer using image-based machine learning models and assess their performance. | 10,000 colonoscopy images from Kaggle (classified into cancerous and non-cancerous) | DenseNet121, CNN, ResNet50, image preprocessing, data augmentation, fine-tuning, Adam optimizer, cross-entropy loss | DenseNet121 achieved 99.6% accuracy, best among all. CNN: 87%, ResNet50: 80.24%. DenseNet121 had best recall and F1-score in adenocarcinoma class. | Mahadi Hasan et al., 2023 Int’l Conf. on AI Robotics, Signal and Image Processing (AIRoSIP). | Abhay | ||||||||||||||||||
63 | ||||||||||||||||||||||||||
64 | 62 | Automated System for Colon Cancer Detection and Segmentation Based on Deep Learning Techniques | To anlalyse various deep learning methods for classification of colorectal cancer helping in the early prediction of colrectal cancer enabling the patient to get treated on time. | Colon,Colonoscopy,Warwick-QU and CRC-VAL-HE-7K were used in the study.These datasets were divided in to training,testing,validation with 80% of images being used for training,10% for testing and 10% for validation. | The images in the datasets were augmented and CNN was used to classify the colon cancer images.Various optimizers such as Adam,RMSprop,Adadelta,Adamax,Nadam and SGD were used to improve the model accuracy in evaluating the images. | For Dataset-1 CNN-Adadelta produces the best result.For Dataset-2 CNN-Nadam produces the best result.For Dataset-3 shows that no single model was effevtive.Each model excelled in different performance metrics.For Dataset-4 CNN-Adam outperforms the other optimizers.The initial result showed that the CNN-Adam outperformed all other DL optimizers with an average accuuracy of 82%. | Azar, A. T., Tounsi, M., Fati, S. M., Javed, Y., Amin, S. U., Khan, Z. I., Alsenan, S., & Ganesan, J. (2023). Automated system for colon cancer detection and segmentation based on deep learning techniques. International Journal of Sociotechnology and Knowledge Development, 15(1), 1–28. https://doi.org/10.4018/ijskd.326629 | Ankita | ||||||||||||||||||
65 | 63 | Machine learning applications in cancer prognosis and prediction | To present a review on the various ML approaches that are being implemented for prediction of cancer. | ----- | ------ | ----- | Kourou, K., Exarchos, T. P., Exarchos, K. P., Karamouzis, M. V., & Fotiadis, D. I. (2014). Machine learning applications in cancer prognosis and prediction. Computational and Structural Biotechnology Journal, 13, 8–17. https://doi.org/10.1016/j.csbj.2014.11.005 | Ankita | ||||||||||||||||||
66 | 64 | Early cancer detection using deep learning and medical imaging: A survey | To bridge the existing research gap and contribute to improving the cancer detection methods. | ----- | ------ | ----- | Ahmad, I., & Alqurashi, F. (2024). Early cancer detection using deep learning and medical imaging: A survey. Critical Reviews in Oncology/Hematology, 204, 104528. https://doi.org/10.1016/j.critrevonc.2024.104528 | Ankita | ||||||||||||||||||
67 | 65 | Automated Detection of Polyps in CT Colonography images using Deep Learning Algorithms in Colon Cancer Diagnosis | To design a fully automated system using deep learning (CNN) for colon segmentation and polyp detection from CT colonography images, comparing it with traditional ML algorithms. | The Cancer Imaging Archive (TCIA): 825 cases with CT colonography DICOM images; manually labeled blocks (Colon Type 1: 500, Type 2: 400, Noise: 400, Polyp: 245, Normal: 356) | Preprocessing (thresholding, segmentation), CNN (custom), classical ML (RF and KNN), LBP-HOG feature extraction, WEKA for training/testing | CNN outperformed: Colon Type Classification Accuracy – CNN: 87.03%, RF: 84.76%, KNN: 82.92%; Polyp Detection Accuracy – CNN: 88.56%, RF: 85.37%, KNN: 80.30%. CNN also had highest sensitivity and specificity across tasks. | Akshay M. Godkhindi & Rajaram M. Gowda, 2017 International Conference on Energy, Communication, Data Analytics and Soft Computing (ICECDS). DOI: Not listed | Abhay | ||||||||||||||||||
68 | 66 | Machine learning-based colorectal cancer prediction using global dietary data | To combine unsupervised and supervised machine learning algorithms to explore the key dietary features for colon cancer prediction. | A large dataset including 109,343 participants in a dietary-based colorectal cancer ase study from Canada, India, Italy, South Korea, Mexico, Sweden, and the United States was collected by the Center for Disease Control and Prevention | Four types of unsupervised machine learning for nonlinear relationship were used to explore the dimensions of the data including t-distributed stochastic Neighbor embedding (t-SNE), uniform manifold approximation and projection (UMAP), Apriori association rules, principal component analysis (PCA), and factor analysis (FA).Te data was split into 80% for training and 20% for testing. Te data was trained using machine learning (ML) algorithms including neural network (Neuralnet), k-nearest neighbors (kNN), generalized linear model (GLM), and recursive partitioning (Rpart). | Among the unsupervised classifers, t-SNE (Fig. 5) was the best performer.In supervised classifers, all techniques performed very well where accuracy, kappa, sensitivity, and specifcity were above 0.90.Although all classifers were very good predictors of CRC labels, artifcial neural networks had the best accuracy and true positives and true negatives | Rahman, H. A., Ottom, M. A., & Dinov, I. D. (2023). Machine learning-based colorectal cancer prediction using global dietary data. BMC Cancer, 23(1). https://doi.org/10.1186/s12885-023-10587-x | Ankita | ||||||||||||||||||
69 | 67 | Enhancing Lung and Colon Cancer Detection Through Convolutional Neural Networks | To build a CNN-based system for detecting lung and colon cancer from histopathological images and evaluate its performance against traditional diagnostic methods. | CT images and histopathological images; not explicitly named | CNN with multiple convolution, ReLU, pooling, and dense layers; classification with softmax/sigmoid; training and validation accuracy analysis | Achieved ~98% accuracy on both training and validation sets; confusion matrix analysis showed 98% recall, few misclassifications; better than traditional techniques | Ravi Ranjan Kumar et al., ICSES 2024. | Abhay | ||||||||||||||||||
70 | 68 | Deep Learning in Lung and Colon Cancer classifications | To develop a four-CNN framework for classifying five types of colon and lung cancer tissue using histopathology images. | Kaggle "Lung and Colon Cancer Histopathological Images" dataset (25,000 images) | 4 CNNs: (1) Lung vs. Colon, (2) Lung Benign vs. Malignant, (3) Colon Benign vs. Malignant, (4) Lung ACA vs. SCC; preprocessing, augmentation, normalization | Achieved up to 98.3% testing accuracy; designed to reduce misclassification and serve as a second-opinion tool for clinicians | Krishna Mridha et al., ICACCM 2022. | Abhay | ||||||||||||||||||
71 | 69 | Deep Learning - Based Prediction Models for Colon and Rectum Cancer Disease: Enhancing the Precision Diagnosis | To enhance colon and rectal cancer diagnosis using deep learning (CNN, RNN, hybrid models) integrating genomic, imaging, and clinical data. | Multi-modal: genomic, histopathology, medical imaging; sources not explicitly named | CNNs, RNNs, hybrid deep learning models, virtual colonoscopy, CT scans, Optical Coherence Tomography (OCT), serologic testing | Reported 95–100% sensitivity, accuracy, and AUC in some models; OCT achieved 100% sensitivity and ~99% AUC; deep learning enhances early detection and precision | Sunil Kumar Suman et al., COMP-SIF 2025. | Abhay | ||||||||||||||||||
72 | 70 | DB-Scan Algorithm based Colon Cancer Detection And Stratification Analysis | To develop a robust method for colon cancer detection using DB-SCAN clustering and entropy-based decision tree scoring. | 100 colon biopsy images from Zenodo (70 abnormal, 30 normal) | DB-SCAN clustering, entropy-based scoring, decision trees, score-based classification | Achieved up to 99% classification accuracy; demonstrated effective noise removal and outlier stratification | Gundlapalle Rajesh et al., I-SMAC 2020. | Abhay | ||||||||||||||||||
73 | 71 | Early Detection of Colon Cancer using Fine Tuned Light Weight CNN Model | To design a PC-aided system to detect colon cancer from biopsy images using DB-SCAN clustering and entropy-based scoring to classify malignant vs. normal tissue. | 100 colon biopsy images from Zenodo (70 abnormal, 30 normal); 85 for training, 15 for testing | DB-SCAN clustering, entropy-based score calculation, decision tree, information gain-based grouping; implemented in MATLAB | Achieved 99% accuracy, 85.4% sensitivity, and 87.6% specificity. Outperformed other methods like structural/statistical and ensemble classifiers (GECC 98.67%). | Gundlapalle Rajesh et al., I-SMAC 2020, IEEE. ISBN: 978-1-7281-5464-0 | Abhay | ||||||||||||||||||
74 | 72 | Lung and Colon Cancer Classification using EfficientNet B3 Transfer Learning Model | To classify five types of lung and colon cancer tissues using a deep learning model based on EfficientNet B7, comparing it with previous state-of-the-art techniques. | Kaggle dataset of 25,000 histopathology images (5 classes: colon adenocarcinoma, colon benign, lung benign, lung adenocarcinoma, lung squamous cell carcinoma; 20,000 train, 5,000 test) | EfficientNet B7, transfer learning, Adam optimizer, normalized confusion matrix, precision/recall/F1-score, performance plots, comparison with ResNet50, CNN, SVM-RBF | Achieved 98% overall accuracy; Precision and Recall up to 100% in several classes (colon and lung types); outperformed prior models (ResNet50: 93.13%, CNN: 97%, SVM-RBF: 97%) | Rahul Singh, Neha Sharma, Rupesh Gupta, 2023 World Conference on Communication & Computing (WCONF). | Abhay | ||||||||||||||||||
75 | 73 | Efficient Colon Cancer Detection Using Transfer Learning with EfficientNet: A Deep Learning Approach for Enhanced Diagnostic Accuracy | To develop an efficient and scalable deep learning model (EfficientNetB3) for colon cancer detection using histopathological images. | Histopathology image dataset with two classes: Adenocarcinoma and Normal tissue (10,000 colon images) | EfficientNetB3, transfer learning, data augmentation (rotation, flip, scaling), Adamax optimizer, dropout, batch normalization, softmax classification | Achieved 98.86% accuracy, precision of 98%, recall of 97%, F1-score of 0.97. Very low false positives/negatives; model robust and generalizable. | Eshika Jain, Amanveer Singh, 2025 International Conference on Automation and Computation (AUTOCOM). | Abhay | ||||||||||||||||||
76 | 74 | Enhancing Cancer Diagnosis: CNN-Based Classification of Lung and Colon Histopathological Images | To build a CNN model for automated classification of five types of lung and colon cancer tissue using histopathological images. | Kaggle “Lung and Colon Histopathological Images” dataset – 25,000 images (5 classes) | CNN (custom architecture), data augmentation (rotation, flipping, magnification), normalization, 80/20 train-validation split, ReLU and softmax activation | Achieved 97% accuracy. Precision and recall per class: ~0.97 for colon and lung tissues. Confusion matrix showed few misclassifications, mainly between similar lung cancer subtypes. | Dhruv Kumar Soni, Ashu Taneja, 2025 IEEE Int’l Conference on Interdisciplinary Approaches in Technology and Management for Social Innovation (IATMSI). | Abhay | ||||||||||||||||||
77 | 75 | Prediction of Colon Cancer Stages and Survival Period with Machine Learning Approach | To predict tumor in the TNM staging (tumor, node, and metastasis) stage of colon cancer using the most influential histopathology parameters and to predict the five years disease-free survival (DFS) period using machine learning (ML) in clinical research. | From the colorectal cancer (CRC) registry of Chang Gung Memorial Hospital, Linkou, Taiwan, 4021 patients were selected for the analysis. | Different ML classifiers such as Random Forest, Support Vector Machine (SVM), Logistic Regression (LR), Multilayer Perceptrons (MLP), K-Nearest Neighbors (KNN), and Adaptive Boosting (AdaBoost).Implementing different classifiers in Scikit-learn for the multi-class classification, the scheme of one-against-one was used for the SVM and the scheme OvR was used for Logistic Regression (LR) and other models, which gave the average of all metrics used in our analysis. In total, 4003 samples were used for the analysis. In each round of the experiment, the data were randomly split into 80% for the training of the ML models using five-fold cross-validation and 20% was reserved for the independent testing. | It was observed that Random Forest performed well with an accuracy of 0.89 for the tumor staging. Furthermore, the top-performing model Random Forest achieved an accuracy of 0.84 and AUC of 0.82 ± 0.10 for predicting the five years DFS of the colon cancer patients. It was also observed that the patients with TAS ≥9.8 had poor DFS, whereas the DFS were found to exceed 10 years of survival in case of patients with TAS <9.8. | Gupta, P., Chiang, S., Sahoo, P. K., Mohapatra, S. K., You, J., Onthoni, D. D., Hung, H., Chiang, J., Huang, Y., & Tsai, W. (2019). Prediction of Colon Cancer Stages and Survival Period with Machine Learning Approach. Cancers, 11(12), 2007. https://doi.org/10.3390/cancers11122007 | Ankita | ||||||||||||||||||
78 | 76 | Classification and Diagnostic Prediction of Colorectal Cancer Mortality Based on Machine Learning Algorithms: A Multicenter National Study | Aims to predict survival outcomes of CRC patients using machine learning (ML) methods | A retrospective analysis included 1853 CRC patients admitted to three prominent tertiary hospitals in Iran from October 2006 to July 2019. | Six ML methods, namely logistic regression (LR), Naïve Bayes (NB), Support Vector Machine (SVM), Neural Network (NN), Decision Tree (DT), and Light Gradient Boosting Machine (LGBM), were developed with 10-fold cross-validation. Feature selection employed the Random Forest method based on mean decrease GINI criteria. Model performance was assessed using Area Under the Curve (AUC). | Time from diagnosis, age, tumor size, metastatic status, lymph node involvement, and treatment type emerged as crucial predictors of survival based on mean decrease GINI. The NB (AUC = 0.70, 95% Confidence Interval [CI] 0.65–0.75) and LGBM (AUC = 0.70, 95% CI 0.65–0.75) models achieved the highest predictive AUC values for CRC patient survival. | Mohammadi, G., Looha, M. A., Pourhoseingholi, M. A., Tavirani, M. R., Sohrabi, S., Khaneh, A. Z. S., Piri, H., Alaei, T., Parvani, N., Vakilzadeh, I., Javadi, S., Cheshmeh, Z. M. H., Razzaghi, Z., Robati, R. M., Azodi, M. Z., Shahraki, S. Z., Hadavi, M., Talebi, R., Yazdani, J., . . . Khodakarim, S. (2024). Classification and diagnostic prediction of colorectal cancer mortality based on machine learning algorithms: a multicenter national study. Asian Pacific Journal of Cancer Prevention, 25(1), 333–342. https://doi.org/10.31557/apjcp.2024.25.1.333 | Ankita | ||||||||||||||||||
79 | 77 | Explainable Machine Learning Models for Colorectal Cancer Prediction Using Clinical Laboratory Data | This study aims to develop machine learning (ML) models for CRC risk prediction using clinical laboratory data. | This retrospective, single-center study analyzed laboratory examination data from healthy controls (HC), polyp patients (Polyp), and CRC patients between 2013 and 2023. | Five ML algorithms, including adaptive boosting (AdaBoost), extreme gradient boosting (XGBoost), decision tree (DT), logistic regression (LR), and random forest (RF), were employed to classify subjects into HC vs Polyp vs CRC, HC vs CRC, and Polyp vs CRC, respectively. | This study included 31 539 subjects: 11 793 HCs, 10 125 polyp patients, and 9621 CRC patients. The XGBoost model achieved the highest AUCs of 0.966 for differentiating HC from CRC and 0.881 for Polyp from CRC, outperforming carcino-embryonic antigen (CEA) and fecal occult blood testing (FOBT) tests. This model could also identify CEA-negative or FOBT-negative CRC patients. Incorporating stool miR-92a detection into the model further improved diagnostic performance. Shapley additive explanations (SHAP) plots indicated that FOBT, CEA, lymphocyte percentage (LYMPH%), and hematocrit (HCT) were the most significant features contributing to CRC diagnosis. | Li, R., Hao, X., Diao, Y., Yang, L., & Liu, J. (2025). Explainable machine learning models for colorectal cancer prediction using clinical laboratory data. Cancer Control, 32. https://doi.org/10.1177/10732748251336417 | Ankita | ||||||||||||||||||
80 | 78 | A Machine Learning Approach for Detection and Classification of Colon Cancer using Convolutional Neural Network Architecture | To develop a deep learning model based on CNN architecture to accurately classify and detect colon cancer. | The model was trained with a dataset containing 10000 histopathological images of the colon. The dataset was divided into 2 classes: colon adenocarcinoma and normal benign tissue of the colon, each containing 5000 images respectively. We have downloaded the dataset from Kaggle, an online data science repository from google. The original dataset (LC25000) consisted of 25000 histopathological images of 2 variants of cancer: lung cancer and colon cancer. In our work, but only 2 classes of colon histopathological images were considered in this study. | The CNN model used for this study had 7 convolutional layers, 4 maxpool layers, 2 fully connected layer, and 1 output layer. The layer design followed a schema of: [{(2 Conv x 1 MaxPool)* 2}+ {(3 Conv x 2 MaxPool)* 1}+ 1Dropout + 2 FC + 1 Output]. During training, 80% of the images were used for training, 10% for testing, and 10% for validation. The kernel size used was 3x3, which performs convolution on the input image through element-by-element matrix multiplication.Adam optimizer was used to lessen the loss, occurring during training of the model and the learning rate was set to 0.0001. The model was trained for a period of 30 epochs. | The proposed model has achieved a remarkable accuracy of 98.79% and was quite successful in classifying the two classes of histopathological colon images, with one class being malignant adenocarcinoma | Sinha, N. S. (2024). A Machine Learning Approach for Detection and Classification of Colon Cancer using Convolutional Neural Network Architecture. Deleted Journal, 20(7s), 1065–1071. https://doi.org/10.52783/jes.3543 | Ankita | ||||||||||||||||||
81 | 79 | Early Colorectal Cancer Detected by Machine Learning Model Using Gender, Age, and Complete Blood Count Data | To validate a machine learning colorectal cancer detection model on a US community-based insured adult population.The overall purpose of this CRC detection model is to detect adults who are likely at an elevated risk of having CRC based on their demographics, and previous laboratory test results in a population of adults with comprehensive medical history data available. | The colorectal cancer cases were selected from the KP Tumor Registry using the following selection criteria: (1) diagnosed with colorectal cancer—International Classification of Diseases-Oncology (ICD-O) sites C18.0–C18.9, C19.9, and C20.9; (2) had one or more CBCs within 6 months of the CRC diagnosis date; (3) had at least 180 days of continuous KPNW enrollment prior to CRC diagnosis date (enrollment gaps of 90 days or less were considered continuous enrollment); (4) CRC patients with any cancer diagnosis prior to the CRC diagnosis date were excluded; and (5) CRC patients with other cancers diagnosed on the same date as the CRC diagnosis date were flagged so that this variable was available to the detection modeling effort. | CNN was used. | Area under the receiver operating characteristics curve for detecting colorectal cancer was 0.80 ± 0.01. At 99% specificity, the odds ratio for association of a high-risk detection score with colorectal cancer was 34.7 (95% CI 28.9–40.4). The detection model had the highest accuracy in identifying right-sided colorectal cancers. | Hornbrook, M. C., Goshen, R., Choman, E., O’Keeffe-Rosetti, M., Kinar, Y., Liles, E. G., & Rust, K. C. (2017). Early colorectal cancer detected by machine learning model using gender, age, and complete blood count data. Digestive Diseases and Sciences, 62(10), 2719–2727. https://doi.org/10.1007/s10620-017-4722-8 | Ankita | ||||||||||||||||||
82 | 80 | ALC: Automated Lung Cancer Detection Framework in Thoracic CT Scans Based on Deep Learning | To propose ALC-Net, a high-precision lightweight model for lung cancer detection in chest CT scans, addressing low contrast and complex anatomy. | Lung-Dataset (4536 CT images from Wuhan Third People’s Hospital and Jingzhou Central Hospital; 3 classes: nodules, cancer, adenocarcinoma) | YOLOv8n baseline enhanced with Edge Learning Attention (ELA), High-Resolution Feature Pyramid Block (HGBlock), Dynamic Block; trained in PyTorch | Achieved mAP50: 98.2%, mAP50-95: 79.3%, F1-score: 95.5%, Recall: 94.7%. Outperformed YOLOv9c, Mask R-CNN, YOLOv8s while requiring fewer FLOPs (10.8G) and parameters (293.1M) | Mingzi Wu et al., 2025 5th Intl. Conf. on Artificial Intelligence and Industrial Tech Applications (AIITA). | Abhay | ||||||||||||||||||
83 | 81 | Deep Learning - Based Prediction Models for Colon and Rectum Cancer Disease: Enhancing the Precision Diagnosis | To review deep learning applications for improving early diagnosis and treatment of colon and rectal cancers using various imaging and genomic data. | Mixed datasets: imaging, histopathology, genomics, clinical data (sources like CT, OCT, etc. discussed but not centralized) | CNNs, RNNs, hybrid DL models; image preprocessing, classification layers, regularization, dropout; applications include OCT, CT scans, colonoscopy; TensorFlow & Keras used | DL models showed excellent performance: e.g., CNN with 100% sensitivity and ~99% AUC (OCT); high accuracy in CT-based classification. Future potential in integrating multi-omics data. | Sunil Kumar Suman et al., 2025 Intl. Conf. on Computing for Sustainability and Intelligent Future (COMP-SIF). | Abhay | ||||||||||||||||||
84 | 82 | Automated Approaches For Detecting And Classifying Colon Cancer: A Comprehensive Study | To review and evaluate automated ML and DL techniques for colon cancer detection/classification, summarizing models, datasets, accuracy, and limitations. | Multiple datasets (CRC-VAL-HE-7K, LC25000, GenBank, Microarray, CT colonography, SEER, CRC-HE-VAL-7K, etc.) | CNNs, RNNs, hybrid DL models; image preprocessing, classification layers, regularization, dropout; applications include OCT, CT scans, colonoscopy; TensorFlow & Keras used | Accuracies range from 85.5% to 98.8% across models. DL models (CNN, GoogLeNet, TL) showed better generalizability. Overfitting and limited datasets remain major issues. | Namitha T. M., Vinod Kumar R. S., 2025 IEEE 2nd Int’l Conf. on Trends in Engineering Systems and Technologies (ICTEST). | Abhay | ||||||||||||||||||
85 | 83 | An Effective Approach for Detecting Colon Cancer Using Deep Convolutional Neural Network | To develop a DCNN-based automated system using adaptive Wiener filter and GLCM features for effective colon cancer classification. | PLCO Colon Dataset (~155,000 participants), images processed via MATLAB | DCNN, adaptive Wiener filter (noise removal), k-means clustering (segmentation), GLCM (feature extraction), softmax classification | Achieved 94.7% classification accuracy; significant reduction in noise, enhanced segmentation, and effective performance compared to other models | Jamel Smida et al., 2024 IEEE Intl. Conf. on Advanced Systems and Emerging Technologies (IC_ASET). | Abhay | ||||||||||||||||||
86 | 84 | A Novel Deep Learning Approach for Colon and Lung Cancer Classification Using Histopathological Images | To propose a novel deep learning model (DeepLCCNet) to classify five types of colon and lung cancer tissues using histopathological images. | LC25000 dataset (25,000 images: 5 classes - CAC, CBT, LAC, LBT, LSCC) | DeepLCCNet (20-layer CNN), 5-fold cross-validation, Leaky ReLU, Batch Normalization, Max-pooling, Dropout, MATLAB R2020a implementation | Achieved avg. accuracy of 99.67%, precision 99.12%, recall 99.04%, F1-score 99.08%; outperformed EfficientNet, SqueezeNet, ResNet18, DenseNet201, and MobileNetV2 | Naeem Ullah et al., 2023 19th IEEE Int'l Conf. on e-Science. | Abhay | ||||||||||||||||||
87 | 85 | Enhanced Colon Cancer Screening with Endoscopic Images Using Deep Learning Techniques | To detect colon cancer through endoscopy image classification using multiple pre-trained deep learning models. | KVASIR and ETIS-Larib-Polyp DB (800 images expanded to 3,200 via augmentation) | Transfer learning with InceptionV3, EfficientNetB2, ResNet50, MobileNetV3; augmented dataset; metrics: Accuracy, Precision, Recall, F1-score, MCC, Kappa | InceptionV3 performed best: Accuracy 94%, Precision 94%, Recall 94%, F1-score 94%, MCC 0.917, Kappa 0.92. EfficientNetB2 was close at 93%. | R.P. Rupesh et al., 2024 5th Int’l Conf. on Communication, Computing & Industry 6.0 (C2I6). | Abhay | ||||||||||||||||||
88 | 86 | Advancements and Challenges in the Image-Based Diagnosis of Lung and Colon Cancer: A Comprehensive Review | To provide valuable insights into the recent developments and challenges in image-based diagnosis for lung and colon cancers, underscoring both the remarkable progress and the hurdles that still need to be overcome to optimize cancer care. | ----- | ------ | ----- | Patharia, P., Sethy, P. K., & Nanthaamornphong, A. (2024). Advancements and Challenges in the Image-Based Diagnosis of Lung and Colon Cancer: A Comprehensive Review. Cancer Informatics, 23. https://doi.org/10.1177/11769351241290608 | Ankita | ||||||||||||||||||
89 | 87 | Improved Water Strider Algorithm With Convolutional Autoencoder for Lung and Colon Cancer Detection on Histopathological Images | To develop the IWSACAE-LCCD method integrating MobileNetv2 and CAE for enhanced classification of lung and colon cancer from histopathology images. | LC25000 (25,000 images: 5 classes – Con-Adc, Con-BeT, Lug-Adc, Lug-BeT, Lug-Scc) | Median filtering, MobileNetv2 (feature extraction), Improved Water Strider Algorithm (hyperparameter tuning), Convolutional Autoencoder (classification) Median filtering, MobileNetv2 (feature extraction), Improved Water Strider Algorithm (hyperparameter tuning), Convolutional Autoencoder (classification) | Achieved high accuracy (not explicitly numerically stated); demonstrated better performance than other methods across standard metrics (accuracy, precision, recall, F1-score). | Hamed Alqahtani et al., IEEE Access 2024. | Abhay | ||||||||||||||||||
90 | 88 | Cloud-super-computing virtual colonoscopy with motion-based navigation for colon cancer screening | To develop a high-performance, mobile-accessible virtual colonoscopy system using cloud computing and Kinect-based motion navigation. | Multi-center lfVC datasets, no exact count mentioned | Cloud infrastructure (HPCC), real-time CABP and CADe, ParaViewWeb for visualization, Kinect motion interface for intuitive 3D navigation | Improved workflow and image interpretation accuracy for colorectal screening. Enabled real-time, mobile-based visualization and interaction. | Hiroyuki Yoshida, 2013 IEEE ICCE-Berlin. | Abhay | ||||||||||||||||||
91 | 89 | Detection of Colon Wall Outer Boundary and Segmentation of the Colon Wall Based on Level Set Methods | To accurately segment both the inner and outer walls of the colon from CT colonography using advanced level set methods. | 3 CT colonography datasets (512×512×512 voxels) | Level set segmentation, geodesic active contours, directional gradient filters, sigmoid mapping | Achieved subvoxel-accurate outer colon wall segmentation; effective even in low-contrast and adjacent organ areas | R. Van Uitert et al., 2006 IEEE EMBS Annual Intl. Conf. | Abhay | ||||||||||||||||||
92 | 90 | Automatic polyp detection of colon using high resolution CT scans | To design an efficient system for polyp detection using fuzzy connectivity and concave region search in CT colonography. | 17 CT colon datasets, total 30 polyps, slice dimensions: 512×512×300–700 | Colon segmentation via fuzzy connectivity; concave region detection for potential polyps; 3D object reconstruction | Successfully detected all polyps before classification; future work focuses on reducing false positives in classification stage | Jamshid Dehmeshki et al., 2003 IEEE ISPA. | Abhay | ||||||||||||||||||
93 | 91 | Advancing Cancer Classification with Hybrid Deep Learning: Image Analysis for Lung and Colon Cancer Detection | To classify lung and colon cancer from high dimensional histopathological images, employing an advanced hybrid deep learning model. | Dataset encompasses 20,000 high-resolution images across five distinct classes, posing significant challenges in terms of computational efficiency and model accuracy.There are 1,000 photos of colon tissue (500 benign and 500 colon adenocarcinomas) and 750 photographs of lung tissue (500 benign, 250 lung adenocarcinomas, and 250 lung squamous cell carcinomas) in the original sample. | Integrates two distinct Convolutional Neural Networks (CNNs) to adeptly handle high-dimensional image datasets, specifically targeting lung and colon cancer classification. The model commences with an input layer designed to accommodate 224x224 pixel images, ensuring compatibility with high-resolution medical imagery. The first CNN, cnn_model_1, employs 3x3 convolutional filters across three convolutional blocks, each followed by max-pooling layers, to extract finegrained features and patterns from the input images, capturing intricate details crucial for medical image analysis. Concurrently, the second CNN, cnn_model_2, utilizes larger 5x5 convolutional filters, also across three convolutional blocks with subsequent max-pooling, aiming to grasp broader and more global features from the images. | Achieves an impressive classification accuracy of 99%, demonstrating the model’s capability to handle high-dimensional datasets with precision. | Sobur, A., Rana, M. I. C., Hossain, M. Z., Hossain, A., Kabir, M. F., & International Journal of Creative Research Thoughts (IJCRT). (2024). Advancing Cancer Classification with Hybrid Deep Learning: Image Analysis for Lung and Colon Cancer Detection [Journal-article]. International Journal of Creative Research Thoughts (IJCRT), 12(2), c38–c39. https://ijcrt.org/papers/IJCRT2402237.pdf | Ankita | ||||||||||||||||||
94 | 92 | Lung and Colon Cancer Detection from CT Images Using Deep Learning | To aid the early detection of lung and colon cancer,proposes a computer-aided diagnostic approach that employs a Deep Learning (DL) architecture to enhance the detection of these cancer types from Computed Tomography (CT) images of suspected body parts. | The dataset used in this work is the lung and colon cancer histopathological dataset [3] which is also referred to as LC25000. It originally contains 250 images each of Lung benign tissue (Lung n), Lung adenocarcinoma (Lung aca), Lung squamous cell carcinoma (Lung scc), Colon adenocarcinoma (Colon aca), and Colon benign (Colon n) tissue totaling 750 images which were then augmented to 25 000 images with a total of 5000 images in each of the five classes. For the purpose of testing the performance of our network, we created a hold-out test set that contained 20% of the total size of the dataset (i.e. 5000 images). The remaining 20 000 images were used for training and validation in order to experiment and improve the architecture before finally testing it on the test set. The validation split is 10% of the remaining 20 000 images (i.e. 2000 images), while the remaining 18 000 images were used for training. | Reused the pre-trained EfficientNetB7 without its top layer. Then, added 4 dense layers on top of the pre-trained architecture as shown in Figure 2. The EfficientNet model architecture is very deep due to the compound scaling method adopted. The architecture was trained with images of dimension 224 × 224 × 3, thus, that is the input dimension specified in the input layer of the architecture.The final dense layer of the network has five units for classifying the input image into one of the five classes. The network was trained with a base learning rate of 5·10−5 , using the Adam optimizer and a categorical cross-entropy loss function. We also included a dropout layer to drop 30% of the neurons in the previous layer in order to combat overfitting and obtain a robust model. The model was trained with early stopping for 44 epochs. We set a seed value of 42 to ensure reproducibility in the random selection of the test set. The modified EfficientNet-B7 architecture was trained on 18 000 CT images, validated on 2000 CT images and tested on a hold-out set of 5000 images. | Research findings showed detection accuracies of 99.63%, 99.50%, and 99.72% for training, validation, and test sets, respectively | Akinyemi, J. D., Akinola, A. A., Adekunle, O. O., Adetiloye, T. O., & Dansu, E. J. (2023). Lung and colon cancer detection from CT images using Deep Learning. Machine Graphics and Vision, 32(1), 85–97. https://doi.org/10.22630/mgv.2023.32.1.5 | Ankita | ||||||||||||||||||
95 | ||||||||||||||||||||||||||
96 | ||||||||||||||||||||||||||
97 | ||||||||||||||||||||||||||
98 | 96 | Optimizing Cancer Detection: A CNN Approach for Lung and Colon Cancer | To develop a robust CNN model to classify histopathological images of lung and colon cancer into five categories. | Public dataset with 5 classes: Lung adenocarcinoma, Lung squamous cell carcinoma, Lung benign, Colon adenocarcinoma, Colon benign | CNN architecture with multiple convolutional, pooling, and dense layers; ReLU, softmax; data augmentation; trained for 10 epochs | Achieved 98.6% test accuracy; F1-score ≥ 0.95 for all classes; strongest results for benign cases; confusion matrix showed minimal misclassification | Seerat Singla & Rupesh Gupta, ICUIS-2024. | Abhay | ||||||||||||||||||
99 | 97 | An Early Stage Determination of Colon Cancer Through Deep Neural Network | To detect colon cancer genes early using dimensionality reduction and classification through a DNN model | NCBI colon gene expression dataset (2000 genes, 62 samples; 40 tumor, 22 normal) | Feature selection: FSUF + RFMF; DNN classifier; compared with KNN and Naive Bayes; Z-score normalization; 5 hidden layers | FSUF-RFMF-DNN achieved 94.44% accuracy, outperforming KNN (91%) and NB (81%); macro precision: 83%, macro recall: 93%, macro F1: 86% | M. Kalaivani, Dr. K. Abirami, Dr. K. Dharmarajan, ICAIHI 2023. | Abhay | ||||||||||||||||||
100 | 98 | Image segmentation using thresholding for cell nuclei detection of colon tissue | To detect nuclei in HE-stained colon tissue images for early-stage colon cancer diagnosis using threshold-based segmentation techniques. | HE-stained colon tissue images (6 samples tested; dataset not explicitly named) | Gaussian filtering, Thresholding (Otsu's method), Sobel & Canny edge detection, HSV color space transformation, Vector Gradient Method, Pseudocoloring, Morphological operations | Vector Gradient and HSV thresholding methods gave best segmentation results; successfully highlighted nuclei positions using composite images | Archana Nawandhar, Lakshmi Yamujala, Navin Kumar, ICACCI 2015. ISBN: 978-1-4799-8792-4 | Abhay |