Computer Vision |
Brain Tumor Detection
A model that could identify a specific patient's brain condition
Argon Riadi | HCK004
Why do diagnostic errors occur?
?
Miscommunication between patient and doctor
Incorrect test is ordered
Test is read incorrectly
There are several thousand known diseases but only so many symptoms
https://www.aarp.org/health/conditions-treatments/info-2022/avoid-diagnostic-errors.html
Meningioma Tumor
Originate in the meninges, which are the outer three layers of tissue between the skull and the brain that cover and protect the brain just under the skull
Brain Tumor Dataset
Glioma Tumor
Gliomas are brain tumours that start in glial cells. These are the supporting cells of the brain and the spinal cord
Pituitary Tumor
Most pituitary tumors don’t cause symptoms. They are not diagnosed or they are found only during a routine brain imaging test. About 25% of people may have small pituitary tumors without knowing it.
Train
Test
0
250
500
750
1,000
Meningioma Tumor
Glioma Tumor
Pituitary Tumor
No Tumor
Meningioma Tumor
28.9%
Glioma Tumor
28.7%
Pituitary Tumor
28.7%
No Tumor
13.7%
Meningioma Tumor
29.2%
No Tumor
26.6%
Glioma Tumor
25.4%
Pituitary Tumor
18.8%
Training
Test
Exploratory Data Analysis
Dataset Example
PROCESS FLOW
SEQUENTIAL MODEL
(TUNING-FLATTEN)
SEQUENTIAL MODEL
TUNING-GLOBAL AVERAGE POOLING
1.
2.
--------COMPARISON--------
SEQUENTIAL - FLATTEN
1.
1
2
3
4
5
6
0.8
0.6
0.4
0.2
0
1
2
3
4
5
6
2
1.5
1
0.5
0
Accuracy
Training 0.76
Val 0.51
Loss
Training 0.59
Val 0.71
SEQUENTIAL - GLOBAL AVERAGE POOLING
2.
Accuracy
Training 0.66
Val 0.68
Loss
Training 0.85
Val 0.75
1
2
3
4
5
6
0.8
0.6
0.4
0.2
0
1
2
3
4
5
6
1.5
1
0.5
0
SEQUENTIAL - FLATTEN
1.
1
2
3
4
5
6
0.8
0.6
0.4
0.2
0
1
2
3
4
5
6
2
1.5
1
0.5
0
Accuracy
Training 0.76
Val 0.51
Loss
Training 0.59
Val 0.71
SEQUENTIAL - GLOBAL AVERAGE POOLING
2.
Accuracy
Training 0.66
Val 0.68
Loss
Training 0.85
Val 0.75
1
2
3
4
5
6
0.8
0.6
0.4
0.2
0
1
2
3
4
5
6
1.5
1
0.5
0
Accuracy | | | 0.73 | 492 |
Macro Avg | 0.77 | 0.72 | 0.73 | 492 |
Weighted Avg | 0.75 | 0.73 | 0.73 | 492 |
| Precision | Recall | F1-Score | Support |
Glioma Tumor | 0.83 | 0.65 | 0.73 | 154 |
Meningioma Tumor | 0.64 | 0.75 | 0.69 | 153 |
No Tumor | 0.87 | 0.59 | 0.70 | 68 |
Pituitary Tumor | 0.72 | 0.91 | 0.80 | 117 |
CLASSIFICATION REPORT -
SEQUENTIAL FLATTEN
THANK YOU!