1 of 16

LungcareAI

DevOps, Enterprise and Machine learning/AI

Illustration by Smart-Servier Medical Art

2 of 16

Our team

Jeremiah O. O.

Daniel Ope

Chief Technology Officer (CTO)

Chief Visionary Officer (CVO)

3 of 16

Problem Statement

2.5M new cases, 1.8M deaths in 2022 globally on lung health issue.

4,251 new cases in West Africa, 1,675 in Nigeria on lung health issue.

Rural clinics image only ~12% of patients vs 50% urban on lung health issue.

Illustration by Smart-Servier Medical Art

4 of 16

Solution

Project Type

Beginner-friendly low-code / no-code health web application for lung image classification and AI chat.�
Designed for use by doctors and patients with minimal technical skills, lack of access to resources and researcher.

LungcareAI

Illustration by Smart-Servier Medical Art

5 of 16

Solution

Trained on 16,000+ lung images — including ~15,000 histopathology slides and ~1,000 CT scans — using Keras on Google Colab.� Achieved 98.37% accuracy for histopathology and 73.02% accuracy for CT scans.

🧠 The model can detect:

Benign
Adenocarcinoma
Squamous Cell Carcinoma
Normal tissue

LungcareAI

Illustration by Smart-Servier Medical Art

6 of 16

Solution

🧩 Alternative Deployment (No-Code)

An additional model was trained on 799+ labeled histopathology images using Azure Custom Vision, achieving 100% training accuracy.

✅ No coding required�
🖱️ Drag-and-drop interface for instant image classification�
🌐 Accessible via Azure portal for rapid testing and deployment�

Ideal for clinicians or researchers who prefer a GUI-based workflow over writing code.

LungcareAI

Illustration by Smart-Servier Medical Art

7 of 16

Solution

LungcareAI architectural diagram

Illustration by Smart-Servier Medical Art

8 of 16

Solution

LungcareAI can be used on multiple devices with sable internet connection.

LungcareAI

Computers

Smart phones

Tablets/Ipads

Illustration by Smart-Servier Medical Art

9 of 16

How it works

Visit the Breath Safe Web App� → Users access the app via browser (mobile or desktop).�

Click “Get Started”� → Launches main dashboard for image analysis.�

Upload a CT or Histopathology Image� → Image is analyzed using our Keras model (trained on 16,000 images).� → Output: Benign / Adenocarcinoma / Squamous Cell / Normal (for CTs).�

(Optional) Chat with BreathSafe AI Assistant� → Users can ask questions about lung cancer, reports, or next steps.�

(Optional) Request Azure Custom Vision Access� → For no-code clinics: request login to upload & classify images via drag‑and‑drop UI� → Uses secondary model (trained on 799+ images).

Illustration by Smart-Servier Medical Art

10 of 16

Impact: Target market

Nigeria

1,675 new lung cancer cases annually (GLOBOCAN 2022)�

🏥 Current rural detection covers only ~12% (~201 patients)�

🎯 Breath Safe could enable screening for all 1,675 cases�

🔍 Potential for +1,474 additional early detections each year

11 of 16

Impact: SDG 3 (Good Health and well being )

✅ Supports SDG 3.4: Aims to reduce premature deaths from non-communicable diseases (NCDs) by one-third by 2030 through early lung cancer detection.�
✅ Amplifies healthcare capacity: Enables one doctor to support 2,500+ patients using AI-assisted triage and diagnosis tools.�
✅ Reduces mortality through early detection: Identifies benign vs. malignant subtypes (Adenocarcinoma, Squamous Cell) before critical progression.�
✅ Scalable to other high-burden regions: Designed for low-resource settings like India, South Asia, and Latin America with adaptable models and no-code options.

12 of 16

Uniqueness

Dual Training Paths

A powerful deep learning model (DenseNet121) trained locally on 16K lung images.
A parallel no-code model trained on 800 curated samples using Azure Custom Vision for accessible cloud-based use.

Two Modalities, One Platform

Handles both CT scans and histopathology slides with equal ease.

Offline-Ready Architecture

Local model designed for containerized inference in remote clinics.

Embedded AI Agent

Helps interpret model outputs and provides clinical context in plain English.

13 of 16

Social Good

Bridging diagnostic gaps in rural clinics (currently <15% imaging coverage).
Advancing SDG 3.4: Early detection could help save over 600,000 lives annually.
Empowering non-specialists: AI helps community health workers participate in diagnosis and triage.
Fostering collaboration: Viewer access and open-source codebase encourage transparency and feedback.

14 of 16

Financial Next step: Breakdown

Use Case	Amount	Why It Matters
📦 Expand dataset (1,000+ images)	$2,000	Improve model robustness across subtypes & demographics
🧠 Hire part-time medical advisor	$1,500	Validate model results and improve trust in rural clinics
🖥️ Build offline-ready web app	$2,000	Enable low-connectivity clinics to use AI model without Azure
📊 Pilot in 5 rural clinics	$1,500	Real-world testing & feedback loop for reliability
📣 Community outreach + education	$1,000	Drive awareness in rural areas & train nurses to use the app

15 of 16

Next Steps

Hardware use

support people with limited access to lung health

Light Coding

Build native mobile application

Partnership

Partner with existing health organization

Scale app up

Pilot in additional regions

Illustration by Smart-Servier Medical Art

16 of 16

Thank you

LungCareAI

🌐View website at: https://gibbon-clever-bream.ngrok-free.app/lungcareai

▶️Watch presentation video at: https://youtu.be/aiCDGri1ctM