How nithin2006-l/ai_pneumonia_detection works
This is a Flask clinical-portal prototype for chest X-ray classification and report management. Users access role-specific portal views, submit scans for preprocessing and model inference, and receive a prediction with a Grad-CAM visualization; the application records reports and audit history in SQLite and flags repeat uploads. Separate capabilities include account and doctor administration, a clinical-assistant integration, and an optional training workflow. The main workflow is visible in the app and model excerpts; some portal and integration details are documented but not fully sampled.
Main entry points: app.py Portal routes · train.py Training pipeline · model.py Hybrid classifier · templates Portal views
Components 19
Clinical portal
- Portal routes
app.py - Portal views
templates - Accounts and roles
app.py - Clinical assistant
app.py
Image analysis
- Scan upload
app.py - Repeat detection
app.py - CLAHE preprocessing
model.py - Hybrid classifier
model.py - Grad-CAM overlays
model.py
Reports and records
- Report management
app.py - Portal database
- Audit history
app.py
Model training
- Training pipeline
train.py - Image dataset
sample data set - Model checkpoint
Other components
- Patient
- Doctor
- Administrator
- Gemini service
Connections 21
- Patient → Portal routes: uses portal
README.md - Doctor → Portal routes: uses portal
README.md - Administrator → Portal routes: uses portal
README.md - Portal routes → Portal views: renders views
app.py - Portal routes → Accounts and roles: handles accounts
app.py - Portal routes → Scan upload: accepts scans
app.py - Scan upload → Repeat detection: checks hashes
README.md - Scan upload → CLAHE preprocessing: enhances image
app.py - Scan upload → Hybrid classifier: runs inference
app.py - Scan upload → Grad-CAM overlays: generates overlay
README.md - Portal routes → Report management: manages reports
app.py - Report management → Portal database: stores reports
app.py - Accounts and roles → Portal database: stores profiles
app.py - Audit history → Portal database: stores history
app.py - Portal routes → Audit history: records actions
README.md - Portal routes → Clinical assistant: serves assistant
app.py - Clinical assistant → Gemini service: requests responses
- Training pipeline → Image dataset: loads images
train.py - Training pipeline → Hybrid classifier: trains model
train.py - Training pipeline → Model checkpoint: saves weights
train.py - Portal routes → Model checkpoint: loads weights
README.md