How rama-alsadeq/osteo-by-ai works
Osteo-by-AI is a Flask diagnostic web application for osteoarthritis and osteoporosis. Users submit knee X-rays and/or osteoporosis inputs; the Flask routes dispatch OA images through joint localization and OA grading, and dispatch OP inputs to BMD rules or knee-image analysis. The README also describes a clinical risk-factor model, dynamic multi-source fusion, and SHAP explanations, which are retained as documented capabilities. Runtime relationships are based on the shown excerpts and SOURCE INDEX; unsampled subsystem details are not inferred.
Main entry points: osteo_app.py Flask routes · OP_Frame2.py OP knee analysis · OA_pipeline2.py OA grading · models Clinical risk model
Components 14
Web interface
- Diagnostic interface
templates - Flask routes
osteo_app.py
Osteoarthritis
- Knee localization
pipelines/OA_pipeline.py - OA grading
pipelines/OA_pipeline2.py
Osteoporosis
- BMD rules
pipelines/OP_Frame1.py - OP knee analysis
pipelines/OP_Frame2.py - Clinical risk model
models - Dynamic fusion
models
Models and explanations
- OA model weights
models - OP model weights
models - YOLO helper weights
models - Image explanations
pipelines/OA_pipeline2.py - Clinical explanations
models
Other components
- Clinician
Connections 18
- Clinician → Diagnostic interface: submits inputs
README.md - Diagnostic interface → Flask routes: sends requests
- Flask routes → Diagnostic interface: renders responses
osteo_app.py - Flask routes → Knee localization: localizes knees
osteo_app.py - Flask routes → OA grading: grades crops
osteo_app.py - Knee localization → OA grading: returns resized crops
osteo_app.py - Flask routes → BMD rules: requests BMD diagnosis
osteo_app.py - Flask routes → OP knee analysis: analyzes X-rays
osteo_app.py - Flask routes → Clinical risk model: predicts risk
- BMD rules → Dynamic fusion: provides modality result
- OP knee analysis → Dynamic fusion: provides modality result
- Clinical risk model → Dynamic fusion: provides modality result
- OA grading → Image explanations: generates Grad-CAM
pipelines/OA_pipeline2.py - OP knee analysis → Image explanations: generates Grad-CAM
pipelines/OP_Frame2.py - Clinical risk model → Clinical explanations: explains risk factors
- OA model weights → OA grading: loaded by
- YOLO helper weights → Knee localization: used for detection
- OP model weights → OP knee analysis: used for inference