A machine learning pipeline for non-invasive brain tumor risk prediction using circulating biomarkers (IL-6, VEGF, GFAP, etc.). Includes:
- Automated data preprocessing
- Explainable AI (SHAP/LIME)
- Flask/FastAPI deployment
- Clinical validation notebooks
graph TD
A[Serum Samples] --> B(Data Cleaning)
B --> C{Feature Selection}
C -->|Top Biomarkers| D[Train Model]
C -->|SHAP Analysis| E[Explainability]
D --> F[Deploy as API]
SHAP analysis reveals which serum biomarkers drive predictions:
| Biomarker | Impact Direction | Clinical Relevance |
|---|---|---|
| IL-6 | ↑ Tumor Risk | Linked to inflammation |
| GFAP | ↑ Tumor Risk | Glial cell damage marker |
| VEGF | ↓ Tumor Risk | Anti-angiogenic effect |