HogbaFranklin/mamacare-ai-app

Explainable multimodal breast cancer screening decision-support platform for low-resource clinical settings — FastAPI app combining a fine-tuned MobileNetV2 (mammogram images, Grad-CAM) and an MLP (clinical features, SHAP) with role-based auth, radiologist review, and audit trail. CSCD602 Advanced Software Engineering exam project. # MammaCare AI — CSCD602 Exam Project Explainable breast cancer screening decision-support platform, built on the pretrained models from the author's MSc thesis (fine-tuned MobileNetV2 for mammogram images, MLP for WBCD clinical features). ## Local setup (Windows / VS Code) Open this folder in VS Code, open a terminal, then: ```powershell python -m venv venv venv\Scripts\activate pip install -r requirements.txt ``` If you hit a Windows long-path error during install, enable long path support once (as Administrator, then restart): ```powershell New-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\FileSystem" -Name "LongPathsEnabled" -Value 1 -PropertyType DWORD -Force ``` ## Seed demo accounts (run once) ```powershell python -m app.seed ``` Creates: | Role | Email | Password | |---|---|---| | Clinician | clinician@mammacare.demo | Demo1234! | | Radiologist | radiologist@mammacare.demo | Demo1234! | | Admin | admin@mammacare.demo | Demo1234! | ## Run ```powershell uvicorn app.main:app --reload ``` Then open 127.0.0.1 ## Run the automated test suite ```powershell python -m tests.run_functional_tests ``` Runs 23 in-process functional tests covering auth, RBAC, both ML assessment pipelines, validation, review workflow, audit trail, and admin functions. Results are also written to tests/test_results.json. ## What's in this package - `app/` — the FastAPI application (routers, services, templates, models) - `models/` — pretrained model artifacts from the thesis (finetuned_model.h5, tabular_only_model.h5, scaler.pkl, breast_cancer_model.tflite) - `tests/` — automated functional test suite + latest results - `documentation/` — Testing_Report.docx - `requirements.txt`, `README.md` — this file ## Architecture notes - Single FastAPI service — auth, patient/case management, both independent ML assessment branches (image + Grad-CAM, tabular + SHAP), radiologist review, audit trail, admin dashboard. - Patient references are auto-generat …