gouravkar007/ai-based-maternal-health-preeclampsia-risk-prediction-system

Purpose and Entry Points

This repository delivers an educational maternal-health screening product: a Streamlit application performs staged ML risk assessment for preeclampsia, provides guidance, reporting, analytics, and a Gemini-backed chatbot. The public landing page is index.html; the runnable application entry point is AI-Based-Maternal-Health-Preeclampsia-Risk-Prediction-System/app/app.py.

Deployment Topology

GitHub Pages serves the static landing page, which links users to the separately hosted Streamlit service. Render is the app hosting target and its service definition is in AI-Based-Maternal-Health-Preeclampsia-Risk-Prediction-System/render.yaml. Python 3.11+ is the runtime; primary libraries include Streamlit and scikit-learn, declared through AI-Based-Maternal-Health-Preeclampsia-Risk-Prediction-System/pyproject.toml.

Streamlit Application Shell

The UI orchestration, navigation, patient input forms, emergency checks, dashboard visualizations, prediction presentation, and report-generation flow are centralized in AI-Based-Maternal-Health-Preeclampsia-Risk-Prediction-System/app/app.py. Streamlit presentation settings are bounded in AI-Based-Maternal-Health-Preeclampsia-Risk-Prediction-System/.streamlit/config.toml. This is a server-rendered interactive Python app, not a separate frontend/API architecture.

Staged Clinical Risk Flow

The primary control flow is:

1. Collect maternal demographic and clinical intake values.

2. Run the general maternal-health classifier.

3. Return low-risk guidance directly, or escalate high-risk results.

4. Collect/use the targeted preeclampsia feature set and run its classifier.

5. Present risk-specific recommendations and optionally generate a downloadable PDF summary.

The two persisted model artifacts are AI-Based-Maternal-Health-Preeclampsia-Risk-Prediction-System/app/models/maternal_health_model.pkl and AI-Based-Maternal-Health-Preeclampsia-Risk-Prediction-System/app/models/preeclampsia_model.pkl.

Model Serving Boundary

AI-Based-Maternal-Health-Preeclampsia-Risk-Prediction-System/app/models/loader.py is the model-loading boundary between UI logic and serialized scikit-learn artifacts. The repository contains inference assets only; dataset ingestion, feature engineering, training, evaluation, and model selection are described as upstream work rather than represented as executable source directories. Runtime correctness therefore depends on app-provided features matching each serialized model’s expected preprocessing/schema.

Guidance, Safety, and State

Clinical advice and threshold-style guidance are data-driven through AI-Based-Maternal-Health-Preeclampsia-Risk-Prediction-System/app/doctor_advice.json, consumed by the app’s result and safety flows. AI-Based-Maternal-Health-Preeclampsia-Risk-Prediction-System/app/cache.py provides the caching boundary for expensive or repeatable runtime operations, likely including model/resource reuse. Outputs are screening guidance only, with no indicated persistence layer, user account system, or clinical-record integration.

Conversational AI Integration

AI-Based-Maternal-Health-Preeclampsia-Risk-Prediction-System/app/chatbot.py encapsulates chatbot behavior and the external Gemini API integration. This is the principal external runtime dependency beyond deployment hosting: Gemini credentials/network availability affect chat functionality, while core local ML predictions rely on packaged .pkl artifacts.