JeevVeda is an integrated AI healthcare suite for early-stage cancer screening, combining a symptom-based medical assistant chatbot, blood report analyser (OCR + Gemini API), MRI image analyser (CNN), and an interactive DICOM viewer.
JeevVeda aims to shorten time-to-diagnosis and improve diagnostic accuracy by providing clinicians and patients with a compact, end-to-end screening toolkit:
- Medical Assistance Chatbot — NLP-based symptom intake and risk estimation (Low / Medium / High).
- Blood Report Analyzer — OCR + parsing pipeline to extract lab values and forward them to Gemini 2.5 Pro API for cancer risk analysis.
- MRI Image Analyzer — CNN-based (ResNet50) model to highlight suspicious nodules and provide malignancy probabilities.
- DICOM Viewer — Browser-based canvas viewer for interactive review of medical images (window/level, zoom, pan, multi-frame navigation).
This repo contains the code and resources for prototype development and research experiments.
├── public/ # Static assets (icons, svg files)
├── src/
│ ├── app/ # Next.js app router
│ │ ├── api/ # API routes (server actions)
│ │ │ ├── assess-risk/route.ts # Symptom risk scoring
│ │ │ ├── blood-analyzer/route.ts # OCR + Gemini blood report analyzer
│ │ │ ├── chat/route.ts # Chatbot interaction
│ │ │ └── users/ # User auth (login, signup, etc.)
│ │ ├── dashboard/ # Dashboard pages for different modules
│ │ │ ├── blood-analyzer/page.tsx
│ │ │ ├── chatbot/page.tsx
│ │ │ ├── dicom-viewer/page.tsx
│ │ │ ├── mri-analysis/page.tsx
│ │ │ └── ... (reports, screening tools, etc.)
│ │ ├── login/page.tsx
│ │ ├── signup/page.tsx
│ │ ├── layout.tsx
│ │ └── page.tsx
│ ├── components/ # Reusable UI + custom components
│ ├── dbConfig/ # Database configuration
│ ├── hooks/ # Custom React hooks (auth, mobile state)
│ ├── lib/ # Utility functions (DICOM parser, types, helpers)
│ └── models/ # Mongoose/DB models
├── BLOOD_ANALYZER_WORKFLOW.md # Workflow documentation for blood analyzer
├── CHAT_WORKFLOW.md # Workflow documentation for chatbot
├── SETUP_INSTRUCTIONS.md # Step-by-step setup
├── README.md # Project overview (this file)
└── ...
- Natural-language symptom intake with feature mapping.
- Tesseract OCR + PDF/image parsing for blood reports.
- Extracted values sent to Gemini 2.5 Pro API for risk assessment and insights.
- ResNet50 CNN pipeline (preprocessing → inference → localization overlays) for MRI scans.
- Next.js API routes powering the backend inside
/src/app/api/. - Interactive React dashboard with modules for chatbot, analyzer, MRI, and DICOM.
- Node.js 18+
- npm or yarn
# 1. Clone
git clone https://github.com/immohitsen/JeevVeda.git
cd JeevVeda
# 2. Install dependencies
npm install
# 3. Run dev server
npm run dev
# 4. Open browser
http://localhost:3000- MRI model: ResNet50-based CNN trained on annotated MRI/CT slices. Weights stored in
models/mri/or hosted in cloud storage. - Blood analysis: Relies on Gemini 2.5 Pro API for biomarker interpretation and risk scoring.
- Symptom model: Small LR/NN trained on symptom→diagnosis mapping.
Datasets: Keep sensitive/PHI data out of the repo. Provide synthetic/anonymized samples in data/sample/ for demos.
-
Use metrics: Accuracy, Precision, Recall, F1, AUC for classifiers.
-
For MRI detections: IoU, sensitivity, ROC curves.
-
Explainability:
- Gemini API structured outputs for blood analysis.
- Grad-CAM / CAM overlays for MRI model interpretability.
This project is a research / prototype tool. It is not a regulated medical device. For any clinical deployment:
- Obtain IRB / ethical approvals.
- Ensure PHI protection (HIPAA / local laws).
- Validate clinically with radiologists.
- Add disclaimers in UI/docs.
Planned enhancements for JeevVeda include:
- Multi-modal AI fusion combining MRI, pathology, and blood biomarkers.
- Real-time DICOM collaboration and annotation tools.
- Federated learning support for privacy-preserving model training.
- Deployment optimization for edge devices and low-resource hospitals.
- Advanced explainability pipelines using Grad-CAM++ and attention visualization.
- Secure patient report storage with encryption and audit logging.
- Expanded support for additional cancer screening modalities.
We welcome contributions from developers, researchers, and healthcare professionals interested in AI-driven medical diagnostics.
- Fork the repo
- Create a feature branch (
git checkout -b feat/your-feature) - Commit and open a PR with description + tests
- We’ll review and merge
- Mohit Sen — Fullstack / AI
- Anurag Pandey — ML / Prompt engineering
- Dr. J. Satya Eswari — Faculty advisor
MIT License — see LICENSE file.
JeevVeda is building an AI-first, accessible cancer diagnostic toolkit. Early detection saves lives.