About MedVision

An explainable AI platform for medical-image research

MedVision demonstrates an end-to-end medical-imaging workflow: dataset preparation, model training, validation, API development, explainability, structured reporting, and frontend design.

Deep learning

DenseNet-121 models support chest multi-label analysis, musculoskeletal body-part detection, and abnormality estimation.

Explainability

Grad-CAM heatmaps show which image regions influenced model output without claiming lesion segmentation.

Responsible presentation

The platform separates model output from diagnosis and presents limitations throughout the experience.

Technology stack

PyTorch

Deep-learning model development and inference

FastAPI

Python API and model-serving backend

Next.js

Responsive web application and analysis interface

Grad-CAM

Visual explanation of model attention

Chest modeling uses CheXpert. Musculoskeletal modeling uses MURA. Displayed performance reflects this project's validation results and should not be interpreted as clinical performance.