riyaarahim/mri-tumor-segmentation-demo

MRI Brain Tumor Segmentation — Demo

Streamlit frontend for interactively running and visualizing brain tumor segmentation on MRI scans.

This is a thin client: it has no model weights or inference logic of its own. All preprocessing, inference, and image rendering happens server-side, via a separately deployed FastAPI service: riyaarahim/mri-tumor-segmentation-api.

Architecture

Frontend (this Space, Streamlit)
        │  HTTP requests
        ▼
API (mri-tumor-segmentation-api, FastAPI)
        │  loads checkpoint at startup
        ▼
Model (riyaarahim/mri-tumor-segmentation-unet, 3D U-Net)

Usage

Upload all 4 MRI modalities (FLAIR, T1, T1ce, T2 as .nii.gz) and click "Run Segmentation" to see tumor region statistics and a slice-by-slice visual overlay of necrotic core, edema, and enhancing tumor regions.