ISLES_Deploy

ISLES Stroke Lesion Segmentation System

This repository now targets ISLES-2022 style ischemic stroke lesion segmentation workflows.

What This Repo Does

• Upload DICOM or standard image slices
• Run an ISLES-style ensemble segmentation pipeline (runtime surrogate)
• Return lesion mask, overlay visualization, lesion pixels, and infarct volume estimate
• Provide metadata and health endpoints for deployment

Main API

• POST /upload: segmentation inference response
• GET /health: runtime health and dataset context
• POST /convert-to-jpeg: convert DICOM to JPEG

Expected /upload Response Fields

• image_path
• mask_path
• overlay_path
• infarct_volume_ml
• lesion_pixels
• confidence
• metadata (patient_id, study_date, modality, shape)

Dataset Context

• Current target dataset: ISLES-2022
• Task: Ischemic Stroke Lesion Segmentation
• Modality focus: DWI/ADC/FLAIR-aligned lesion analysis workflows

Run Locally

python -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txt
python app.py

Then open http://127.0.0.1:5000.

Notebook

Use isles-2022-dataset-analysis.ipynb for dataset EDA and lesion-intensity analysis.

This site is open source. Improve this page.