Prostate Anatomies and Lesion Automatic Segmentor

Whole-gland, zonal & lesion prostate segmentation from multi-parametric MRI. Cascaded nnU-Net v2 (anatomies) + ProLesA-Net (lesions). A three-stage pipeline that segments the whole gland (WG), peripheral zone (PZ), transition zone (TZ) and (when ADC + DWI are provided) the prostate lesions from prostate MRI. The anatomies are produced by two cascaded nnU-Net v2 models on T2; the lesion mask is produced by ProLesA-Net (Keras / TensorFlow 2.11) operating on T2 + ADC + DWI inside the WG mask

Topics

Collections

Details

License:
EUPL-1.2
Maturity:
Emerging
Cost:
Free of charge
Tool Type:
workflow
Operating Systems:
Linux, Mac, Windows
Programming Languages:
Python
Added:
4/30/2026
Last Updated:
9/24/2026

Operations

Data Inputs & Outputs

Image annotation

Inputs

Outputs

Publications

Documentation

Quick start guide
https://github.com/dzaridis/Prostate-Anatomies-and-Lesion-Segmentor
Quick Start Commands for the Software, including pulling from Harbor Container Registry, input folders structure
General
https://drive.eucaim.cancerimage.eu/s/HEKfYpL9DrCqJ84
Video with input folders structure, docker image execution & visualisation

Downloads

Links