Hevi AI Prostate Zonal Segmentation

Logo for Hevi AI Prostate Zonal Segmentation

About

Creators:

User Mugshot krgz.ahmet1 

User Mugshot joeran.bosma 

Contact email:
krgz.ahmet@hotmail.com
Version:
fc63b5a0-0757-422a-9f58-a0c6b4fdced7
Last updated:
Feb. 26, 2023, 12:29 p.m.
Associated publication:
Karagoz A, Seker ME, Yergin M, et al.. Prostate Lesion Estimation using Prostate Masks from Biparametric MRI. arXiv. Published online January 25, 2023.
Inputs:
  • Transverse T2 Prostate MRI  (Transverse T2 MRI of the Prostate)
Outputs:
  • Prostate Zonal Segmentation  (Zonal segmentation of the prostate where 1 is the Peripheral Zone, 2 is the Central Gland, 3 is the Prostatic Urethra, and 4 is the Anterior Fibromuscular Stroma.)
  • Softmax Prostate Peripheral Zone Segmentation  (Single-class heatmap of prostate peripheral zone segmentation, with each voxel indicating the likelihood in the range [0, 1])
  • Softmax Prostate Central Gland Segmentation  (Single-class heatmap of prostate central gland segmentation, with each voxel indicating the likelihood in the range [0, 1])

Model Facts

Summary

This model segments the central zone + anterior stroma + transition zone (TZ), and peripheral zone (PZ). The model requires a NIFTI MR image as input and creates a probability map and segmentation mask as output. The model is based on nnUnet.

Mechanism

This algorithm is a deep learning-based model (nnU-Net model). We trained these models with a total of 204 prostate T2 MRI scans paired with a manual prostate segmentation. These scans were sourced from ProstateX. Annotations for ProstateX cases were retrieved from https://github.com/rcuocolo/PROSTATEx_masks/.

Input:

  • Datatype: Transverse T2 MR image of the prostate
  • File format: NIFTI image
  • Target group: Male patients

Output:

  • Results: Prostate zonal segmentation (segmentation), Softmax prostate peripheral zone segmentation (probabilistic map), Softmax prostate central zone segmentation (probabilistic map)

  • Source code: https://github.com/ahmetkrgztr/HeviAI_picai/tree/main/AbdomenMRUS-prostate-segmentation

Validation and Performance

Uses and Directions

  • This algorithm was developed for research purposes only. This algorithm is intended to be used only on prostate T2 MRI examinations.
  • This model is intended to be used by radiologists for predicting prostate volume in biparametric MRI examinations. The model is not meant to guide or drive clinical care.

Warnings

  • Even if used appropriately, clinicians using this model can estimate prostate volume incorrectly.
  • This model is not designed to guide clinical diagnosis and treatment for prostate cancer.

Common Error Messages

Information on this algorithm has been provided by the Algorithm Editors, following the Model Facts labels guidelines from Sendak, M.P., Gao, M., Brajer, N. et al. Presenting machine learning model information to clinical end users with model facts labels. npj Digit. Med. 3, 41 (2020). 10.1038/s41746-020-0253-3