Related Experiment Video
Updated: Jul 3, 2026
09:06
Whole-brain Segmentation and Change-point Analysis of Anatomical Brain MRI—Application in Premanifest Huntington's Disease
Published on: June 9, 2018
Multidisciplinary Consensus Prostate Contours on Magnetic Resonance Imaging: Educational Atlas and Reference Standard
Yuze Song1, Anna M Dornisch2, Robert T Dess3
1Department of Radiation Medicine and Applied Sciences, University of California San Diego, La Jolla, California; Department of Electrical and Computer Engineering, University of California San Diego, La Jolla, California.
Summary
This study created a prostate segmentation benchmark using expert consensus. The best AI tools show high accuracy, surpassing traditional methods, but physician review is still crucial.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Radiation Oncology
Background:
- Prostate segmentation for treatment planning lacks a reliable ground truth.
- Evaluating artificial intelligence (AI) algorithms for this task is therefore challenging.
Purpose of the Study:
- To establish an expert consensus benchmark dataset for prostate segmentation.
- To evaluate the performance of various AI tools against this gold standard.
Main Methods:
- Consensus prostate segmentations were developed by a panel of 4 experts on T2-weighted MRI from 68 patients.
- Six AI tools (3 commercial, 3 academic) were evaluated using Dice scores, surface distance, and volume error.
Main Results:
- Expert consensus segmentation was achieved for all 68 cases.
- AI tools demonstrated median Dice scores from 0.80 to 0.94.
- Mean surface error ranged from 1.3 to 2.4 mm, with volume errors from 4.3% to 31.4%.
Conclusions:
- An expert consensus benchmark for prostate segmentation was successfully established.
- Top AI tools exhibit accuracy exceeding that of radiation oncologists using CT scans.
- Physician oversight remains necessary to identify significant AI errors.
