by
Gus Iversen, Editor in Chief | February 18, 2025
Researchers in Switzerland have developed an AI model capable of automatically segmenting MR images, regardless of imaging sequence.
According to a study published in
Radiology, the model, called TotalSegmentator MRI, outperformed other publicly available segmentation tools. A similar tool, TotalSegmentator CT, is already in use by over 300,000 users worldwide.
Manual segmentation — outlining anatomical structures — is time-consuming and varies between radiologists. “MR images have traditionally been manually segmented, which is a time-consuming process that requires intensive effort by radiologists, and is subject to inter-reader variability,” said Jakob Wasserthal, Ph.D., a research scientist in the radiology department at University Hospital Basel in Switzerland. “Automated systems can potentially reduce radiologists’ workload, minimize human errors and provide more consistent and reproducible results.”
Ad Statistics
Times Displayed: 99051
Times Visited: 6189 MIT labs, experts in Multi-Vendor component level repair of: MRI Coils, RF amplifiers, Gradient Amplifiers Contrast Media Injectors. System repairs, sub-assembly repairs, component level repairs, refurbish/calibrate. info@mitlabsusa.com/+1 (305) 470-8013
The researchers based TotalSegmentator MRI on nnU-Net, a self-configuring AI framework designed for medical image segmentation. For the study, the team trained the tool using a data set of 616 MR and 527 CT exams. The model segmented 80 anatomical structures, including organs, bones, and muscles, which are used for volume measurements, disease characterization, and surgical planning. Wasserthal noted that the model’s strength lies in its ability to adapt to different MR scanners and imaging settings.
The model’s performance was evaluated using Dice scores, which measure the similarity between AI-generated and radiologist-validated segmentations. It achieved an overall Dice score of 0.839 on an internal MR test set, outperforming two publicly available segmentation models. It also matched the performance of TotalSegmentator CT.
“To our knowledge, our model is the only one that can automatically segment the highest number of structures on MRs of any sequence,” Wasserthal said. “It’s a tool that helps improve radiologists’ work, makes measurements more precise and enables other measurements to be done that would have taken too much time to do manually.”
Beyond research applications, the AI model could potentially assist in clinical settings by aiding in treatment planning, monitoring disease progression, and identifying incidental findings through opportunistic screening.
