https://atlas.rsna.org/schemas/2025-11/model.json

TotalSegmentator MRI: Robust Sequence-independent Segmentation of Multiple Anatomic Structures in MRI

https://doi.org/10.1148/radiol.241613

jakob.wasserthal@usb.ch

Authors declared no funding for this work.

This retrospective study was approved by the Ethics Committee of Northwest and Central Switzerland (EKNZ BASEC 2023–00446), and the requirement for informed consent was waived as this was a retrospective study.

This study developed and evaluated TotalSegmentator MRI, an open-source, easy-to-use segmentation model for automatic and robust segmentation of 80 major anatomic structures independent of MRI sequence. The model was trained on a diverse dataset of MRI and CT scans and demonstrated high accuracy and outperformed other publicly available models.

nnU-Net framework, a U-Net–based implementation. This framework automatically configures all hyperparameters based on the dataset characteristics.

The ready-to-use online tool is available at https://totalsegmentator.com; the model, at https://github.com/wasserth/TotalSegmentator; and the dataset, at http://zenodo.org/records/14710732.

Reduces radiologists’ workload, minimizes human errors, provides more consistent and reproducible results, and has clinical applications in treatment planning, disease progression monitoring, opportunistic screening, volumetric calculations of anatomic structures, and assessing organ size based on factors such as age, sex, and disease state.

Diagnostic processes, Treatment planning, Disease progression monitoring, Opportunistic screening

Fully automated

Robust segmentation of major anatomic structures independent of MRI sequence. Segmentation of 80 anatomic structures relevant for use cases such as organ volumetry, disease characterization, surgical planning, and opportunistic screening.

MRI and CT scans (3D datasets, 1.5-mm or 3-mm isotropic resolution)

The ready-to-use online tool is available at https://totalsegmentator.com; the model, at https://github.com/wasserth/TotalSegmentator.

Lower performance on MRI scans than on CT images due to greater heterogeneity in MRI acquisition protocols and scanner configurations, and lower spatial resolution and signal-to-noise ratio. Retrospective study design.

Can be easily integrated into existing clinical workflows and can operate in real time to assist radiologists during diagnostic processes. Can also be used in various research projects, for example, analyzing age-dependent changes in the volume of different abdominal structures.

The proposed open-source model and annotated training dataset are publicly available.