Related Experiment Video
Updated: Nov 21, 2025
Author Spotlight: Bridging Gaps in Anatomy and Establishing a Foundation for Algorithmic Studies
Published on: December 15, 2023
Automated joint skull-stripping and segmentation with Multi-Task U-Net in large mouse brain MRI databases
Riccardo De Feo1, Artem Shatillo2, Alejandra Sierra3
1Sapienza Università di Roma, Rome 00184, Italy; Centro Fermi-Museo Storico della Fisica e Centro Studi e Ricerche Enrico Fermi, Rome 00184, Italy; A.I. Virtanen Institute for Molecular Sciences, University of Eastern Finland, Kuopio 70210, Finland.
We developed Multi-task U-Net (MU-Net), a deep learning tool for automated mouse brain MRI segmentation and skull-stripping. MU-Net significantly improves accuracy and speed compared to manual methods, aiding preclinical research.
Area of Science:
- Neuroimaging
- Computational Biology
- Medical Image Analysis
Background:
- Manual skull-stripping and region segmentation in preclinical MRI are time-consuming and prone to variability.
- Automated methods are needed to improve efficiency and consistency in analyzing mouse brain MRI data.
Purpose of the Study:
- To introduce Multi-task U-Net (MU-Net), a novel convolutional neural network for simultaneous skull-stripping and brain region segmentation in mouse MRI.
- To evaluate MU-Net's performance against existing methods and assess its robustness on diverse datasets.
Main Methods:
- Development of MU-Net, a U-Net architecture designed for multi-task learning.
- Training and validation on 128 T2-weighted mouse MRI volumes and the MRM NeAT dataset.
- Testing on a large dataset of 1782 mouse brain MRI volumes from healthy and Huntington's disease models.
Main Results:
- MU-Net achieved high segmentation accuracy with Dice scores of 0.906 (striata), 0.937 (cortex), and 0.978 (brain mask).
- Inference time was rapid at 0.35 seconds per volume, with no pre-processing required.
- Performance was robust across different architectural variations and animal age ranges.
Conclusions:
- MU-Net is an effective and efficient tool for automated mouse brain MRI segmentation and skull-stripping.
- The method reduces inter- and intra-rater variability compared to manual segmentation.
- Publicly available code and model facilitate adoption in preclinical research.
