Meeting Banner
Abstract #4875

Multi-Task Learning with Hierarchical Label Consolidation for Efficient and Accurate Brain MRI Segmentation of 169 Regions

Pin-Chuan Chen1, Teng-Yi Huang1, Yi-Ru Lin2, Tzu-Chao Chuang3, and Hsiao-Wen Chung4
1Department of Electrical Engineering, National Taiwan University of Science and Technology, Taipei, Taiwan, 2Department of Electronic and Computer Engineering, National Taiwan University of Science and Technology, Taipei, Taiwan, 3Department of Electrical Engineering, National Sun Yat-Sen University, Kaohsiung, Taiwan, 4Department of Electrical Engineering, National Taiwan University, Taipei, Taiwan

Synopsis

Keywords: Gray Matter, Segmentation, brain

Motivation: Efficient brain MRI segmentation is crucial, but current models with many labels require high computational resources, limiting their practicality in resource-constrained settings.

Goal(s): To reduce computational demands without compromising accuracy by using label consolidation and multi-task learning in brain MRI segmentation models.

Approach: We consolidated 169 labels into 65 using hierarchical label consolidation and employed multi-task learning within a 3D U-Net framework, significantly reducing memory usage and processing time.

Results: Our model outperforms traditional models in accuracy, uses less GPU and CPU memory, and cuts CPU processing time by 87%, enabling faster processing on limited-resource systems.

Impact: The method enhances efficiency and accuracy in brain MRI segmentation, allowing integration of additional tasks like cortical thickness estimation, leading to a multifunctional open-source brain MRI processing toolbox.

How to access this content:

For one year after publication, abstracts and videos are only open to registrants of this annual meeting. Registrants should use their existing login information. Non-registrant access can be purchased via the ISMRM E-Library.

After one year, current ISMRM & ISMRT members get free access to both the abstracts and videos. Non-members and non-registrants must purchase access via the ISMRM E-Library.

After two years, the meeting proceedings (abstracts) are opened to the public and require no login information. Videos remain behind password for access by members, registrants and E-Library customers.

Click here for more information on becoming a member.

Keywords