Meeting Banner
Abstract #4284

Improve the Accuracy of Right Ventricle Segmentation in Cardiac Magnetic Resonance Images by UNet++

Chih-Wei Lin1 and Hsu-Hsia Peng2
1Interdisciplinary Program in Life Sciences and Medicine, National Tsing Hua University, New Taipei City, Taiwan, 2Institution of Biomedical Engineering and Environment sciences, Hsinchu, Taiwan

Synopsis

Keywords: Analysis/Processing, Machine Learning/Artificial Intelligence

Motivation: To improve the performance of deep learning based right ventricle (RV) segmentation in cine cardiac MRI, especially for images with low image quality.

Goal(s): To develop an efficient and accurate RV segmentation by UNet++ in combination with different encoders.

Approach: The UNet++ model was combined with encoders of EfficientNet-b5, EfficientNet-b4, and ResNet50, and optimized with a combined Dice Loss and BCEWithLogitsLoss.

Results: The model achieved a high DSC in both validation and testing sets, outperforming traditional UNet methods, especially in images with low-quality.

Impact: The UNet++ model combined with special encoders provided a robust solution to RV segmentation, with the potential to enhance cardiac imaging analysis and support clinical decision-making by reducing manual intervention and increasing accuracy in heart function assessment.

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