A hybrid transformer-deep learning model for improved cardiac MRI left ventricle segmentation

Author Identifier

Syed Mohammed Shamsul Islam: https://orcid.org/0000-0002-3200-2903

Md Moniruzzaman: https://orcid.org/0000-0003-1130-7078

Document Type

Conference Proceeding

Publication Title

Proceedings - 2024 25th International Conference on Digital Image Computing: Techniques and Applications, DICTA 2024

First Page

435

Last Page

441

Publisher

IEEE

School

School of Science

Comments

Islam, K. T., Shamsul Islam, S. M., Moniruzzaman, M., & Ihdayhid, A. (2024). A hybrid transformer-deep learning model for improved cardiac MRI left ventricle segmentation. In Proceedings of the 2024 International Conference on Digital Image Computing: Techniques and Applications (DICTA) (pp. 435-441). Perth, Australia. https://doi.org/10.1109/DICTA63115.2024.00070

Abstract

Cardiac left ventricle segmentation is crucial in diagnosing and managing various cardiac diseases. Accurate and efficient segmentation c an leadtop recise e valuations of left ventricular size, shape, and function, ultimately enhancing clinical outcomes. With the advent of deep learning, transformer architectures have emerged as powerful tools for medical image segmentation. This study introduces an innovative hybrid architecture that combines the strengths of the transformer model with established deep-learning models. Our approach demonstrates superior performance in cardiac left ventricle segmentation from MRI images, achieving near-perfect accuracy and significantly outperforming traditional models in both Dice coefficient and Intersection over Union (IoU) metrics. These results confirm the model's ability to capture complex image dependencies while maintaining class balance, marking a notable advancement in cardiac MRI segmentation.

DOI

10.1109/DICTA63115.2024.00070

Access Rights

subscription content

Link to Full Text

Share

 
COinS