MyoViTs-Net: Enhanced Vision Transformer-Based Segmentation Network for Heart Segmentation Using Cine-MRI Dataset

Recommended citation:

         @inproceedings{Sehar2026MyoViTs, AUTHOR = {Uroosa Sehar and Nouman Ahmad and Jianhua Zhou and Zeeshan Tariq}, TITLE = {MyoViTs-Net: Enhanced Vision Transformer-Based Segmentation Network for Heart Segmentation Using Cine-MRI Dataset}, BOOKTITLE = {International Conference on Machine Learning and Artificial Intelligence Applications (MLAIA 2025)}, PUBLISHER = {SPIE}, VOLUME = {14134}, YEAR = {2026}, PAGES = {246--250}, ABSTRACT = {Cardiovascular disease (CVD) remains one of the leading causes of death worldwide, making early diagnosis and treatment critically important. Automated cardiovascular disease diagnosis using artificial intelligence techniques, particularly computer vision and medical image analysis, has become an effective approach to assist clinicians. Accurate heart segmentation plays a crucial role in enabling reliable cardiac analysis and surgical planning. However, existing deep learning-based segmentation approaches still struggle with capturing the complex three-dimensional structure of the heart. In this work, we propose MyoViTs-Net, an enhanced vision transformer-based segmentation network designed to improve heart segmentation performance using cine-MRI datasets. The proposed architecture leverages the global context modeling capability of vision transformers while preserving spatial details necessary for medical image segmentation. By effectively capturing the 3D spatial characteristics of cardiac structures, the model aims to achieve improved segmentation accuracy and robustness compared with traditional convolutional neural network-based methods. Experimental evaluation demonstrates that the proposed framework effectively handles the complex spatial structure of the heart and provides improved segmentation performance for cine-MRI data, supporting more reliable automated cardiac analysis and clinical decision-making.} }