Related Experiment Video
Updated: Jun 19, 2026
11:17
A Microscopic 2,3,5-Triphenyltetrazolium Chloride Assay for Accurate and Reliable Analysis of Myocardial Injury
Published on: November 28, 2025
Myocardial Temporal-Mechanical Self-Supervision Model for Contrast-Free Myocardial Infarction Segmentation With
IEEE Transactions on Medical Imaging
|June 17, 2026
Summary
This study introduces MTMS, a novel contrast-free myocardial infarction segmentation model that bypasses the need for paired imaging data. MTMS effectively integrates cardiac biomechanics for improved segmentation accuracy without contrast agents.
Area of Science:
- Medical Imaging
- Cardiovascular Disease Analysis
- Artificial Intelligence in Medicine
Background:
- Contrast-free myocardial infarction (MI) segmentation is crucial for patient safety, avoiding risks associated with contrast agents (CAs).
- Existing methods struggle with inter-modality slice misalignments and require paired CINE and contrast-enhanced images, which are difficult to acquire.
Purpose of the Study:
- To develop the first label-free training and contrast-free MI segmentation model (MTMS) that does not require paired datasets.
- To incorporate cardiac biomechanical knowledge into contrast-free MI segmentation using a self-supervised paradigm.
Main Methods:
- MTMS utilizes dual upstream guidance: a Spatiotemporal Structural Evolution Module for deformation trajectories and a Cardiac Mechanics-Driven Analysis Module for stress responses.
- Self-supervised learning is achieved through iterative pseudo-label refinement, combining structural and biomechanical cues via a Dual-Domain Interaction Module.
Main Results:
- MTMS achieved a Dice score of 0.698 and HD95 of 19.634 on 370 clinical cases.
- The model outperformed seven state-of-the-art methods, showing significant improvements in Dice and HD95 metrics.
Conclusions:
- MTMS demonstrates the potential of integrating biomechanical knowledge into contrast-free MI segmentation through self-supervised learning.
- This approach advances the development of effective and safer MI segmentation techniques, reducing reliance on contrast agents.
