STCMT-Net: A spatiotemporal consistency motion transfer network for enhancing cardiac motion estimation

Area of Science:

• Medical imaging analysis
• Computational cardiovascular science
• Biomedical engineering

Background:

• Accurate cardiac motion estimation is vital for understanding heart function and mechanics.
• Existing methods struggle with complex cardiac anatomy and inter-image correlations, leading to inaccurate motion tracking.
• Four-dimensional (4D) cardiac imaging presents challenges in modeling intricate spatiotemporal dynamics.

Purpose of the Study:

• To develop an unsupervised Spatiotemporal Consistency Motion Transfer Network (STCMT-Net) for enhanced motion estimation in 4D cardiac images.
• To improve the modeling of cardiac anatomy and inter-structural correlations for more precise motion analysis.
• To provide a robust tool for clinical applications in cardiac condition assessment.

Main Methods:

• Proposed an unsupervised Spatiotemporal Consistency Motion Transfer Network (STCMT-Net).
• Utilized unsupervised keypoint detection for modeling cardiac motion and structural anchoring.
• Employed spatial distribution uniformity and temporal local anatomical correspondence constraints.
• Formulated dense motion reconstruction using basis motion vectors and residual refinements.

Main Results:

• STCMT-Net demonstrated enhanced accuracy in cardiac motion estimation across 4D computed tomography and cine magnetic resonance imaging datasets.
• The method effectively captured both global and fine-grained cardiac motions.
• Cardiac strain analysis using STCMT-Net effectively distinguished between patients with different cardiac conditions.

Conclusions:

• STCMT-Net offers a robust and accurate solution for unsupervised cardiac motion estimation in 4D imaging.
• The method shows significant potential for improving the clinical assessment of cardiac function and disease.
• Keypoint-based modeling and spatiotemporal consistency enhance the reliability of motion analysis.