GPU accelerated non-rigid registration for the evaluation of cardiac function.

Area of Science:

• Medical Imaging
• Computational Biology
• Cardiovascular Research

Background:

• Accurate tracking of cardiac motion is crucial for diagnosing cardiovascular diseases.
• Current methods for analyzing cardiac magnetic resonance (CMR) images can be time-consuming.
• Non-rigid registration techniques are essential for precise motion analysis in dynamic imaging.

Purpose of the Study:

• To develop and validate a fast and efficient GPU-accelerated method for tracking cardiac motion in CMR cines.
• To assess the accuracy and speed of the proposed method compared to expert manual contouring.

Main Methods:

• Implemented a GPU-accelerated Levenberg-Marquardt non-linear least squares optimization for finite element non-rigid registration using OpenGL.
• Employed forward and backward incremental registration for frame-to-frame point tracking.
• Tracked endocardial and epicardial borders in short-axis CMR cines from 36 patients with vascular disease, using expert semi-automatic ventricular analysis as the gold standard.

Main Results:

• The GPU-accelerated method achieved a processing time of 0.5 seconds per frame.
• Maximum Hausdorff errors were less than 2 mm on average, comparable to inter-observer variability among experts.
• Demonstrated high accuracy and efficiency in tracking cardiac motion in CMR images.

Conclusions:

• GPU acceleration of Levenberg-Marquardt non-linear optimization significantly enhances the speed and accuracy of cardiac motion tracking in CMR.
• This method provides a valuable tool for efficient and reliable analysis of cardiovascular dynamics.
• The developed technique shows potential for improving clinical workflows in cardiac imaging analysis.