Extrapolation of incomplete marker tracks by lower rank approximation

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Computational Biology

Background:

• Measuring object motion and deformation, such as the heart, often relies on tracking optical or radiopaque markers.
• Experimental challenges like marker occlusion or poor contrast can lead to fragmented, unidentifiable marker tracks.

Purpose of the Study:

• To develop and evaluate a method for extrapolating incomplete marker tracks using available track data.
• To improve the accuracy and continuity of motion and deformation measurements in scenarios with marker detection failures.

Main Methods:

• A novel extrapolation method using iterative lower-rank matrix fitting to noisy, incomplete marker tracks.
• Performance evaluation using computer-simulated data and animal experimental data with artificially removed track segments (3-44% of track length).

Main Results:

• For simulated data, the root mean square (RMS) error of extrapolations closely matched the noise level.
• For animal experimental data (256x256 pixels), the RMS error was +/- 0.5 pixels, significantly smaller than the total marker excursion (20 pixels).
• The proposed method outperformed BMDPAM estimation, which yielded approximately twice the RMS error.

Conclusions:

• The developed method effectively extrapolates missing marker track data, even with significant track loss.
• This approach offers a robust solution for handling marker detection failures in motion analysis, enhancing data continuity and accuracy.
• The technique demonstrates superior performance compared to existing statistical methods for missing data estimation in this context.