2D human pose tracking in the cardiac catheterisation laboratory with BYTE

Area of Science:

• Medical Imaging and Computer Vision
• Healthcare Workflow Optimization

Background:

• Cardiac Catheterisation Laboratories (Cath Labs) generate valuable data for workflow analysis.
• Human pose tracking from video offers insights into laboratory efficiency and safety.
• Challenges in Cath Labs include occlusions and visual similarity of personnel, hindering accurate re-identification.

Purpose of the Study:

• To develop and evaluate a novel human pose tracker for Cardiac Catheterisation Laboratories.
• To address challenges of occlusion and personnel re-identification in Cath Lab environments.
• To assess the tracker's performance across different surgical phases and compare it with state-of-the-art methods.

Main Methods:

• A human pose tracker was developed using object keypoint similarity and a third-order motion model, excluding visual re-identification.
• The algorithm was tested on real coronary angiogram recordings from a Cath Lab.
• Performance was measured using Higher-Order Tracking Accuracy (HOTA) across five distinct surgical steps.

Main Results:

• The proposed tracker achieved up to 0.71 HOTA, outperforming state-of-the-art trackers (up to 0.65 HOTA).
• The tracker demonstrated more consistent performance across workflow phases (10 pp variation) compared to others (up to 23 pp).
• Achieved processing speeds of 22.5 frames per second, 9 fps faster than current methods.

Conclusions:

• The novel pose tracker is effective for Cath Lab workflow analysis, offering improved accuracy and consistency.
• Its speed and stability support real-time applications in healthcare settings.
• The publicly available code facilitates further research and development in this area.