Multi-Camera Hand Motion Analysis For Puncture Technique Training

• Main Authors: Li, Zhe, Kanazuka, Aya, Hojo, Atsushi, Hara, Yushin, Nomura, Yukihiro, Nakaguchi, Toshiya
• Format: Conference Proceeding
• Language: English
• Subjects: 3D reconstruction; Accuracy; Clinical injection training; MediaPipe hand; Motion analysis; Motion capture; Non-contact; Robustness; Solid modeling; Three-dimensional displays; Tracking; Training; Wearable sensors
• ISSN: 2694-0604
• DOI: 10.1109/EMBC53108.2024.10782622

Injection skill training is a crucial element in the clinical medical curriculum for medical students. However, existing feedback methodologies are predominantly based on subjective evaluations by physicians, which lack critical objective measures for assessing the precision and dependability of injection technique. Additionally, implementing wearable sensors is exorbitantly costly, and their usage can negatively influence the students' movements and performance due to the discomfort and encumbrance they cause. To address these challenges, our research presents a novel MediaPipe Hand-based system for non-contact tracking of finger movements during simulated injection training. This system employs four sets of cameras positioned at various angles to track medical students' injection movements, effectively resolving the complex issue of finger occlusion. It also reconstructs a 3D model of the hand during the injection process. To ascertain the accuracy of our system, we used a high-precision motion capture system (VICON) as a standard. Data on hand injection techniques were gathered from ten medical students. The system's robustness was verified by comparing the injection finger joint lengths, while the system's precision was evaluated by contrasting the finger joints' positions. The methodology proposed in this study can provide objective, accurate feedback on the trainees' techniques to enhance their injection skills and elevate the standard of clinical injection training.