Video and accelerometer-based motion analysis for automated surgical skills assessment

Area of Science:

• Medical Education
• Biomedical Engineering
• Computer Science

Background:

• Basic surgical skills like suturing and knot tying are crucial for medical training.
• Automated assessment systems can optimize training efficiency and expert time allocation.
• Existing automated methods utilize video or accelerometer data, with varying success.

Purpose of the Study:

• To present a novel approach for automated assessment of Objective Structured Assessment of Technical Skills (OSATS)-like surgical skills.
• To analyze the effectiveness of different features, including novel entropy-based ones, on multi-modal data (video and accelerometer).

Main Methods:

• A large dataset of video and accelerometer data for suturing and knot-tying tasks was collected.
• Introduced "entropy-based" features (approximate entropy and cross-approximate entropy) to quantify time-series data predictability.
• Compared proposed features against existing methods like Sequential Motion Texture, Discrete Cosine Transform, and Discrete Fourier Transform.

Main Results:

• Entropy-based features achieved high accuracies: 95.1% for suturing and 92.2% for knot tying using video data.
• The method showed improved performance for suturing with accelerometer data (86.8% accuracy).
• Fusing video and accelerometer features further enhanced overall skill assessment performance.

Conclusions:

• Automated surgical skill assessment with high accuracy is feasible using proposed entropy features.
• This automated system holds significant potential to improve surgical training efficiency in medical institutions.