Surgical Suturing Skill Assessment Using Estimated Hand Roll Angle from a Deep-learning Computer Vision Algorithm.

Area of Science:

• Medical technology
• Computer vision
• Surgical education

Background:

• Assessing surgical suturing skills traditionally relies on subjective evaluation or specialized equipment like IMUs.
• Quantifying hand roll reversals (number of rolls) is a key metric for suturing proficiency.
• Existing methods for measuring hand roll angles may be invasive or limited in scope.

Purpose of the Study:

• To develop and validate a deep-learning computer vision algorithm for estimating hand roll angles during surgical suturing.
• To enable metric-based assessment of surgical suturing skills using non-contact video analysis.
• To evaluate the algorithm's accuracy and its ability to differentiate surgical skill levels.

Main Methods:

• A deep-learning algorithm was trained on hand roll estimation using videos from the SutureCoach simulator.
• The algorithm's accuracy was tested for robustness against varying video backgrounds.
• The 'number of rolls' metric was calculated using the computer vision algorithm and compared to IMU-based calculations.

Main Results:

• The deep-learning algorithm demonstrated robust accuracy in estimating hand roll angles across different video backgrounds.
• The 'number of rolls' metric derived from computer vision significantly differentiated between various surgical skill levels (novices, residents, attending surgeons).
• This pattern of skill differentiation was consistent with results obtained using IMU data.

Conclusions:

• The proposed computer vision algorithm offers a reliable, non-contact method for estimating hand roll angles in surgical suturing.
• This technology facilitates objective, inter-operative assessment of surgical skills.
• The findings support the integration of computer vision for enhanced surgical training and evaluation.