Computer-Based Tracking of Microsurgical Instruments: A Novel Assessment Tool for Robot-Assisted and Conventional...

Efficient and objective tools for self-assessment of microsurgical skills are needed to ensure high-quality microsurgical training and optimized use of surgeons' time and resources. In addition, the successful clinical integration of microsurgical robots in operating rooms will critically depend on effective training and evaluation strategies for microsurgeons, necessitating the development, usability testing, and validation of such assessment tools for both conventional and robotically assisted microsurgery.

Two deep convolutional neural network-based computer algorithms were developed to enable automated tracking of conventional and robotic microsurgical instruments. To train these models, supervised and semisupervised learning was applied to 84 microsurgical training videos, and the results were statistically analyzed using t tests, ANOVA, linear regression, and correlation analyses.

Computer algorithms that automatically track conventional and robotic microinstruments in recorded microsurgical training videos were developed. The total trajectory length showed a positive correlation with procedure time and Structured Assessment of Microsurgical Skill scores, reflecting operative efficiency and flow. Both procedure time and total trajectory length of robot-assisted procedures were significantly longer among experienced microsurgeons compared with the conventional approach, but not among microsurgical beginners. The mean deviation intensity, quantifying hand tremor throughout microsurgical performances, was significantly lower with the robot-assisted compared with the conventional microsurgical approach across all experience levels.

The proposed computer algorithms address critical gaps in objective microsurgical skill assessment, enabling accessible, efficient, and quantitative self-evaluation, and allow for direct comparison of robot-assisted and conventional microsurgical performances.