Artificial Intelligence in Surgical Training: A New Study Highlights the Role of Machine Learning

A new study published in Laryngoscope Investigative Otolaryngology has investigated the use of artificial intelligence in surgical training, highlighting the potential of machine learning to provide automated feedback and assessment of surgical performance. The research, conducted at the University of Manitoba, aimed to develop a software classifier model that can predict the level of expertise of surgical trainees based on recorded drill trajectory and hand motion tracking.

Key Takeaways:

  • The study used machine learning to develop a software classifier model that can predict the level of expertise of surgical trainees with an accuracy of 92.8% and a sensitivity of 87.5%.
  • The classifier was able to detect stroke with a precision of 80.2%, 82.7%, and 84.8% during cortical mastoidectomy, thinning procedures, and facial recess exposure, respectively.
  • The study used a supervised classification approach to develop the classifier, which was trained on data from 3D-printed temporal bone models.
  • The research aimed to provide an essential first step towards an autonomous training paradigm in surgical education.
  • The study has significant implications for the development of artificial intelligence in surgical training, highlighting the potential of machine learning to provide automated feedback and assessment of surgical performance.

Statistics:

  • 92.8% accuracy in predicting the level of expertise of surgical trainees.
  • 87.5% sensitivity in detecting expertise.
  • 80.2%, 82.7%, and 84.8% precision in detecting stroke during cortical mastoidectomy, thinning procedures, and facial recess exposure, respectively.
  • 92.8% accuracy in predicting the level of expertise with a sensitivity of 87.5%.

Sources:

  • Classification in Virtual Temporal Bone Surgical Education: A First Step Towards Automated Virtual Education With Use of Machine Learning. Laryngoscope Investigative Otolaryngology, 2025,10(4):n/a-n/a.
  • http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)2378-8038
  • https://doi-org.sdpl.idm.oclc.org/10.1002/lio2.70188