Page 109 - Read Online

Page 165 Choksi et al. Art Int Surg. 2025;5:160-9 https://dx.doi.org/10.20517/ais.2024.84

Figure 2. Illustration of the end-to-end system to predict trainee proficiency level.

sample size and more annotations to train our model.

This study was limited by the number of participants. With a larger video training set, we aim to be able to
improve our model to enhance accuracy. This dry lab model was also a preliminary model. The model can
be improved for better object and task recognition by CV. For example, utilizing a larger needle or larger
suture size may allow more differentiation between the background and the needle for the CV algorithm to
better pick up the needle movements. In the future, this could be assessed by a separate pre-assessment
model to determine how well CV is doing in object identification. With the rapid improvement in CV
models, this model could even be further optimized with the use of not only object identification but also
gesture recognition. Otiato et al. showed that surgical gestures during dissection can vary based on
[18]
experience and correlate with patient outcomes . Using a multi-modal model on a dataset with an