Multimodal deep learning for objective skill assessment in robot-assisted vesico-urethral anastomosis

Comprehensive Summary

With increasing advancements in medical technology, robot-assisted surgery (RAS) has become increasingly more prominent. This study aimed to assess skill levels in using RAS (inexperienced, competent, and experienced) when performing vesico-urethral anastomosis (VUA) through Convolutional Neural Networks (CNNs) and Long-Short-Term Memory (LSTM) models. Data analyzed were composed of electroencephalogram (EEG) and eye-tracking signals during VUA subtasks of (1) needle grasping, positioning, and entry and (2) needle driving with wrist rotation and suture pull-out. Results indicated that multimodal deep learning using EEG and eye-tracking data allowed for objective classification of skill-level during robot-assisted VUA. However, assessing skill classification was task-dependent, highlighting the importance of tailoring assessment tools to corresponding surgical subtasks.

Outcomes and Implications

Technological tools with the ability to objectively assess robot-assisted surgery performance gives rise to the ability to continually learn and improve on surgical procedures. By assessing the experience of VUA performed with RAS, surgeons and health professionals can identify areas that require further practice and attention in order to minimize surgical risks and improve patient outcomes.