Development and implementation of eye surgery (Capsulorhexis) skill assessment techniques on video data and its feedback

Master Thesis Defense

Development and implementation of eye surgery (Capsulorhexis) skill assessment techniques on video data and its feedback

ARAS-FARABI Capsulorhexis Video Dataset
Please Join me in the M.Sc. defense of ARAS master’s student Mohammad Javad Ahmadi, who has worked on “Development and implementation of eye surgery (Capsulorhexis) skill assessment techniques on video data and its feedback”. We accommodate both in-person and virtual attendance.
Date: Monday, September 19, 2022 (28 Shahrivar 1401)
Time: 16:30-18:00 (+4:30 GMT Tehran local time)
Event Place: K. N. Toosi Uni. of Tech., Faculty of EE, Room 212
Virtual attendance: https://lnkd.in/eJpxrf-D
More info: https://lnkd.in/eNMxinPb

It is essential to train trainee surgeons and transfer skills from experts to them to reduce medical risks and costs. Assessment of surgeons’ skills is one of the most critical aspects of their training. An automatic evaluation of cataract surgery videos and its most fateful process, capsulorhexis, is the purpose of this research. The reason for using only video data is that this data is available in almost all operating rooms and, unlike sensory data, does not require expensive equipment.

In collaboration with the surgeons of Farabi hospital, this thesis has developed a surgical skill evaluation structure specific to capsulorhexis surgery. As a result, a comprehensive annotated dataset was created for various medical and machine vision applications. This study combines deep learning, feature extraction, and optical flow with robotic knowledge of frequency space to assess video surgery skills automatically.The proposed structures have been evaluated on two other similar datasets, as well as some exploration of feature maps extracted from surgical videos to check accuracy and generalizability.

Although deep learning-based surgical skill assessment is considered a cutting-edge technology and a future medical need, it still remains a black box. To address this concern, this thesis analyzes the motion characteristics of capsulorhexis surgery videos, emphasizing tangible skill-based indicators. This motion information was derived from video and the tracker-oriented software developed in this research. This work has received significant attention and trust from surgeons because it introduces and analyzes concrete surgical skill indicators mapped their surgical skills to tangible engineering metrics.

Besides knowing the successful indicators for distinguishing surgeon skills, this mapping allows them to evaluate quantitatively and without bias what they previously evaluated subjectively and qualitatively. This thesis has developed the previous works of the ARAS research group in evaluating surgical skills on sensory and simulated datasets. The findings of this thesis have brought the ARAS research group closer to a more effective presence in surgical and surgical training processes. The overall outcome of this thesis is very promising, for further future collaboration and development in this field.