Authors:

Atin Dewan

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

Surgical complications make a major contribution to global death rates, and surgical proficiency has a direct impact on postoperative outcomes. This study aims to develop computational models identifying and classifying surgical tasks such as knot tying and suturing in open surgery. In this study, we will apply advanced deep learning and computational video analysis techniques to over 2,000 video-based technical skill assessment clips from the IU School of Medicine, including YOLO (You Only Look Once) CNN (Convolutional Neural Network) versions. The video dataset comprises four frequently occurring surgical tasks (Two Handed Square Knot, One-handed Half-hitch Slip Knot, Simple Interrupted with Instrument Tie and Deep Dermal Suture with Instrument Tie) and are, on average, around 70 seconds in duration. For task classification, we will use YOLO models to detect and classify the various surgical actions. The preprocessing steps include background rejection, and we will investigate model performance at varying frame rates as well as spatial and temporal analysis of video clips to highlight particularly informative video sections and areas in frames. These steps are essential to enhance model accuracy by ensuring relevant features are highlighted and irrelevant information is minimized. Overall, we expect this research to facilitate automatic, scalable task detection and classification, enabling future development of intraoperative decision support models. We expect this work to provide a basis for intraoperative surgical decision support models with potential applications across patient care and residency training. Keywords: Surgical Task Classification; Deep Learning; Video Analysis; Intraoperative Decision Support

Source:

Purdue University / 2024

Topics:

No topics listed

Co-authors:

Atin Dewan