Automated skills assessment in open surgery: A scoping review
Hawa Hamza, Dehlela Shabir, Omar M. Aboumarzouk, Abdulla Al‐Ansari, Khaled Shaban, Nikhil V. Navkar
Abstract
Surgical skills proficiency lowers the incidence of adverse clinical outcomes during surgeries. Artificial intelligence (AI) has been applied for surgical skills assessment, especially in the field of minimally invasive surgeries (MIS). This paves the way for integrating AI for skills assessment in open surgeries as well. An overview of its applications can inform the scientific community and facilitate further developments. In this scoping review , we present the open surgeries and clinical settings where AI-based skill assessment has been applied, the kind of surgical data acquired for the AI-based algorithms, and the types of AI-based models used for automated skills assessment. A total of 40 articles were identified and included. Majority of the articles focused on macrosurgical suturing (45 %, n = 18). Most of the studies acquired data by capturing surgeon's hands (50 %, n = 20). About 35 % utilized deep learning algorithms , specifically convolutional neural networks (CNN) ( n = 14). The assessment input for the automation algorithms were predominantly hand movement. Around 37.5 % ( n = 15) of the studies assessed algorithm performance using classification accuracy . In the review, we compare conventional methods such as statistical modeling and custom algorithms with the emerging AI-based approaches. We also explore the utilization of object detection and temporal information for surgical skills assessment. We highlight the progress in automated skills assessment during open surgery with advancements in sensor technology, and AI algorithms with high prediction accuracies. Further developments in data acquisition and processing methods are essential to facilitate clinical implementation of such technologies.