Automated Surgical-Phase Recognition Using Rapidly-Deployable Sensors

Surgical-phase recognition is important to many future applications in clinical care, from building context-aware operating rooms to automatically providing feedback to surgeons in training. In this work, we focus on learning-based phase recognition in laparoscopic gallbladder removals (cholescystectomies). Using data from 15 sensors across 42 surgeries, we 1) compare performance using support vector machines, hidden Markov models, and conditional random fields and 2) demonstrate that it is possible to achieve 74% accuracy using only 8 rapidly-deployable sensors.     «

Surgical-phase recognition is important to many future applications in clinical care, from building context-aware operating rooms to automatically providing feedback to surgeons in training. In this work, we focus on learning-based phase recognition in laparoscopic gallbladder removals (cholescystectomies). Using data from 15 sensors across 42 surgeries, we 1) compare performance using support vector machines, hidden Markov models, and conditional random fields and 2) demonstrate that it is...     »