Artificial intelligence software available for medical devices: surgical phase recognition in laparoscopic cholecystectomy

Figures

• 

  Fig. 1. Analysis of the duration of the surgical phases in 115 LC cases. The duration differed for each phase and varied strongly between cases. The duration was 36.9 ± 19.8 min for P2, which was the longest surgical process, and 1.8 ± 2.1 min for P6, which was the shortest surgical process

• 

  Fig. 2. Schematic diagram of a representative transition in the surgical phase. The colors show each surgical phase. The horizontal axis of the color bar shows the time course of surgery, indicating the transition in surgical phase for each time point

• 

  Fig. 3. Transitions in surgical phases over time for ground truth, offline testing, and online testing. The left-hand column shows the order from the highest recall value in the nine test datasets. The right-hand column shows the order from the lowest recall value in the nine test datasets. The postprocessing results show inference by the artificial intelligence (AI) model, inference accuracy, and mode algorithm. The original results show inference from the AI model using EfficienteNet-B7 and SAM optimizer. The ground truth results show the time-dependent transition of the correct surgical process