Inter-hospital transferability of AI: A case study on phase recognition in cholecystectomy

Published: 2025, Last Modified: 01 Mar 2026Comput. Biol. Medicine 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•Transfer models to new environments for surgical phase recognition based on public data.•High degree of heterogeneity is preferable to a lot of homogeneous data in training.•Combining public data in balanced fashion leads to significant improved performance.•Phase transitions are training depended due to ambiguous frames in that time.•xAI reveals manufacturer-specific instrument dependencies for model performance.
Loading