Automatic recognition of anatomical structures and surgical phases in robot-assisted minimally invasive esophagectomy (RAMIE) using deep learning: a retrospective cohort study - Summary - MDSpire
Advertisement
Automatic recognition of anatomical structures and surgical phases in robot-assisted minimally invasive esophagectomy (RAMIE) using deep learning: a retrospective cohort study
To develop real-time deep learning algorithms for anatomy and surgical phase recognition in the thoracic part of Robot-Assisted Minimally Invasive Esophagectomy (RAMIE).
Approach:
Study Design: A retrospective single-center cohort study was conducted at University Medical Center Utrecht, collecting operation recordings from patients who underwent transthoracic RAMIE for esophageal cancer.
Data Collection: Operation recordings were prospectively collected from January 2018 to July 2021, with a focus on the thoracic phase until the division of the esophagus.
Dataset Creation: Two datasets were created: one for anatomy segmentation and another for surgical phase recognition, with annotations performed by trained research fellows and reviewed by an expert surgeon.
Key Findings:
1504 frames were selected from 53 edited RAMIE videos for the anatomy segmentation dataset.
Twelve classes were annotated, including four surgical instruments and eight anatomical structures.
The annotation protocol was refined based on feedback from upper-gastrointestinal surgeons.
Interpretation:
The study highlights the potential for deep learning algorithms to assist in real-time anatomy recognition during RAMIE, which could enhance surgical orientation and reduce complications.
Limitations:
The study was limited to a single center and may not be generalizable.
The annotation process relied on a limited number of experts, which may affect the dataset's robustness.
Conclusion:
The development of real-time recognition algorithms could improve intraoperative guidance and potentially enhance surgical outcomes in RAMIE.
by Romy C. van Jaarsveld, Yiping Li, Ronald L. P. D. de Jong, Franco Badaloni, Gino M. Kuiper, Tim J. M. Jaspers, Marcel Breeuwer, Fons van der Sommen, Richard van Hillegersberg, Yasmina Al Khalil, Jelle P. Ruurda
