259 DEVELOPMENT OF AN ONTOLOGY FOR LAPAROSCOPIC TRANSABDOMINAL ADRENALECTOMY FOR SURGICAL TRAINING AND VIDEO ANALYSIS WITH MACHINE LEARNING ALGORITHMS AND ITS VALIDATION VIA A COMPREHENSIVE MODIFIED DELPHI SURVEY

Abstract

Abstract Background Although the safety of laparoscopic transabdominal lateral adrenalectomy (TLA) depends on the learning curve and experience of surgeons, surgical training remains a challenge as adrenalectomies are relatively rare. Structured laparoscopic video analysis supports individual training and could pave the way for automatic workflow recognition and safety alerts via Artificial Intelligence systems. We aimed to develop and validate a surgical ontology for TLA video analysis suitable for multicentric use. Methods An ontology hierarchically structured in phases and steps of laparoscopic right (RTLA) and left (LTLA) TLA was developed and submitted to a web-based survey via a 2-rounds modified Delphi process shared among 17 experienced adrenal surgeons across Europe. Consensus was defined at ≥ 80% agreement for each statement. Results RTLA was subdivided in 6 phases and 27 steps, and LTLA in 6 phases and 25 steps. Phases on both sides were: I) Preparation (5 statements), II) Exposition (8 statements), III) Dissection of the main adrenal vein (RTLA 14 statements, LTLA 12 statements), IV) Dissection of adrenal gland (6 statements), V) Extraction and disassembling (6 statements), and VI) any complementary intervention (1 statement). Panelist response rate was 88 % (15/17). Consensus was reached for LTLA in all 38 statements, and for RTLA in 39/40 statements (97,5%), as identification of the renal vein was not unanimously considered mandatory. Conclusions An expert consensus validated the developed ontology for TLA techniques as performed across Europe, suitable for multicentric video assessments and for surgical training and video analysis with machine learning algorithms.

Ähnliche Arbeiten