Computational Methods to Analyse Intra-operative Adverse Events in Surgery at Scale

Projektinformationen

CompSURG

ID Finanzhilfevereinbarung: 101088553

DOI

10.3030/101088553

EK-Unterschriftsdatum 31 Juli 2023

Startdatum 1 November 2023

Enddatum 31 Oktober 2028

Finanziert unter

• European Research Council (ERC)

Gesamtkosten

€ 1 951 931,00

EU-Beitrag

€ 1 951 931,00

Koordiniert durch

UNIVERSITE DE STRASBOURG
France

Periodic Reporting for period 1 - CompSURG (Computational Methods to Analyse Intra-operative Adverse Events in Surgery at Scale)

Berichtszeitraum: 2023-11-01 bis 2026-04-30

Zusammenfassung vom Kontext und den Gesamtzielen des Projekts

The operating room is the most frequent location for hospital-related errors leading to increased morbidity, mortality, and costs. However, intra-operative adverse events (IAEs) are seldom well documented, which impedes their thorough root-cause analysis, the definition of appropriate safety measures, and the development of intra-operative support systems to reduce their occurrence. Recent manual video-based assessments of surgical procedures have shown that not only are IAEs frequent, but also that intra-operative near miss events, previously thought to be inconsequential, are in fact predictors of major errors and correlate with complications and poor surgical outcomes. CompSURG leverages these recent findings to propose a radically new, computational approach to improve intra-operative surgical safety.

The CompSURG project proposes to focus on automatically detecting and analyzing surgical variability and IAEs in endoscopic videos via novel computer vision methods that model the surgical tool-tissue interactions and infer the quality of intra-operative events, as needed to study and prevent the activity patterns leading to poor outcomes. It also aims to introduce new machine learning training methods for scaling these approaches to different types of surgeries and centers using a limited set of manual annotations.

These computational methods will allow the automated reporting and analysis of surgical activities and IAEs at a scale unfeasible with manual methods. They will be used to analyze patterns of IAEs, to identify the activities and phases of the surgical procedures that would benefit from new safety measures, and to define and implement best practices for surgical safety.

Arbeit, die ab Beginn des Projekts bis zum Ende des durch den Bericht erfassten Berichtszeitraums geleistet wurde, und die wichtigsten bis dahin erzielten Ergebnisse

A collaboration with 7 international hospitals has been established to generate a multi-centric, multi-procedure dataset annotated with tool-tissue interaction semantics and IAEs, as required to develop the CompSURG methods. CompSURG has also built tools for surgical data collection at scale, obtained ethical approval for a prospective observational study, and started to collect the multimodal data required for the project.

In parallel, CompSURG has developed methods to learn representations from surgical video lectures, a new endeavor aimed at reducing the need for annotations. The project has demonstrated that these methods can serve to build generalist models able to tackle different types of recognition tasks on surgical videos and on data from different procedures and hospitals in a zero-shot setting, namely without any manual annotation.

CompSURG is currently defining with the team of clinical collaborators several new clinical questions that will be tackled with the CompSURG AI approaches on the collected dataset. CompSURG is also developing a methodology based on graph and video-languages models to analyze vast amount of surgical data.

Fortschritte, die über den aktuellen Stand der Technik hinausgehen und voraussichtliche potenzielle Auswirkungen (einschließlich der bis dato erzielten sozioökonomischen Auswirkungen und weiter gefassten gesellschaftlichen Auswirkungen des Projekts)

The initial work on unsupervised learning from surgical video lectures has opened a new research direction and provided several comparative baselines already used in the Surgical Data Science community. CompSURG’s current work aims to define new safety measures and AI-based intra-operative support for different types of procedures, and to build a methodology to analyze the vast amount of video data with limited supervision.