Artificial intelligence improves quality criteria in screening colonoscopy – a randomised, controlled, multicenter trial in a real world setting in daily clinical routine – ColonAI
In diesem bisher weltweit einzigartigen Studiendesign haben wir transsektoral in Kooperation mit den niedergelassenen Gastroenterologen der Region mehr als 2000 Patienten prospektiv, randomisiert untersucht und dabei den genauen Stellenwert der künstlichen Intelligenz bei der Vorsorgekoloskopie unter „real-world-Bedingungen evaluiert.
Als Endpunkte wurden dabei nicht nur die klassischen Kriterien der Polypen- und Adenomdetektionsrate, sondern diverse weitere Qualitätskriterien kritisch beleuchtet.
Basierend auf unseren Daten verbessert der Einsatz der künstlichen Intelligenz die Behandlungs- und Ergebnisqualität der Vorsorgekoloskopie signifikant und kann langfristig womöglich die Inzidenz des kolorektalen Karzinoms in der Bevölkerung reduzieren.
Background:
Recent studies demonstrated that artificial intelligence(AI) algorithms can increase the polyp/adenoma detection rates(PDR/ADR) in colonoscopy. Whether this integration translates into improved quality criteria remains unclear.
Aim:
To assess the key quality criteria of screening colonoscopy in a real-world, multicenter, randomized controlled trial, with/without the assistance of AI algorithms.
Methods:
Three private practices and one university hospital participated in this study. Individuals scheduled for colonoscopy were randomized to: (A): AI-assisted colonoscopy(AIAC): HDWLcolonoscopy+ GIGenius(Medtronic); (B, control): HDWL. PDR/ADR were defined as primary endpoints.
Results:
2,117 patients were included (AIAC=1,051; Control=1,066). The AIAC group demonstrated a PDR of 53.5% compared to 36.7% in the control group (p<0.0001; ADR: 43.9% vs. 27.3%, p<0.0001). In trainees PDR/ADR were lower compared to experts (p<0.0001), but improved significantly by AIAC. In the control group, withdrawal time significantly accelerated during the day. This difference vanished (p>0.05) in the AIAC group.
Conclusion:
AI algorithms significantly improve quality criteria of screening colonoscopy in a real-world setting potentially improving patient safety.
Ansprechpartner:
Univ.-Prof. Dr. med. Mark Ellrichmann
Universitätsklinikum Schleswig Holstein
Campus Kiel
Medizinische Klinik I
