Efficacy of an artificial intelligence system for lesion detection and characterization ( CADe and CADx ) during colonoscopy following positive faecal immunochemical test in a colorectal cancer screening programme: A randomized clinical trial
Wirksamkeit eines KI-Systems zur Detektion und Charakterisierung von Läsionen (CADe und CADx) bei der Koloskopie nach positivem fäkalem immunchemischem Test in einem Darmkrebs-Screeningprogramm: randomisierte klinische Studie
Colorectal Disease · 2026-03 · 28 (3)
DOI: 10.1111/codi.70426 ↗Auf einen Blick
Bei FIT-positiven Personen verbesserte die CAD EYE™-Assistenz die Läsionsdetektion in der Koloskopie nicht signifikant. Die CADx-Charakterisierung zeigte jedoch höhere Spezifität und PPV.
Fragestellung
Verbessert ein KI-System mit CADe und CADx die Läsionsdetektion und -charakterisierung bei Koloskopien nach positivem FIT im Darmkrebs-Screening?
Methode
Monozentrische, randomisierte, parallelgruppige klinische Studie; Vergleich konventionelle Koloskopie vs. CAD EYE™-assistierte Koloskopie bei FIT-positiven Personen von 50–74 Jahren; Oktober 2023 bis Februar 2025.
Ergebnisse
361 Patient:innen analysiert. ADR 61,5 % mit KI vs. 69,8 % ohne KI (p = 0,097); fortgeschrittene kolorektale Neoplasie 37,9 % vs. 36,3 % (p = 0,753). CADx-Genauigkeit >85 %.
Schlussfolgerung der Autor:innen
Bei FIT-positiven diagnostischen Koloskopien mit hoher Ausgangs-ADR verbesserte KI die Läsionsdetektion nicht signifikant. Der Hauptnutzen liegt laut Autor:innen in der optischen Charakterisierung mit höherer Spezifität und PPV.
Kernaussagen
- ADR: 61,5 % mit KI vs. 69,8 % ohne KI
- Fortgeschrittene Neoplasie: 37,9 % vs. 36,3 %
- CADx: Spezifität 66,7 % vs. 51,9 %, PPV 92,4 % vs. 89,9 %
Deutsche Zusammenfassung: KI-gestützt aus dem Original-Abstract erstellt; alle Zahlen wurden automatisch mit dem Original abgeglichen. Maßgeblich ist der Original-Abstract.
Kurz gefasst
Randomisierte kontrollierte Studie in der Zeitschrift Colorectal Disease, online erschienen am 19.03.2026. Verfasst von Daniel Robles de la Osa und 17 weiteren Autor:innen. Beteiligte Einrichtung: Gastroenterology Department, University Care Complex of Palencia, Palencia, Spain. Hauptthemen laut MEDLINE: Colorectal Neoplasms, Colonoscopy, Early Detection of Cancer, Adenoma. Die Arbeit wurde bisher 6-mal zitiert (Crossref) und nennt 31 Literaturangaben.
Abstract (Originalsprache)
Aim
Artificial intelligence (AI) has emerged as a promising tool to enhance lesion detection (CADe) and characterization (CADx) during colonoscopy. However, its effectiveness in faecal immunochemical test (FIT)‐based colorectal cancer (CRC) screening remains controversial.
Method
This single‐centre, randomized, parallel‐group clinical trial compared conventional colonoscopy with CAD EYE™‐assisted colonoscopy in FIT‐positive individuals aged 50–74 years undergoing CRC screening between October 2023 and February 2025. The primary endpoints were the adenoma detection rate (ADR) and the advanced colorectal neoplasia detection rate.
Results
A total of 361 patients were analysed. ADR (61.5% with AI vs. 69.8% without AI, p = 0.097) and advanced colorectal neoplasia detection rate (37.9% vs. 36.3%, p = 0.753) did not differ significantly between groups. Similarly, detection rates stratified by lesion type, location, size or morphology, as well as the mean number of lesions per colonoscopy, were comparable between groups. Longer withdrawal time was associated with higher detection of advanced neoplasia (OR = 1.3; p < 0.001). The CADx system revealed diagnostic accuracy exceeding 85%, with greater specificity (66.7% vs. 51.9%) and positive predictive value (PPV) (92.4% vs. 89.9%) compared with endoscopist‐based optical diagnosis.
Conclusion
In FIT‐positive diagnostic colonoscopies with high baseline ADRs, AI assistance did not significantly improve lesion detection. Its primary utility lies in optimizing optical characterization by increasing specificity and PPV, suggesting a potential role in reducing false positives. Further multicentre studies incorporating cost‐effectiveness analyses are warranted ( ClinicalTrials.gov NCT07125300).
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Gastroenterology Department, University Care Complex of Palencia, Palencia, Spain.
- Pediatrics Department, University Clinical Hospital of Valladolid, Valladolid, Spain.
ORCID-Profile: Daniel Robles de la Osa, Javier Santos Fernández, Claudia Pérez Urra, Javier Barcenilla Laguna, Laura López Allúe, Antonio Germán Pérez Millán
Zitieren
APA
Robles de la Osa, D., Santos Fernández, J., Pérez Urra, C., Espinel Pinedo, P., Bulnes Labrador, C. B., Martín Ibáñez, C., González de Castro, E., Pérez Citores, L., Montero Moretón, Á. M., Santos Santamarta, F., Cimavilla Román, M., Moreira da Silva, B. A., Maestro Antolín, S., Barcenilla Laguna, J., Rancel Medina, F. J., Rizzo Rodríguez, M. A., López Allúe, L., & Pérez Millán, A. G. (2026). Efficacy of an artificial intelligence system for lesion detection and characterization ( CADe and CADx ) during colonoscopy following positive faecal immunochemical test in a colorectal cancer screening programme: A randomized clinical trial. Colorectal Disease, 28(3). https://doi.org/10.1111/codi.70426
Vancouver
Robles de la Osa D, Santos Fernández J, Pérez Urra C, Espinel Pinedo P, Bulnes Labrador CB, Martín Ibáñez C, et al. Efficacy of an artificial intelligence system for lesion detection and characterization ( CADe and CADx ) during colonoscopy following positive faecal immunochemical test in a colorectal cancer screening programme: A randomized clinical trial. Colorectal Disease. 2026;28(3). doi:10.1111/codi.70426
BibTeX
@article{RoblesdelaOsa2026,
author = {Robles de la Osa, Daniel and Santos Fernández, Javier and Pérez Urra, Claudia and Espinel Pinedo, Pablo and Bulnes Labrador, Carmen Beatriz and Martín Ibáñez, Clara and González de Castro, Elena and Pérez Citores, Laura and Montero Moretón, Ángela Martina and Santos Santamarta, Fernando and Cimavilla Román, Marta and Moreira da Silva, Bruno Antonio and Maestro Antolín, Sergio and Barcenilla Laguna, Javier and Rancel Medina, Francisco José and Rizzo Rodríguez, María Antonella and López Allúe, Laura and Pérez Millán, Antonio Germán},
title = {Efficacy of an artificial intelligence system for lesion detection and characterization ( CADe and CADx ) during colonoscopy following positive faecal immunochemical test in a colorectal cancer screening programme: A randomized clinical trial},
journal = {Colorectal Disease},
year = {2026},
volume = {28},
number = {3},
doi = {10.1111/codi.70426}
}
EUVIMED