Endoscopy 2018; 50(04): S29-S30
DOI: 10.1055/s-0038-1637113
ESGE Days 2018 oral presentations
20.04.2018 – Colon: Improving characterization
Georg Thieme Verlag KG Stuttgart · New York

OPTICAL CLASSIFICATION OF NEOPLASTIC COLORECTAL POLYPS – A COMPUTER ASSISTED APPROACH (THE COACH STUDY)

P Klare
1   Klinikum rechts der Isar der Technischen Universität München, Klinik für Innere Medizin II, Munich, Germany
,
J Renner
1   Klinikum rechts der Isar der Technischen Universität München, Klinik für Innere Medizin II, Munich, Germany
,
B Haller
2   Klinikum rechts der Isar der Technischen Universität München, Institut für Medizinische Statistik und Epidemiologie, Munich, Germany
,
Y Saint-Hill-Febles
3   Technical University Munich, Chair for Computer Aided Medical Procedures & Augmented Reality, Munich, Germany
,
D Mateus
3   Technical University Munich, Chair for Computer Aided Medical Procedures & Augmented Reality, Munich, Germany
,
T Ponchon
4   Edouard Herriot Hospital, Department of Endoscopy and Gastroenterology, Lyon, France
,
C Schlag
1   Klinikum rechts der Isar der Technischen Universität München, Klinik für Innere Medizin II, Munich, Germany
,
A Poszler
1   Klinikum rechts der Isar der Technischen Universität München, Klinik für Innere Medizin II, Munich, Germany
,
RM Schmid
1   Klinikum rechts der Isar der Technischen Universität München, Klinik für Innere Medizin II, Munich, Germany
,
S von Delius
5   RoMed Klinikum Rosenheim, Medizinische Klinik II, Rosenheim, Germany
,
F Navarro-Avila
3   Technical University Munich, Chair for Computer Aided Medical Procedures & Augmented Reality, Munich, Germany
› Author Affiliations
Further Information

Publication History

Publication Date:
27 March 2018 (online)

Also available at
 

Aims:

Clinical data suggest that quality of optical diagnoses of colorectal polyps differ markedly among endoscopists. Therefore new technologies are wanted in order to assist endoscopists in their decisions on optical polyp characterization. The aim of this study was to develop a computer program being able to differentiate neoplastic from non-neoplastic polyps using unmagnified endoscopic pictures.

Methods:

During colonoscopy procedures polyp photografies were performed using the standard unmagnified high definition white light (HDWL) mode as well as the Narrow Band Image (NBI) mode. All detected polyps were resected and sent to pathology. Histopathological diagnoses served as the ground truth. Machine learning was used in order to generate a computer assisted optical biopsy (CAOB) approach. In a test phase 186 pictures from 100 polyps were presented to CAOB in order to obtain optical diagnoses. The same pictures were presented to two experts in optical polyp characterization. The primary endpoint of the study was accuracy of CAOB diagnoses in the test phase.

Results:

A total of 275 polyps were found in 250 patients. Altogether 788 polyp pictures were available for machine learning and CAOB testing. Accuracy of the CAOB approach was 78.0%. Sensitivity and Negative (NPV) were 92.3% and 88.2%. Accuracy obtained by two expert endoscopists was 84.0% and 77.0%. Regarding accuracy of optical diagnoses CAOB predictions did not differ significantly compared to experts (p = 0.307 and p = 1.000 respectively).

Conclusions:

Computer assisted optical biopsy showed a good accuracy on the basis of unmagnified endoscopic pictures. Performance of CAOB predictions did not differ significantly from experts decisions. The concept of computer assistance for colorectal polyp characterization needs to involve towards a real time application prior of being used in a broader setup.