Diagnostic Performance of a Convolutional Neural Network for Diminutive Colorectal Polyp Recognition (POLAR)

Diagnostic Performance of a Convolutional Neural Network for Diminutive Colorectal Polyp Recognition. A Multicentre, Prospective Observational Study

Rationale: Diminutive colorectal polyps (1-5mm in size) have a high prevalence and very low risk of harbouring cancer. Current practice is to send all these polyps for histopathological assessment by the pathologist. If an endoscopist would be able to correctly predict the histology of these diminutive polyps during colonoscopy, histopathological examination could be omitted and practise could become more time- and cost-effective. Studies have shown that prediction of histology by the endoscopist remains dependent on training and experience and varies greatly between endoscopists, even after systematic training. Computer aided diagnosis (CAD) based on convolutional neural networks (CNN) may facilitate endoscopists in diminutive polyp differentiation. Up to date, studies comparing the diagnostic performance of CAD-CNN to a group of endoscopists performing optical diagnosis during real-time colonoscopy are lacking.

Objective: To develop a CAD-CNN system that is able to differentiate diminutive polyps during colonoscopy with high accuracy and to compare the performance of this system to a group of endoscopist performing optical diagnosis, with the histopathology as the gold standard.

Study design: Multicentre, prospective, observational trial. Study population: Consecutive patients who undergo screening colonoscopy (phase 2)

Main study parameters/endpoints: The accuracy of optical diagnosis of diminutive colorectal polyps (1-5mm) by CAD-CNN system compared with the accuracy of the endoscopists. Histopathology is used as the gold standard.

Study Overview

• Status: Completed
• Conditions: Artificial Intelligence; Colorectal Polyp
• Intervention / Treatment: Device: CAD-CNN system

• Study Type: Observational
• Enrollment (Actual): 292

Contacts and Locations

Study Locations

• Netherlands
  • Noord-Holland
    • Amsterdam, Noord-Holland, Netherlands, 1105AZ
      • Academic Medical Centre

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: 18 years and older (Adult, Older Adult)
• Accepts Healthy Volunteers: N/A
• Genders Eligible for Study: All

• Sampling Method: Probability Sample

Study Population

Phase 1APatients that underwent colonoscopy between 2011-2018 in the Bergman Clinics Amsterdam, in the context of the Dutch bowel cancer screening or surveillance program or because of symptoms.

Phase 1B Patients older than 18 years that underwent colonoscopy in one of the participating centres.

Phase 2 All patients older than 18 years old undergoing screenings colonoscopy in one of the participating centres.

Description

Phase 1A -

- Patients with one polyp subtype (based on histology)

Phase 1B Patients older than 18 years that underwent colonoscopy in one of the participating centres.

Phase 2:- Validation CAD-CNN system

Inclusion Criteria:

All patients older than 18 years old undergoing screenings colonoscopy in one of the participating centres.

Exclusion Criteria:

• Diagnosis of inflammatory bowel disease, Lynch syndrome or (serrated) polyposis syndrome.
• Boston Bowel Preparation Scale (BBPS) <2 in one of the colon segments
• Patients who are unwilling or unable to give informed consent

Study Plan

How is the study designed?

Number of groups / cohorts

1

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
Patients; Patients older than 18 years undergoing colonoscopy in one the participating centres.	Device: CAD-CNN system; The CAD-CNN system will be trained in predicting the histology of diminutive polyps. Before training, the dataset will be split up into a training set and a test set. To ensure a completely independent test and training set there will be no overlap between patients (i.e. if polyps from a patient A is present in the training set it cannot be in the test set as well).

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
The accuracy of the CAD-CNN system for predicting histology of diminutive colorectal polyps (1-5mm) compared with the accuracy of the prediction of the endoscopist. Both the CAD-CNN system and the endoscopist will use NBI for their predictions.	Accuracy is defined as the percentage of correctly predicted optical diagnoses of the CAD-CNN system and / or endoscopist compared to the gold standard pathology. For the calculation of the accuracy, adenomas and SSLs will be dichotomized as neoplastic polyps, while HPs are considered non-neoplastic	2 year

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
The mean duration in seconds of the CAD-CNN system to make a per polyp diagnosis.	The mean duration in seconds of the CAD-CNN system to make a per polyp diagnosis.	2 year
The mean number of attempts of the CAD-CNN to make a diagnosis per polyp	The mean number of attempts of the CAD-CNN to make a diagnosis per polyp	2 year
The ratio of unsuccessful diagnosis from all diagnosis of the CAD-CNN system. An unsuccessful diagnosis/failure of the CAD-CNN system is defined as more than 3 unsuccessful attempts	The ratio of unsuccessful diagnosis from all diagnosis of the CAD-CNN system. An unsuccessful diagnosis/failure of the CAD-CNN system is defined as more than 3 unsuccessful attempts	2 year
The number of diminutive polyps per colonoscopy that is resected and discarded without histopathological analysis with optical diagnosis strategy (the CAD-CNN system or endoscopist)	The number of diminutive polyps per colonoscopy that is resected and discarded without histopathological analysis with optical diagnosis strategy (the CAD-CNN system or endoscopist)	2 year
The percentage of colonoscopies in which diminutive polyps are characterized based on optical diagnosis, removed and discarded without histopathological evaluation (i.e. proportion of polyps assessed with high confidence)	The percentage of colonoscopies in which diminutive polyps are characterized based on optical diagnosis, removed and discarded without histopathological evaluation (i.e. proportion of polyps assessed with high confidence)	2 year
The percentage of colonoscopies in which the surveillance interval is based on the optical diagnosis of the CAD-CNN system and the patient can be directly informed of the surveillance interval after colonoscopy	The percentage of colonoscopies in which the surveillance interval is based on the optical diagnosis of the CAD-CNN system and the patient can be directly informed of the surveillance interval after colonoscopy	2 year
The percentage of colonoscopies in which diminutive hyperplastic polyps in the rectosigmoid are left in situ.	The percentage of colonoscopies in which diminutive hyperplastic polyps in the rectosigmoid are left in situ.	2 year
The diagnostic sensitivity for optical diagnosis of the CAD-CNN system and the endoscopists	The diagnostic sensitivity for optical diagnosis of the CAD-CNN system and the endoscopists	2 year
The diagnostic sensitiviy for optical diagnosis of the CAD-CNN system and the endoscopists	The diagnostic sensitiviy for optical diagnosis of the CAD-CNN system and the endoscopists	2 year
The accuracy rates on a per polyp basis	Accuracy on a polyp basis is defined as the percentage of correctly predicted optical diagnoses of the CAD-CNN system and / or endoscopist compared to the gold standard pathology. For the calculation of the accuracy on a polyp basis, adenomas, SSLs and HPs are considered different subtypes.	2 year
Agreement between recommended surveillance intervals, based on optical diagnosis of diminutive polyps with high confidence, compared to surveillance recommendations based on histology of all polyps	Agreement between recommended surveillance intervals, based on optical diagnosis of diminutive polyps with high confidence, compared to surveillance recommendations based on histology of all polyps	2 year
The diagnostic specificity for optical diagnosis of the CAD-CNN system and the endoscopists	The diagnostic specificity for optical diagnosis of the CAD-CNN system and the endoscopists	2 year
The diagnostic PPV for optical diagnosis of the CAD-CNN system and the endoscopists	The diagnostic PPV for optical diagnosis of the CAD-CNN system and the endoscopists	2 year
The diagnostic NPV for optical diagnosis of the CAD-CNN system and the endoscopists	The diagnostic NPV for optical diagnosis of the CAD-CNN system and the endoscopists	2 year

Collaborators and Investigators

• Sponsor: Academisch Medisch Centrum - Universiteit van Amsterdam (AMC-UvA)
• Collaborators: Medical Centre Leeuwarden; Bergman Clinics
• Investigators: Principal Investigator: Evelien NA Dekker, Msc, Amsterdam Umc, Location Vumc

Study record dates

Study Major Dates

• Study Start (Actual): October 16, 2018
• Primary Completion (Actual): October 16, 2021
• Study Completion (Actual): October 16, 2021

Study Registration Dates

• First Submitted: January 21, 2019
• First Submitted That Met QC Criteria: January 29, 2019
• First Posted (Actual): January 30, 2019

Study Record Updates

• Last Update Posted (Actual): December 29, 2021
• Last Update Submitted That Met QC Criteria: December 9, 2021
• Last Verified: December 1, 2021

More Information

Terms related to this study

• Keywords: Artificial Intelligence; Computer aided diagnosis; Diminutive colorectal polyps; Optical diagnosis
• Additional Relevant MeSH Terms: Pathological Conditions, Anatomical; Polyps
• Other Study ID Numbers: W18_422

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: UNDECIDED
• IPD Plan Description: There is not yet a plan to share data. However, patients will be asked informed consent with the possibility to share the data.

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No