Real-time deep learning-based colorectal polyp localization on clinical video footage achievable with a wide array of hardware configurations

Abstract

Background and study aims Several computer-assisted polyp detection systems have been proposed, but they have various limitations, from utilizing outdated neural network architectures to a requirement for multi-graphics processing unit (GPU) processing, to validating on small or non-robust datasets. To address these problems, we developed a system based on a state-of-the-art convolutional neural network architecture able to detect polyps in real time on a single GPU and tested on both public datasets and full clinical examination recordings. Methods The study comprised 165 colonoscopy procedure recordings and 2678 still photos gathered retrospectively. The system was trained on 81,962 polyp frames in total and then tested on footage from 42 colonoscopies and CVC-ClinicDB, CVC-ColonDB, Hyper-Kvasir, and ETIS-Larib public datasets. Clinical videos were evaluated for polyp detection and false-positive rates whereas the public datasets were assessed for F1 score. The system was tested for runtime performance on a wide array of hardware. Results The performance on public datasets varied from an F1 score of 0.727 to 0.942. On full examination videos, it detected 94 % of the polyps found by the endoscopist with a 3 % false-positive rate and identified additional polyps that were missed during initial video assessment. The system’s runtime fits within the real-time constraints on all but one of the hardware configurations. Conclusions We have created a polyp detection system with a post-processing pipeline that works in real time on a wide array of hardware. The system does not require extensive computational power, which could help broaden the adaptation of new commercially available systems.

Polyp characterization using deep learning and a publicly accessible polyp video database

“…A plausible explanation is that annotating high-and moderate-quality video frames compensated for the modest number of polyps. 20 CADx models are often developed using retrospective still images from endoscopy reports of expert endoscopists, but this inherently limits the data to one or a few images per polyp. These models may also generalize less well to nonexpert endoscopists, who may be less likely to reproduce the highquality images of experts.…”

“…Our CNN achieved favorable results without using a confidence threshold despite a smaller sample size of polyps for training than other models. A plausible explanation is that annotating high‐ and moderate‐quality video frames compensated for the modest number of polyps 20 . CADx models are often developed using retrospective still images from endoscopy reports of expert endoscopists, but this inherently limits the data to one or a few images per polyp.…”

Computer-Aided Detection False Positives in Colonoscopy

“…Lee et al proposed to reduce FPs by using the median filter (which reduced the FPR from 12.5% to 6.3%), a nonlinear spatial filter that is particularly effective for eliminating salt-and-pepper noise [ 21 ]. To filter out most short flashes, Podlasek et al suggested setting a threshold of persistent time for FPs to show up; however, this method might introduce a minor detection lag, depending on the desired sensitivity [ 22 ]. These methods are beyond the expertise of the clinical endoscopists.…”

Deep Learning-based Polyp Detection in Wireless Capsule Endoscopy Images

“…The F1 score of CVC-ClinicDB ranged from 0.727 to 0.942. With a 3% FP rate, full examination films sample detected 94% of polyps [9].…”

“…Different studies were concerned with developing automated systems for polyp detection. Podlasek et al [9] designed a system for polyp detection with a real-time post-processing pipeline that runs on a variety of devices. The F1 score of CVC-ClinicDB ranged from 0.727 to 0.942.…”