Part-Based Multiderivative Edge Cross-Sectional Profiles for Polyp Detection in Colonoscopy

Abstract

This paper presents a novel technique for automated detection of protruding polyps in colonoscopy images using edge cross-section profiles (ECSP). We propose a part-based multiderivative ECSP that computes derivative functions of an edge cross-section profile and segments each of these profiles into parts. Therefore, we can model or extract features suitable for each part. Our features obtained from the parts can effectively describe complex properties of protruding polyps including the shape of the parts, texture, and protrusion and smoothness of the polyp surface. We evaluated our method against two existing polyp image detection techniques on 42 different polyps, including those with little protrusion. Each polyp has a large variation of appearance in viewing angles, light conditions, and scales in different images. The evaluation showed that our technique outperformed the existing techniques in both accuracy and analysis time. Our method has a higher area under the free-response receiver operating characteristic curve. For instance, when both techniques have a true positive rate for polyp image detection of 81.4%, the average number of false regions per image of our technique is 0.32 compared to 1.8 of the best existing technique under study. Additionally, our technique can precisely mark edges of candidate polyp regions as visual feedback. These results altogether indicate that our technique is promising to provide visual feedback of polyp regions in clinical practice.

Polyp-Alert: Near real-time feedback during colonoscopy

“…Our previous work published in 2013 [34] compared our part-based multi-derivative edge cross-section profile (ECSP) features with Local Binary Pattern (LBP) and Opponent Color LBP (OCLBP) [6]. OCLBP was shown to perform best on over 15,000 images [6].…”

“…Our technique took 7.1 s to process one image implemented using MATLAB. From Table 1, our technique [34] is by far the most promising one to provide feedback in real time. Tajbakhsh et al [35] presented an edge like descriptor based on the similar principle as our work in Ref.…”

“…Tajbakhsh et al [35] presented an edge like descriptor based on the similar principle as our work in Ref. [34]. The technique detected polyp edges by sliding a square patch on the image, extracting features based on the pixel information surrounding the edge normal directions and classifying features using Random Forest classifier.…”

“…The technique in [35] implemented a sampler edge descriptor based on our edge normal and cross-section property in Ref. [34]. However, the technique does not directly compare with our technique.…”

“…The technique was tested on only 150 test images and the analysis time was not reported. In this paper, we focus on improving our technique in [34] in terms of analysis speed while maintaining high recall and low false alarms in order to provide near real-time visual feedback.…”