Computational learning of features for automated colonic polyp classification

Kangkana Bora,Zhongming Zhao,M K Bhuyan,Kunio Kasugai,Saurav Mallik

doi:10.1038/s41598-021-83788-8

Abstract

Shape, texture, and color are critical features for assessing the degree of dysplasia in colonic polyps. A comprehensive analysis of these features is presented in this paper. Shape features are extracted using generic Fourier descriptor. The nonsubsampled contourlet transform is used as texture and color feature descriptor, with different combinations of filters. Analysis of variance (ANOVA) is applied to measure statistical significance of the contribution of different descriptors between two colonic polyps: non-neoplastic and neoplastic. Final descriptors selected after ANOVA are optimized using the fuzzy entropy-based feature ranking algorithm. Finally, classification is performed using Least Square Support Vector Machine and Multi-layer Perceptron with five-fold cross-validation to avoid overfitting. Evaluation of our analytical approach using two datasets suggested that the feature descriptors could efficiently designate a colonic polyp, which subsequently can help the early detection of colorectal carcinoma. Based on the comparison with four deep learning models, we demonstrate that the proposed approach out-performs the existing feature-based methods of colonic polyp identification.

Highlights

Colonoscopy image processing has been an area of active research
The results are highly satisfactory with Generic Fourier Descriptor (GFD)+NonsSubsampled Contourlet Transform (NSCT) feature, which is giving highest accuracy of 95.24% and 94.68%
We propose a novel approach to quantify shape, texture, and color features for detecting the stages of dysplasia in polyps

Summary

Introduction

Colonoscopy image processing has been an area of active research. Studies cover polyp localization and segmentation[6], as well as feature analysis for various applications. 3-D reconstruction and polyp segmentation in endoscopic videos are reported in the literature[7]. There are many studies of the classification of polyps using traditional and deep feature learning based methods. Fu et al[8] have worked on a large dataset comprising 365 generated images They extracted texture features from the first component of the principal component transform, representing both spatial and spectral domains. Sebastian et al[22] tried to classify colon polyps based on the Kudo’s classification schema using VGG16 filter as feature extraction method and achieved an accuracy of 83%. Shape feature is extracted using Generic Fourier Descriptor (GFD), while texture and color components are obtained by NonsSubsampled Contourlet Transform (NSCT) with different filters. Our extensive experiments show that the proposed method outperforms the existing feature-based (conventional) approaches for colonic polyp detection. The proposed work is compared with four classical deep learning models

Objectives

Methods

Results

Conclusion