Bionnica: A Deep Neural Network Architecture for Colorectal Polyps' Premalignancy Risk Evaluation

Diogen Babuc,Todor Ivascu,Melania Ardelean,Darian Onchis

DOI: https://doi.org/10.1101/2024.06.19.24309153

2024-06-20

Abstract:The third most prevalent cancer nowadays is colorectal cancer. Colonoscopy is an important procedure in the stage of detection of polyps' malignancy because it helps in early identification and establishes effective therapy. This paper explores specific deep-learning architectures for the binary classification of colorectal polyps and considers the evaluation of their premalignancy risk. The main scope is to create a custom-based deep learning architecture that classifies adenomatous, hyperplastic, and serrated polyps' samples into benign and premalignant based on images from the colonoscopic dataset. Each image's output is modified through masked autoencoders which enhance the classification performance of the proposed model, called Bionnica. From the four evaluated state-of-the-art deep learning models (ZF NET, VGG-16, AlexNet, and ResNet-50), our experiments showed that ResNet-50 and ZF NET are most accurate (above 84%), with ResNet-50 excelling at indicating patients with premalignant colorectal polyps (above 92%). ZF NET is the fastest at handling 700 images. Our proposed deep learning model, Bionnica, is more performant than ZF NET and provides an efficient classification of colorectal polyps given its simple structure. The advantage of our model comes from the custom enhancement interpretability with a rule-based layer that guides the learning process and supports medical personnel in their decisions.

Health Informatics

What problem does this paper attempt to address?

The paper aims to address the issue of evaluating the malignant risk of colorectal polyps through deep learning methods. In the study, the authors designed a deep neural network architecture called Bionnica for binary classification (benign vs. malignant precursor) of colorectal polyps. The main objective of the paper is to create a rule-based customized deep learning model that can differentiate adenomatous, hyperplastic, and serrated polyps based on colonoscopy images and determine their malignant potential. By evaluating several advanced deep learning models such as ZF NET, VGG-16, AlexNet, and ResNet-50, it was found that ResNet-50 performs best in identifying high-risk colorectal polyps, while ZF NET is the fastest in terms of processing speed. Although the Bionnica model has a simple structure, it outperforms ZF NET in terms of performance, providing effective classification of colorectal polyps and enhancing the interpretability of the model by adding a rule layer to support medical decision-making.

Bionnica: A Deep Neural Network Architecture for Colorectal Polyps' Premalignancy Risk Evaluation

An Ensemble Framework of Deep Neural Networks for Colorectal Polyp Classification

A Deep Ensemble Learning Method for Colorectal Polyp Classification with Optimized Network Parameters

Comparative analysis of machine learning frameworks for automatic polyp characterization

An Ensemble-Based Deep Convolutional Neural Network for Computer-Aided Polyps Identification From Colonoscopy

Deep Learning in Image Classification using Residual Network (ResNet) Variants for Detection of Colorectal Cancer

Towards a Computed-Aided Diagnosis System in Colonoscopy: Automatic Polyp Segmentation Using Convolution Neural Networks

BetterNet: An Efficient CNN Architecture with Residual Learning and Attention for Precision Polyp Segmentation

Automatic Polyp Segmentation Using Convolutional Neural Networks

An Interpretable Deep Neural Network for Colorectal Polyp Diagnosis under Colonoscopy

A Deep Convolutional Neural Network for the Detection of Polyps in Colonoscopy Images

Deep neural networks for automated classification of colorectal polyps on histopathology slides: A multi-institutional evaluation

Y-Net: A deep Convolutional Neural Network for Polyp Detection

NanoNet: Real-Time Polyp Segmentation in Video Capsule Endoscopy and Colonoscopy

Deep Bayesian segmentation for colon polyps: Well-calibrated predictions in medical imaging

Deep Learning for Improved Polyp Detection from Synthetic Narrow-Band Imaging

Colon cancer classification and detection by novel CMNV2 model and methods of deep learning

Automatic Polyp Detection in Colonoscopy Images: Convolutional Neural Network, Dataset and Transfer Learning

Deep learning to find colorectal polyps in colonoscopy: A systematic literature review

AI based colorectal disease detection using real-time screening colonoscopy

Automatic Polyp Segmentation with Multiple Kernel Dilated Convolution Network