Detection and visualization of abnormality in chest radiographs using modality-specific convolutional neural network ensembles

Convolutional neural networks (CNNs) trained on natural images are extremely successful in image classification and localization due to superior automated feature extraction capability. In extending their use to biomedical recognition tasks, it is important to note that visual features of medical im...

Full description

Bibliographic Details
Main Authors: Sivaramakrishnan Rajaraman, Incheol Kim, Sameer K. Antani
Format: Article
Language:English
Published: PeerJ Inc. 2020-03-01
Series:PeerJ
Subjects:
Online Access:https://peerj.com/articles/8693.pdf
_version_ 1827610653506928640
author Sivaramakrishnan Rajaraman
Incheol Kim
Sameer K. Antani
author_facet Sivaramakrishnan Rajaraman
Incheol Kim
Sameer K. Antani
author_sort Sivaramakrishnan Rajaraman
collection DOAJ
description Convolutional neural networks (CNNs) trained on natural images are extremely successful in image classification and localization due to superior automated feature extraction capability. In extending their use to biomedical recognition tasks, it is important to note that visual features of medical images tend to be uniquely different than natural images. There are advantages offered through training these networks on large scale medical common modality image collections pertaining to the recognition task. Further, improved generalization in transferring knowledge across similar tasks is possible when the models are trained to learn modality-specific features and then suitably repurposed for the target task. In this study, we propose modality-specific ensemble learning toward improving abnormality detection in chest X-rays (CXRs). CNN models are trained on a large-scale CXR collection to learn modality-specific features and then repurposed for detecting and localizing abnormalities. Model predictions are combined using different ensemble strategies toward reducing prediction variance and sensitivity to the training data while improving overall performance and generalization. Class-selective relevance mapping (CRM) is used to visualize the learned behavior of the individual models and their ensembles. It localizes discriminative regions of interest (ROIs) showing abnormal regions and offers an improved explanation of model predictions. It was observed that the model ensembles demonstrate superior localization performance in terms of Intersection of Union (IoU) and mean Average Precision (mAP) metrics than any individual constituent model.
first_indexed 2024-03-09T07:53:12Z
format Article
id doaj.art-fe3b1e6eca7544c79c9e28229a19261d
institution Directory Open Access Journal
issn 2167-8359
language English
last_indexed 2024-03-09T07:53:12Z
publishDate 2020-03-01
publisher PeerJ Inc.
record_format Article
series PeerJ
spelling doaj.art-fe3b1e6eca7544c79c9e28229a19261d2023-12-03T01:26:39ZengPeerJ Inc.PeerJ2167-83592020-03-018e869310.7717/peerj.8693Detection and visualization of abnormality in chest radiographs using modality-specific convolutional neural network ensemblesSivaramakrishnan Rajaraman0Incheol Kim1Sameer K. Antani2Lister Hill National Center for Biomedical Communications, National Library of Medicine, National Institutes of Health, Bethesda, MD, United States of AmericaLister Hill National Center for Biomedical Communications, National Library of Medicine, National Institutes of Health, Bethesda, MD, United States of AmericaLister Hill National Center for Biomedical Communications, National Library of Medicine, National Institutes of Health, Bethesda, MD, United States of AmericaConvolutional neural networks (CNNs) trained on natural images are extremely successful in image classification and localization due to superior automated feature extraction capability. In extending their use to biomedical recognition tasks, it is important to note that visual features of medical images tend to be uniquely different than natural images. There are advantages offered through training these networks on large scale medical common modality image collections pertaining to the recognition task. Further, improved generalization in transferring knowledge across similar tasks is possible when the models are trained to learn modality-specific features and then suitably repurposed for the target task. In this study, we propose modality-specific ensemble learning toward improving abnormality detection in chest X-rays (CXRs). CNN models are trained on a large-scale CXR collection to learn modality-specific features and then repurposed for detecting and localizing abnormalities. Model predictions are combined using different ensemble strategies toward reducing prediction variance and sensitivity to the training data while improving overall performance and generalization. Class-selective relevance mapping (CRM) is used to visualize the learned behavior of the individual models and their ensembles. It localizes discriminative regions of interest (ROIs) showing abnormal regions and offers an improved explanation of model predictions. It was observed that the model ensembles demonstrate superior localization performance in terms of Intersection of Union (IoU) and mean Average Precision (mAP) metrics than any individual constituent model.https://peerj.com/articles/8693.pdfDeep learningConvolutional neural networksModality-specific knowledgeChest X-raysEnsemble learningClass-selective relevance mapping
spellingShingle Sivaramakrishnan Rajaraman
Incheol Kim
Sameer K. Antani
Detection and visualization of abnormality in chest radiographs using modality-specific convolutional neural network ensembles
PeerJ
Deep learning
Convolutional neural networks
Modality-specific knowledge
Chest X-rays
Ensemble learning
Class-selective relevance mapping
title Detection and visualization of abnormality in chest radiographs using modality-specific convolutional neural network ensembles
title_full Detection and visualization of abnormality in chest radiographs using modality-specific convolutional neural network ensembles
title_fullStr Detection and visualization of abnormality in chest radiographs using modality-specific convolutional neural network ensembles
title_full_unstemmed Detection and visualization of abnormality in chest radiographs using modality-specific convolutional neural network ensembles
title_short Detection and visualization of abnormality in chest radiographs using modality-specific convolutional neural network ensembles
title_sort detection and visualization of abnormality in chest radiographs using modality specific convolutional neural network ensembles
topic Deep learning
Convolutional neural networks
Modality-specific knowledge
Chest X-rays
Ensemble learning
Class-selective relevance mapping
url https://peerj.com/articles/8693.pdf
work_keys_str_mv AT sivaramakrishnanrajaraman detectionandvisualizationofabnormalityinchestradiographsusingmodalityspecificconvolutionalneuralnetworkensembles
AT incheolkim detectionandvisualizationofabnormalityinchestradiographsusingmodalityspecificconvolutionalneuralnetworkensembles
AT sameerkantani detectionandvisualizationofabnormalityinchestradiographsusingmodalityspecificconvolutionalneuralnetworkensembles