{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,18]],"date-time":"2026-04-18T11:42:51Z","timestamp":1776512571565,"version":"3.51.2"},"reference-count":29,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2024,9,5]],"date-time":"2024-09-05T00:00:00Z","timestamp":1725494400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Artif. Intell."],"abstract":"<jats:sec><jats:title>Introduction<\/jats:title><jats:p>Deep learning (DL) has significantly advanced medical image classification. However, it often relies on transfer learning (TL) from models pretrained on large, generic non-medical image datasets like ImageNet. Conversely, medical images possess unique visual characteristics that such general models may not adequately capture.<\/jats:p><\/jats:sec><jats:sec><jats:title>Methods<\/jats:title><jats:p>This study examines the effectiveness of modality-specific pretext learning strengthened by image denoising and deblurring in enhancing the classification of pediatric chest X-ray (CXR) images into those exhibiting no findings, i.e., normal lungs, or with cardiopulmonary disease manifestations. Specifically, we use a <jats:italic>VGG-16-Sharp-U-Net<\/jats:italic> architecture and leverage its encoder in conjunction with a classification head to distinguish normal from abnormal pediatric CXR findings. We benchmark this performance against the traditional TL approach, <jats:italic>viz.<\/jats:italic>, the VGG-16 model pretrained only on ImageNet. Measures used for performance evaluation are balanced accuracy, sensitivity, specificity, F-score, Matthew\u2019s Correlation Coefficient (MCC), Kappa statistic, and Youden\u2019s index.<\/jats:p><\/jats:sec><jats:sec><jats:title>Results<\/jats:title><jats:p>Our findings reveal that models developed from CXR modality-specific pretext encoders substantially outperform the ImageNet-only pretrained model, <jats:italic>viz.<\/jats:italic>, Baseline, and achieve significantly higher sensitivity (<jats:italic>p<\/jats:italic>\u2009&amp;lt;\u20090.05) with marked improvements in balanced accuracy, F-score, MCC, Kappa statistic, and Youden\u2019s index. A novel attention-based fuzzy ensemble of the pretext-learned models further improves performance across these metrics (Balanced accuracy: 0.6376; Sensitivity: 0.4991; F-score: 0.5102; MCC: 0.2783; Kappa: 0.2782, and Youden\u2019s index:0.2751), compared to Baseline (Balanced accuracy: 0.5654; Sensitivity: 0.1983; F-score: 0.2977; MCC: 0.1998; Kappa: 0.1599, and Youden\u2019s index:0.1327).<\/jats:p><\/jats:sec><jats:sec><jats:title>Discussion<\/jats:title><jats:p>The superior results of CXR modality-specific pretext learning and their ensemble underscore its potential as a viable alternative to conventional ImageNet pretraining for medical image classification. Results from this study promote further exploration of medical modality-specific TL techniques in the development of DL models for various medical imaging applications.<\/jats:p><\/jats:sec>","DOI":"10.3389\/frai.2024.1419638","type":"journal-article","created":{"date-parts":[[2024,9,5]],"date-time":"2024-09-05T05:19:14Z","timestamp":1725513554000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":1,"title":["Noise-induced modality-specific pretext learning for pediatric chest X-ray image classification"],"prefix":"10.3389","volume":"7","author":[{"given":"Sivaramakrishnan","family":"Rajaraman","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Zhaohui","family":"Liang","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Zhiyun","family":"Xue","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Sameer","family":"Antani","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"1965","published-online":{"date-parts":[[2024,9,5]]},"reference":[{"key":"ref1","doi-asserted-by":"publisher","first-page":"1","DOI":"10.7717\/peerj-cs.607","article-title":"Chest X-ray pneumothorax segmentation using U-net with EfficientNet and ResNet architectures","volume":"7","author":"Abedalla","year":"2021","journal-title":"PeerJ Comput. Sci."},{"key":"ref2","doi-asserted-by":"publisher","first-page":"1120989","DOI":"10.3389\/fdata.2023.1120989","article-title":"AI-based radiodiagnosis using chest X-rays: a review","volume":"6","author":"Akhter","year":"2023","journal-title":"Front. Big Data"},{"key":"ref3","doi-asserted-by":"publisher","first-page":"815","DOI":"10.1038\/s41598-022-05052-x","article-title":"Objective evaluation of deep uncertainty predictions for COVID-19 detection","volume":"12","author":"Asgharnezhad","year":"2022","journal-title":"Sci. Rep."},{"key":"ref4","doi-asserted-by":"publisher","first-page":"1601","DOI":"10.1177\/0284185120973630","article-title":"Identifying cardiomegaly in chest X-rays: a cross-sectional study of evaluation and comparison between different transfer learning methods","volume":"62","author":"Bougias","year":"2021","journal-title":"Acta Radiol."},{"key":"ref5","doi-asserted-by":"publisher","first-page":"e551","DOI":"10.7717\/PEERJ-CS.551","article-title":"ECOVNet: a highly effective ensemble based deep learning model for detecting COVID-19","volume":"7","author":"Chowdhury","year":"2021","journal-title":"PeerJ Comput. Sci."},{"key":"ref6","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1148\/ryai.2021200126","article-title":"Magician\u2019s corner: 9. Performance metrics for machine learning models","volume":"3","author":"Erickson","year":"2021","journal-title":"Radiol. Artif. Intell."},{"key":"ref7","doi-asserted-by":"publisher","first-page":"1037","DOI":"10.1167\/9.8.1037","article-title":"ImageNet: constructing a large-scale image database","volume":"9","author":"Fei-Fei","year":"2010","journal-title":"J. Vis."},{"key":"ref8","doi-asserted-by":"publisher","first-page":"W126","DOI":"10.2214\/AJR.14.13116","article-title":"Radiographic techniques, contrast, and noise in X-ray imaging","volume":"204","author":"Huda","year":"2015","journal-title":"Am. J. Roentgenol."},{"key":"ref9","doi-asserted-by":"publisher","first-page":"202","DOI":"10.1016\/j.neucom.2022.01.055","article-title":"A fuzzy-enhanced deep learning approach for early detection of COVID-19 pneumonia from portable chest X-ray images","volume":"481","author":"Ieracitano","year":"2022","journal-title":"Neurocomputing"},{"key":"ref10","doi-asserted-by":"publisher","first-page":"1122","DOI":"10.1016\/j.cell.2018.02.010","article-title":"Identifying medical diagnoses and treatable diseases by image-based deep learning","volume":"172","author":"Kermany","year":"2018","journal-title":"Cell"},{"key":"ref11","first-page":"57","article-title":"MobileNetV2 based chest X-rays classification","author":"Kolonne","year":"2021"},{"key":"ref12","doi-asserted-by":"publisher","first-page":"161","DOI":"10.3991\/ijoe.v18i07.30807","article-title":"U-net based chest X-ray segmentation with ensemble classification for COVID-19 and pneumonia","volume":"18","author":"Kumarasinghe","year":"2022","journal-title":"Int. J. Online Biomed. Eng."},{"key":"ref13","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1016\/j.comcom.2021.06.005","article-title":"A cost-effective trilateration-based radio localization algorithm using machine learning and sequential least-square programming optimization","volume":"177","author":"Marques","year":"2021","journal-title":"Comput. Commun."},{"key":"ref14","doi-asserted-by":"publisher","first-page":"109319","DOI":"10.1016\/j.asoc.2022.109319","article-title":"Chest X-ray analysis empowered with deep learning: a systematic review","volume":"126","author":"Meedeniya","year":"2022","journal-title":"Appl. Soft. Comput."},{"key":"ref15","doi-asserted-by":"publisher","first-page":"e210315","DOI":"10.1148\/ryai.210315","article-title":"RadImageNet: an open radiologic deep learning research dataset for effective transfer learning","volume":"4","author":"Mei","year":"2022","journal-title":"Radiol. Artif. Intell."},{"key":"ref16","first-page":"90","article-title":"Automated radiology report generation using transformers","author":"Nimalsiri","year":"2023"},{"key":"ref17","doi-asserted-by":"publisher","first-page":"17532","DOI":"10.1038\/s41598-020-74539-2","article-title":"Automatic classification between COVID-19 pneumonia, non-COVID-19 pneumonia, and the healthy on chest X-ray image: combination of data augmentation methods","volume":"10","author":"Nishio","year":"2020","journal-title":"Sci. Rep."},{"key":"ref18","doi-asserted-by":"publisher","first-page":"6268","DOI":"10.1038\/s41598-019-42557-4","article-title":"Efficient deep network architectures for fast chest X-ray tuberculosis screening and visualization","volume":"9","author":"Pasa","year":"2019","journal-title":"Sci. Rep."},{"key":"ref19","doi-asserted-by":"publisher","first-page":"240","DOI":"10.1038\/s41597-023-02102-5","article-title":"PediCXR: an open, large-scale chest radiograph dataset for interpretation of common thoracic diseases in children","volume":"10","author":"Pham","year":"2023","journal-title":"Sci. Data"},{"key":"ref20","doi-asserted-by":"publisher","first-page":"e0242301","DOI":"10.1371\/journal.pone.0242301","article-title":"Analyzing inter-reader variability affecting deep ensemble learning for COVID-19 detection in chest radiographs","volume":"15","author":"Rajaraman","year":"2020","journal-title":"PLoS One"},{"key":"ref21","doi-asserted-by":"publisher","first-page":"120531","DOI":"10.1016\/j.eswa.2023.120531","article-title":"Can deep adult lung segmentation models generalize to the pediatric population?","volume":"229","author":"Rajaraman","year":"2023","journal-title":"Expert Syst. Appl."},{"key":"ref22","doi-asserted-by":"publisher","first-page":"109098","DOI":"10.1016\/j.compeleceng.2024.109098","article-title":"Multimodal imputation-based stacked ensemble for prediction and classification of air quality index in Indian cities","volume":"114","author":"Rao","year":"2024","journal-title":"Comput. Electr. Eng."},{"key":"ref23","doi-asserted-by":"publisher","first-page":"33","DOI":"10.1016\/j.image.2017.11.001","article-title":"A Haar wavelet-based perceptual similarity index for image quality assessment","volume":"61","author":"Reisenhofer","year":"2018","journal-title":"Signal Process. Image Commun."},{"key":"ref24","article-title":"Very deep convolutional networks for large-scale image recognition","author":"Simonyan","year":"2015"},{"key":"ref25","doi-asserted-by":"publisher","first-page":"81","DOI":"10.1093\/rpd\/nch517","article-title":"Evaluation of a novel method of noise reduction using computer-simulated mammograms","volume":"114","author":"Tischenko","year":"2005","journal-title":"Radiat. Prot. Dosim."},{"key":"ref26","first-page":"111","article-title":"Score-CAM: score-weighted visual explanations for convolutional neural networks","author":"Wang","year":"2020"},{"key":"ref27","doi-asserted-by":"publisher","first-page":"1068","DOI":"10.3390\/diagnostics13061068","article-title":"Cross dataset analysis of domain shift in CXR lung region detection","volume":"13","author":"Xue","year":"2023","journal-title":"Diagnostics"},{"key":"ref28","doi-asserted-by":"publisher","first-page":"21300","DOI":"10.1109\/ACCESS.2023.3249759","article-title":"Assessing inter-annotator agreement for medical image segmentation","volume":"11","author":"Yang","year":"2023","journal-title":"IEEE Access"},{"key":"ref29","doi-asserted-by":"publisher","first-page":"104699","DOI":"10.1016\/j.compbiomed.2021.104699","article-title":"Sharp U-net: depthwise convolutional network for biomedical image segmentation","volume":"136","author":"Zunair","year":"2021","journal-title":"Comput. Biol. Med."}],"container-title":["Frontiers in Artificial Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2024.1419638\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,9,5]],"date-time":"2024-09-05T05:19:17Z","timestamp":1725513557000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2024.1419638\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,9,5]]},"references-count":29,"alternative-id":["10.3389\/frai.2024.1419638"],"URL":"https:\/\/doi.org\/10.3389\/frai.2024.1419638","relation":{},"ISSN":["2624-8212"],"issn-type":[{"value":"2624-8212","type":"electronic"}],"subject":[],"published":{"date-parts":[[2024,9,5]]},"article-number":"1419638"}}