[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-121943-en":3,"doc-seo-121943-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},121943,137441390410,"Hazel","https://ap-avatar.wpscdn.com/avatar/2000252f4ab5702993?_k=1776741390130283984",8,"Research & Report","Addressing Chest Radiograph Projection Bias in Deep Classification Models - Paper Summary","Deep learning models used for chest radiograph disease classification can suffer from projection-related bias when posteroanterior versus anteroposterior imaging introduces learnable, spurious correlations with specific conditions. This work evaluates how projection affects model performance and introduces a label-conditional gradient reversal framework that encourages features to support disease recognition while discouraging projection discrimination. Results indicate projection-aware behavior, and improved robustness to projection-induced bias.","Addressing Chest Radiograph Projection Bias in Deep  \nClassification Models  \nSofia C. Pereira 1 ,2  \n1 Faculty of Engineering, University of Porto  \n2 Institute for Systems and Computer Engineering, Technology and Science, INESC-TEC  \nJoana Rocha 1 ,2 Alex Gaudio 1 ,2 ,3  \n3 Carnegie Mellon University Asim Smailagic3 Aur´elio Campilho 1 ,2 Ana Maria Mendon¸ca 1 ,2  \n[sofia.c.pereira@inesctec.pt](sofia.c.pereira@inesctec.pt)  \n[joana.m.rocha@inesctec.pt](joana.m.rocha@inesctec.pt)[agaudio@andrew.cmu.edu](agaudio@andrew.cmu.edu)  \n[asim@andrew.cmu.edu](asim@andrew.cmu.edu)[ ](asim@andrew.cmu.edu)[campilho@fe.up.pt](campilho@fe.up.pt)[ ](campilho@fe.up.pt)[amendon@fe.up.pt](amendon@fe.up.pt)  \nAbstract  \nDeep learning-based models are widely used for disease classification in chest radiographs.  \nThis exam can be performed in one of two projections (posteroanterior or anteroposterior), depending on the direction that the X-ray beam travels through the body. Since projection visibly affects the way anatomical structures appear in the scans, it may introduce bias inclassifiers, especially when spurious correlations between a given disease and a projection occur. This paper examines the influence of chest radiograph projection on the performance of deep learning-based classification models and proposes an approach to mitigate projection-induced bias. Results show that a DenseNet-121 model is better at classifying images from the most representative projection in the data set, suggesting that projection is taken into account by the classifier. Moreover, this model can classify chest X-ray projection better than any of the fourteen radiological findings considered, without being explicitly trained for that task, putting it at high risk for projection bias. We propose a label-conditional gradient reversal framework to make the model insensitive to projection, by forcing the extracted features to be simultaneously good for disease classification and bad for projection classification, resulting in a framework with reduced projection-induced bias.  \nKeywords: chest X-ray, deep learning, disparity, anteroposterior, posteroanterior  \n1. Introduction  \nArtificial Intelligence (AI) revolutionized many fields, including medicine. It is crucial to ensure that AI-based methods are not only accurate but also unbiased and ethically compliant. Biased diagnoses can have serious consequences, and the potential sources of bias are countless, ranging from biological attributes (age, gender) to socioeconomic factors (race, income) and clinical factors (exam conditions, medical institution, presence of medical devices, among others) . When AI models are trained on biased data, they become prone to shortcut learning: instead of learning meaningful features for the issue at hand, models base their predictions on biases in the data, which are typically easier to learn than meaningful features. Not only this poses serious ethical concerns but, consequentially, these models will fail in data where the same types of biases are not present.  \n© 2023 CC-BY 4.0, S.C. Pereira, J. Rocha, A. Gaudio, A. Smailagic, A. Campilho & A.M. Mendon¸ca.  \nPereira Rocha Gaudio Smailagic Campilho Mendonc¸a  \n(a) Posteroanterior scan  \n(b) Anteroposterior scan  \nFigure 1: Examples of chest radiographs taken in both projections.  \nIn this paper, we focus on the bias that projection can inject in deep learning models for frontal chest X-ray (CXR) classification. These scans can be performed in one of two projections. In the posteroanterior (PA) projection (Figure 1a), the X-ray beam transversesthe patient from posterior to anterior, which is the gold standard projection for frontal CXRs. However, it is not always possible to perform PA CXRs, especially when the patient is not able to stand. When that is the case, an anteroposterior (AP) projection (Figure 1b) is used, which does not require the patient to stand up, enabling the use of portable X-ray units. There are some key diffe","cbCaibrg7eFkJnKC","https://ap.wps.com/l/cbCaibrg7eFkJnKC","pdf",441513,1,12,"English","en",105,"# Introduction\n## Related Work\n# Methodology\n## Label-Conditional Gradient Reversal Framework\n# Experimental Results\n## Projection Sensitivity and Findings","[{\"question\":\"Why can chest radiograph projection bias deep learning classifiers?\",\"answer\":\"Because the posteroanterior and anteroposterior views change how anatomy appears, models may learn spurious correlations between projection type and disease labels.\"},{\"question\":\"What approach does the paper propose to mitigate projection-induced bias?\",\"answer\":\"It proposes a label-conditional gradient reversal framework that makes extracted features useful for disease classification but uninformative for projection classification.\"},{\"question\":\"What do the results indicate about DenseNet-121 and projection?\",\"answer\":\"A DenseNet-121 model classifies images from the most representative projection better, suggesting the classifier accounts for projection even when not explicitly trained for it.\"}]","Addressing Chest Radiograph Projection Bias in Deep Classification Models - Paper Summary | PDF",1785807871,30,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"addressing-chest-radiograph-projection-bias-in-deep-classification-models-paper-summary","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/addressing-chest-radiograph-projection-bias-in-deep-classification-models-paper-summary/121943/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-04",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why can chest radiograph projection bias deep learning classifiers?","Question",{"text":75,"@type":76},"Because the posteroanterior and anteroposterior views change how anatomy appears, models may learn spurious correlations between projection type and disease labels.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What approach does the paper propose to mitigate projection-induced bias?",{"text":80,"@type":76},"It proposes a label-conditional gradient reversal framework that makes extracted features useful for disease classification but uninformative for projection classification.",{"name":82,"@type":73,"acceptedAnswer":83},"What do the results indicate about DenseNet-121 and projection?",{"text":84,"@type":76},"A DenseNet-121 model classifies images from the most representative projection better, suggesting the classifier accounts for projection even when not explicitly trained for it.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":29,"slug":121},"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]