Refining Tuberculosis Detection in CXR Imaging: Addressing Bias in Deep Neural Networks via Interpretability

Automatic classification of active tuberculosis from chest X-ray images has the potential to save lives, especially in low- and mid-income countries where skilled human experts can be scarce. Given the lack of available labeled data to train such systems and the unbalanced nature of publicly availab...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Güler, Özgür Acar, Günther, Manuel, Anjos, André
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Güler, Özgür Acar
Günther, Manuel
Anjos, André
description Automatic classification of active tuberculosis from chest X-ray images has the potential to save lives, especially in low- and mid-income countries where skilled human experts can be scarce. Given the lack of available labeled data to train such systems and the unbalanced nature of publicly available datasets, we argue that the reliability of deep learning models is limited, even if they can be shown to obtain perfect classification accuracy on the test data. One way of evaluating the reliability of such systems is to ensure that models use the same regions of input images for predictions as medical experts would. In this paper, we show that pre-training a deep neural network on a large-scale proxy task, as well as using mixed objective optimization network (MOON), a technique to balance different classes during pre-training and fine-tuning, can improve the alignment of decision foundations between models and experts, as compared to a model directly trained on the target dataset. At the same time, these approaches keep perfect classification accuracy according to the area under the receiver operating characteristic curve (AUROC) on the test set, and improve generalization on an independent, unseen dataset. For the purpose of reproducibility, our source code is made available online.
doi_str_mv 10.48550/arxiv.2407.14064
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2407_14064</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2407_14064</sourcerecordid><originalsourceid>FETCH-arxiv_primary_2407_140643</originalsourceid><addsrcrecordid>eNqFjrsOgkAQRbexMOoHWDk_IIKCGjsFjTQWxMKOrDiSictCZhcffy8Qe6tT3HOTI8TYcx1_HQTuTPKbns7cd1eO57tLvy9UgnfSpHM411fkrFalIQMRWswslRpIQ3hJIC5k3lgb2N5ujMa0jx1J0-4RYgUnrFmqBvZV8sPAkyTE2iJXjFZeSZH9DEXvLpXB0Y8DMTnsz-Fx2nWlFVMh-ZO2fWnXt_hvfAG7KUaq</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Refining Tuberculosis Detection in CXR Imaging: Addressing Bias in Deep Neural Networks via Interpretability</title><source>arXiv.org</source><creator>Güler, Özgür Acar ; Günther, Manuel ; Anjos, André</creator><creatorcontrib>Güler, Özgür Acar ; Günther, Manuel ; Anjos, André</creatorcontrib><description>Automatic classification of active tuberculosis from chest X-ray images has the potential to save lives, especially in low- and mid-income countries where skilled human experts can be scarce. Given the lack of available labeled data to train such systems and the unbalanced nature of publicly available datasets, we argue that the reliability of deep learning models is limited, even if they can be shown to obtain perfect classification accuracy on the test data. One way of evaluating the reliability of such systems is to ensure that models use the same regions of input images for predictions as medical experts would. In this paper, we show that pre-training a deep neural network on a large-scale proxy task, as well as using mixed objective optimization network (MOON), a technique to balance different classes during pre-training and fine-tuning, can improve the alignment of decision foundations between models and experts, as compared to a model directly trained on the target dataset. At the same time, these approaches keep perfect classification accuracy according to the area under the receiver operating characteristic curve (AUROC) on the test set, and improve generalization on an independent, unseen dataset. For the purpose of reproducibility, our source code is made available online.</description><identifier>DOI: 10.48550/arxiv.2407.14064</identifier><language>eng</language><subject>Computer Science - Computer Vision and Pattern Recognition</subject><creationdate>2024-07</creationdate><rights>http://creativecommons.org/licenses/by/4.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,780,885</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2407.14064$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2407.14064$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Güler, Özgür Acar</creatorcontrib><creatorcontrib>Günther, Manuel</creatorcontrib><creatorcontrib>Anjos, André</creatorcontrib><title>Refining Tuberculosis Detection in CXR Imaging: Addressing Bias in Deep Neural Networks via Interpretability</title><description>Automatic classification of active tuberculosis from chest X-ray images has the potential to save lives, especially in low- and mid-income countries where skilled human experts can be scarce. Given the lack of available labeled data to train such systems and the unbalanced nature of publicly available datasets, we argue that the reliability of deep learning models is limited, even if they can be shown to obtain perfect classification accuracy on the test data. One way of evaluating the reliability of such systems is to ensure that models use the same regions of input images for predictions as medical experts would. In this paper, we show that pre-training a deep neural network on a large-scale proxy task, as well as using mixed objective optimization network (MOON), a technique to balance different classes during pre-training and fine-tuning, can improve the alignment of decision foundations between models and experts, as compared to a model directly trained on the target dataset. At the same time, these approaches keep perfect classification accuracy according to the area under the receiver operating characteristic curve (AUROC) on the test set, and improve generalization on an independent, unseen dataset. For the purpose of reproducibility, our source code is made available online.</description><subject>Computer Science - Computer Vision and Pattern Recognition</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNqFjrsOgkAQRbexMOoHWDk_IIKCGjsFjTQWxMKOrDiSictCZhcffy8Qe6tT3HOTI8TYcx1_HQTuTPKbns7cd1eO57tLvy9UgnfSpHM411fkrFalIQMRWswslRpIQ3hJIC5k3lgb2N5ujMa0jx1J0-4RYgUnrFmqBvZV8sPAkyTE2iJXjFZeSZH9DEXvLpXB0Y8DMTnsz-Fx2nWlFVMh-ZO2fWnXt_hvfAG7KUaq</recordid><startdate>20240719</startdate><enddate>20240719</enddate><creator>Güler, Özgür Acar</creator><creator>Günther, Manuel</creator><creator>Anjos, André</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20240719</creationdate><title>Refining Tuberculosis Detection in CXR Imaging: Addressing Bias in Deep Neural Networks via Interpretability</title><author>Güler, Özgür Acar ; Günther, Manuel ; Anjos, André</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-arxiv_primary_2407_140643</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Computer Science - Computer Vision and Pattern Recognition</topic><toplevel>online_resources</toplevel><creatorcontrib>Güler, Özgür Acar</creatorcontrib><creatorcontrib>Günther, Manuel</creatorcontrib><creatorcontrib>Anjos, André</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Güler, Özgür Acar</au><au>Günther, Manuel</au><au>Anjos, André</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Refining Tuberculosis Detection in CXR Imaging: Addressing Bias in Deep Neural Networks via Interpretability</atitle><date>2024-07-19</date><risdate>2024</risdate><abstract>Automatic classification of active tuberculosis from chest X-ray images has the potential to save lives, especially in low- and mid-income countries where skilled human experts can be scarce. Given the lack of available labeled data to train such systems and the unbalanced nature of publicly available datasets, we argue that the reliability of deep learning models is limited, even if they can be shown to obtain perfect classification accuracy on the test data. One way of evaluating the reliability of such systems is to ensure that models use the same regions of input images for predictions as medical experts would. In this paper, we show that pre-training a deep neural network on a large-scale proxy task, as well as using mixed objective optimization network (MOON), a technique to balance different classes during pre-training and fine-tuning, can improve the alignment of decision foundations between models and experts, as compared to a model directly trained on the target dataset. At the same time, these approaches keep perfect classification accuracy according to the area under the receiver operating characteristic curve (AUROC) on the test set, and improve generalization on an independent, unseen dataset. For the purpose of reproducibility, our source code is made available online.</abstract><doi>10.48550/arxiv.2407.14064</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2407.14064
ispartof
issn
language eng
recordid cdi_arxiv_primary_2407_14064
source arXiv.org
subjects Computer Science - Computer Vision and Pattern Recognition
title Refining Tuberculosis Detection in CXR Imaging: Addressing Bias in Deep Neural Networks via Interpretability
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-08T09%3A33%3A24IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Refining%20Tuberculosis%20Detection%20in%20CXR%20Imaging:%20Addressing%20Bias%20in%20Deep%20Neural%20Networks%20via%20Interpretability&rft.au=G%C3%BCler,%20%C3%96zg%C3%BCr%20Acar&rft.date=2024-07-19&rft_id=info:doi/10.48550/arxiv.2407.14064&rft_dat=%3Carxiv_GOX%3E2407_14064%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true