Improving model fairness in image-based computer-aided diagnosis
Abstract Deep learning has become a popular tool for computer-aided diagnosis using medical images, sometimes matching or exceeding the performance of clinicians. However, these models can also reflect and amplify human bias, potentially resulting inaccurate missed diagnoses. Despite this concern, t...
Main Authors: | , , , , , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Nature Portfolio
2023-10-01
|
Series: | Nature Communications |
Online Access: | https://doi.org/10.1038/s41467-023-41974-4 |
_version_ | 1797557925260034048 |
---|---|
author | Mingquan Lin Tianhao Li Yifan Yang Gregory Holste Ying Ding Sarah H. Van Tassel Kyle Kovacs George Shih Zhangyang Wang Zhiyong Lu Fei Wang Yifan Peng |
author_facet | Mingquan Lin Tianhao Li Yifan Yang Gregory Holste Ying Ding Sarah H. Van Tassel Kyle Kovacs George Shih Zhangyang Wang Zhiyong Lu Fei Wang Yifan Peng |
author_sort | Mingquan Lin |
collection | DOAJ |
description | Abstract Deep learning has become a popular tool for computer-aided diagnosis using medical images, sometimes matching or exceeding the performance of clinicians. However, these models can also reflect and amplify human bias, potentially resulting inaccurate missed diagnoses. Despite this concern, the problem of improving model fairness in medical image classification by deep learning has yet to be fully studied. To address this issue, we propose an algorithm that leverages the marginal pairwise equal opportunity to reduce bias in medical image classification. Our evaluations across four tasks using four independent large-scale cohorts demonstrate that our proposed algorithm not only improves fairness in individual and intersectional subgroups but also maintains overall performance. Specifically, the relative change in pairwise fairness difference between our proposed model and the baseline model was reduced by over 35%, while the relative change in AUC value was typically within 1%. By reducing the bias generated by deep learning models, our proposed approach can potentially alleviate concerns about the fairness and reliability of image-based computer-aided diagnosis. |
first_indexed | 2024-03-10T17:24:08Z |
format | Article |
id | doaj.art-2303727d53f34ecabc28cb178ef5db30 |
institution | Directory Open Access Journal |
issn | 2041-1723 |
language | English |
last_indexed | 2024-03-10T17:24:08Z |
publishDate | 2023-10-01 |
publisher | Nature Portfolio |
record_format | Article |
series | Nature Communications |
spelling | doaj.art-2303727d53f34ecabc28cb178ef5db302023-11-20T10:14:45ZengNature PortfolioNature Communications2041-17232023-10-011411910.1038/s41467-023-41974-4Improving model fairness in image-based computer-aided diagnosisMingquan Lin0Tianhao Li1Yifan Yang2Gregory Holste3Ying Ding4Sarah H. Van Tassel5Kyle Kovacs6George Shih7Zhangyang Wang8Zhiyong Lu9Fei Wang10Yifan Peng11Department of Population Health Sciences, Weill Cornell MedicineSchool of Information, The University of Texas at AustinNational Center for Biotechnology Information, National Library of Medicine, National Institutes of Health (NIH)Department of Electrical and Computer Engineering, The University of Texas at AustinSchool of Information, The University of Texas at AustinDepartment of Ophthalmology, Weill Cornell MedicineDepartment of Ophthalmology, Weill Cornell MedicineDepartment of Radiology, Weill Cornell MedicineDepartment of Electrical and Computer Engineering, The University of Texas at AustinNational Center for Biotechnology Information, National Library of Medicine, National Institutes of Health (NIH)Department of Population Health Sciences, Weill Cornell MedicineDepartment of Population Health Sciences, Weill Cornell MedicineAbstract Deep learning has become a popular tool for computer-aided diagnosis using medical images, sometimes matching or exceeding the performance of clinicians. However, these models can also reflect and amplify human bias, potentially resulting inaccurate missed diagnoses. Despite this concern, the problem of improving model fairness in medical image classification by deep learning has yet to be fully studied. To address this issue, we propose an algorithm that leverages the marginal pairwise equal opportunity to reduce bias in medical image classification. Our evaluations across four tasks using four independent large-scale cohorts demonstrate that our proposed algorithm not only improves fairness in individual and intersectional subgroups but also maintains overall performance. Specifically, the relative change in pairwise fairness difference between our proposed model and the baseline model was reduced by over 35%, while the relative change in AUC value was typically within 1%. By reducing the bias generated by deep learning models, our proposed approach can potentially alleviate concerns about the fairness and reliability of image-based computer-aided diagnosis.https://doi.org/10.1038/s41467-023-41974-4 |
spellingShingle | Mingquan Lin Tianhao Li Yifan Yang Gregory Holste Ying Ding Sarah H. Van Tassel Kyle Kovacs George Shih Zhangyang Wang Zhiyong Lu Fei Wang Yifan Peng Improving model fairness in image-based computer-aided diagnosis Nature Communications |
title | Improving model fairness in image-based computer-aided diagnosis |
title_full | Improving model fairness in image-based computer-aided diagnosis |
title_fullStr | Improving model fairness in image-based computer-aided diagnosis |
title_full_unstemmed | Improving model fairness in image-based computer-aided diagnosis |
title_short | Improving model fairness in image-based computer-aided diagnosis |
title_sort | improving model fairness in image based computer aided diagnosis |
url | https://doi.org/10.1038/s41467-023-41974-4 |
work_keys_str_mv | AT mingquanlin improvingmodelfairnessinimagebasedcomputeraideddiagnosis AT tianhaoli improvingmodelfairnessinimagebasedcomputeraideddiagnosis AT yifanyang improvingmodelfairnessinimagebasedcomputeraideddiagnosis AT gregoryholste improvingmodelfairnessinimagebasedcomputeraideddiagnosis AT yingding improvingmodelfairnessinimagebasedcomputeraideddiagnosis AT sarahhvantassel improvingmodelfairnessinimagebasedcomputeraideddiagnosis AT kylekovacs improvingmodelfairnessinimagebasedcomputeraideddiagnosis AT georgeshih improvingmodelfairnessinimagebasedcomputeraideddiagnosis AT zhangyangwang improvingmodelfairnessinimagebasedcomputeraideddiagnosis AT zhiyonglu improvingmodelfairnessinimagebasedcomputeraideddiagnosis AT feiwang improvingmodelfairnessinimagebasedcomputeraideddiagnosis AT yifanpeng improvingmodelfairnessinimagebasedcomputeraideddiagnosis |