Equitable artificial intelligence for glaucoma screening with fair identity normalization
Abstract Glaucoma is the leading cause of irreversible blindness globally. Research indicates a disproportionate impact of glaucoma on racial and ethnic minorities. Existing deep learning models for glaucoma detection might not achieve equitable performance across diverse identity groups. We develop...
Saved in:
| Main Authors: | , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-01-01
|
| Series: | npj Digital Medicine |
| Online Access: | https://doi.org/10.1038/s41746-025-01432-5 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832585375088902144 |
|---|---|
| author | Min Shi Yan Luo Yu Tian Lucy Q. Shen Nazlee Zebardast Mohammad Eslami Saber Kazeminasab Michael V. Boland David S. Friedman Louis R. Pasquale Mengyu Wang |
| author_facet | Min Shi Yan Luo Yu Tian Lucy Q. Shen Nazlee Zebardast Mohammad Eslami Saber Kazeminasab Michael V. Boland David S. Friedman Louis R. Pasquale Mengyu Wang |
| author_sort | Min Shi |
| collection | DOAJ |
| description | Abstract Glaucoma is the leading cause of irreversible blindness globally. Research indicates a disproportionate impact of glaucoma on racial and ethnic minorities. Existing deep learning models for glaucoma detection might not achieve equitable performance across diverse identity groups. We developed fair identify normalization (FIN) module to equalize the feature importance across different identity groups to improve model performance equity. The optical coherence tomography (OCT) measurements were used to categorize patients into glaucoma and non-glaucoma. The equity-scaled area under the receiver operating characteristic curve (ES-AUC) was adopted to quantify model performance equity. With FIN for racial groups, the overall AUC and ES-AUC increased from 0.82 to 0.85 and 0.77 to 0.81, respectively, with the AUC for Blacks increasing from 0.77 to 0.82. With FIN for ethnic groups, the overall AUC and ES-AUC increased from 0.82 to 0.84 and 0.77 to 0.80, respectively, with the AUC for Hispanics increasing from 0.75 to 0.79. |
| format | Article |
| id | doaj-art-a82b532f5db34122b78566063f7d84f5 |
| institution | Kabale University |
| issn | 2398-6352 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | npj Digital Medicine |
| spelling | doaj-art-a82b532f5db34122b78566063f7d84f52025-01-26T12:53:52ZengNature Portfolionpj Digital Medicine2398-63522025-01-018111110.1038/s41746-025-01432-5Equitable artificial intelligence for glaucoma screening with fair identity normalizationMin Shi0Yan Luo1Yu Tian2Lucy Q. Shen3Nazlee Zebardast4Mohammad Eslami5Saber Kazeminasab6Michael V. Boland7David S. Friedman8Louis R. Pasquale9Mengyu Wang10Harvard Ophthalmology AI Lab, Schepens Eye Research Institute of Massachusetts Eye and Ear, Harvard Medical SchoolHarvard Ophthalmology AI Lab, Schepens Eye Research Institute of Massachusetts Eye and Ear, Harvard Medical SchoolHarvard Ophthalmology AI Lab, Schepens Eye Research Institute of Massachusetts Eye and Ear, Harvard Medical SchoolMassachusetts Eye and Ear, Harvard Medical SchoolMassachusetts Eye and Ear, Harvard Medical SchoolHarvard Ophthalmology AI Lab, Schepens Eye Research Institute of Massachusetts Eye and Ear, Harvard Medical SchoolHarvard Ophthalmology AI Lab, Schepens Eye Research Institute of Massachusetts Eye and Ear, Harvard Medical SchoolMassachusetts Eye and Ear, Harvard Medical SchoolMassachusetts Eye and Ear, Harvard Medical SchoolEye and Vision Research Institute, Icahn School of Medicine at Mount SinaiHarvard Ophthalmology AI Lab, Schepens Eye Research Institute of Massachusetts Eye and Ear, Harvard Medical SchoolAbstract Glaucoma is the leading cause of irreversible blindness globally. Research indicates a disproportionate impact of glaucoma on racial and ethnic minorities. Existing deep learning models for glaucoma detection might not achieve equitable performance across diverse identity groups. We developed fair identify normalization (FIN) module to equalize the feature importance across different identity groups to improve model performance equity. The optical coherence tomography (OCT) measurements were used to categorize patients into glaucoma and non-glaucoma. The equity-scaled area under the receiver operating characteristic curve (ES-AUC) was adopted to quantify model performance equity. With FIN for racial groups, the overall AUC and ES-AUC increased from 0.82 to 0.85 and 0.77 to 0.81, respectively, with the AUC for Blacks increasing from 0.77 to 0.82. With FIN for ethnic groups, the overall AUC and ES-AUC increased from 0.82 to 0.84 and 0.77 to 0.80, respectively, with the AUC for Hispanics increasing from 0.75 to 0.79.https://doi.org/10.1038/s41746-025-01432-5 |
| spellingShingle | Min Shi Yan Luo Yu Tian Lucy Q. Shen Nazlee Zebardast Mohammad Eslami Saber Kazeminasab Michael V. Boland David S. Friedman Louis R. Pasquale Mengyu Wang Equitable artificial intelligence for glaucoma screening with fair identity normalization npj Digital Medicine |
| title | Equitable artificial intelligence for glaucoma screening with fair identity normalization |
| title_full | Equitable artificial intelligence for glaucoma screening with fair identity normalization |
| title_fullStr | Equitable artificial intelligence for glaucoma screening with fair identity normalization |
| title_full_unstemmed | Equitable artificial intelligence for glaucoma screening with fair identity normalization |
| title_short | Equitable artificial intelligence for glaucoma screening with fair identity normalization |
| title_sort | equitable artificial intelligence for glaucoma screening with fair identity normalization |
| url | https://doi.org/10.1038/s41746-025-01432-5 |
| work_keys_str_mv | AT minshi equitableartificialintelligenceforglaucomascreeningwithfairidentitynormalization AT yanluo equitableartificialintelligenceforglaucomascreeningwithfairidentitynormalization AT yutian equitableartificialintelligenceforglaucomascreeningwithfairidentitynormalization AT lucyqshen equitableartificialintelligenceforglaucomascreeningwithfairidentitynormalization AT nazleezebardast equitableartificialintelligenceforglaucomascreeningwithfairidentitynormalization AT mohammadeslami equitableartificialintelligenceforglaucomascreeningwithfairidentitynormalization AT saberkazeminasab equitableartificialintelligenceforglaucomascreeningwithfairidentitynormalization AT michaelvboland equitableartificialintelligenceforglaucomascreeningwithfairidentitynormalization AT davidsfriedman equitableartificialintelligenceforglaucomascreeningwithfairidentitynormalization AT louisrpasquale equitableartificialintelligenceforglaucomascreeningwithfairidentitynormalization AT mengyuwang equitableartificialintelligenceforglaucomascreeningwithfairidentitynormalization |