Document Type

Conference Paper

Publication Date

2024

DOI

10.1117/12.3008748

Publication Title

Proceedings Volume 12927: Medical Imaging 2024: Computer-Aided Diagnosis

Volume

12927

Pages

129271M (1-8)

Conference Name

SPIE Medical Imaging 2024: Computer-Aided Diagnosis, 18-23 February 2024, San Diego, California

Abstract

Deep learning models have shown potential in medical image analysis tasks. However, training a generalized deep learning model requires huge amounts of patient data that is usually gathered from multiple institutions which may raise privacy concerns. Federated learning (FL) provides an alternative to sharing data across institutions. Nonetheless, FL is susceptible to a few challenges including inversion attacks on model weights, heterogenous data distributions, and bias. This study addresses heterogeneity and bias issues for multi-institution patient data by proposing domain adaptive FL modeling using several radiomics (volume, fractal, texture) features for O6-methylguanine-DNA methyltransferase (MGMT) classification across multiple institutions. The proposed domain adaptive FL MGMT classification inherently offers differential privacy (DP) for the patient data. For domain adaptation two techniques e.g., mixture of experts (ME) with a gating network and adversarial alignment are used for comparison. The proposed method is evaluated using publicly available multi-institution (UPENN-GBM, UCSF-PDGM, RSNA-ASNR-MICCAI BraTS-2021) data set with a total of 1007 patients. Our experiments with 5-fold cross validation suggest that domain adaptive FL offers improved performance with a mean accuracy of 69.93% ± 4.8 % and area under curve of 0.655 ± 0.055 across multiple institutions. In addition, further analysis of probability density of gating network for domain adaptive FL identifies the institution that may bias the global model prediction due to increased heterogeneity for a given input. Our comparison analysis shows that the proposed method with bias identification offers the best predictive performance when compared to different commonly employed FL and baseline methods in the literature.

Rights

© 2024 Society of Photo-Optical Instrumentation Engineers (SPIE). One print or electronic copy may be made for personal use only. Systematic reproduction and distribution, duplication of any material in this publication for fee or for commercial purposes, and modification of the contents of this publication are prohibited.

"SPIE grants to authors (and their employers) of papers, posters, and presentation recordings published in SPIE Proceedings or SPIE Journals on the SPIE Digital Library (hereinafter "publications") the right to post an author-prepared version or an official version (preferred) of the publication on an internal or external server controlled exclusively by the author/employer or the entity funding the research, provided that (a) such posting is noncommercial in nature and the publication is made available to users without charge; (b) an appropriate copyright notice and citation appear with the publication; and (c) a link to SPIE's official online version of the publication is provided using the item's DOI."

Original Publication Citation

Farzana, W., Witherow, M. A., Longoria, I., Sadique, M. S., Temtam, A., & Iftekharuddin, K. M. (2024). Domain adaptive federated learning for multi-institution molecular mutation prediction and bias identification. In W. Chen & S. M. Astley (Eds.), Medical Imaging 2024: Computer-Aided Diagnosis, Proc. of SPIE 12927 (129271M). SPIE. https://doi.org/10.1117/12.3008748

ORCID

0000-0003-1995-2426 (Farzana), 0000-0002-6578-4657 (Witherow), 0000-0002-6734-6802 (Sadique), 0000-0001-8316-4163 (Iftekharuddin)

Download

Share

COinS