FAST-CAD: A Fairness-Aware Framework for Non-Contact Stroke Diagnosis

2025-11-12

AI Summary

Existing non-contact intelligent stroke diagnosis methods exhibit significant fairness disparities across demographic subgroups—such as age, sex, and posture—exacerbating healthcare inequities. To address this, we propose a unified framework integrating domain-adversarial training with group distributionally robust optimization (Group-DRO), the first to provide both convergence guarantees and theoretical fairness bounds for contactless stroke diagnosis. Our method leverages self-supervised representation learning to jointly perform adversarial domain discrimination and domain adaptation, enabling demographic-invariant feature extraction while explicitly optimizing worst-case subgroup performance. Evaluated on a multimodal dataset spanning 12 demographic subgroups, our approach achieves substantial improvements: diagnostic accuracy increases significantly, inter-subgroup performance variance decreases by 42%, and fairness metrics—including Equalized Odds difference—improve by up to 68%. Both theoretical analysis and empirical validation confirm the framework’s efficacy in enhancing robustness and fairness.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Computer Vision: Adversarial Attacks & Robustness
• Intelligent Robots: Learning & Optimization for ROB

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Responsible Web: Algorithmic accountability and transparency on the web
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Stroke is an acute cerebrovascular disease, and timely diagnosis significantly improves patient survival. However, existing automated diagnosis methods suffer from fairness issues across demographic groups, potentially exacerbating healthcare disparities. In this work we propose FAST-CAD, a theoretically grounded framework that combines domain-adversarial training (DAT) with group distributionally robust optimization (Group-DRO) for fair and accurate non-contact stroke diagnosis. Our approach is built on domain adaptation and minimax fairness theory and provides convergence guarantees and fairness bounds. We curate a multimodal dataset covering 12 demographic subgroups defined by age, gender, and posture. FAST-CAD employs self-supervised encoders with adversarial domain discrimination to learn demographic-invariant representations, while Group-DRO optimizes worst-group risk to ensure robust performance across all subgroups. Extensive experiments show that our method achieves superior diagnostic performance while maintaining fairness across demographic groups, and our theoretical analysis supports the effectiveness of the unified DAT + Group-DRO framework. This work provides both practical advances and theoretical insights for fair medical AI systems.

Problem

• Addresses fairness issues in automated stroke diagnosis across demographic groups
• Combines domain-adversarial training with group distributional robust optimization
• Ensures accurate and equitable diagnosis performance across 12 demographic subgroups

Innovation

• Domain-adversarial training for demographic-invariant representations
• Group distributionally robust optimization for worst-group performance
• Unified framework combining DAT and Group-DRO theoretically