A Framework for Exploring and Disentangling Intersectional Bias: A Case Study in Fetal Ultrasound

TL;DR

This paper introduces a structured framework to explore intersectional bias in fetal ultrasound AI, revealing how image quality factors like pixel spacing influence model performance across demographic groups.

Contribution

It presents a novel framework combining unsupervised discovery and systematic analysis to detect intersectional bias in medical imaging AI models.

Findings

Pixel spacing significantly affects model performance, with up to 24% improvement in some subgroups.

Acquisition parameters like pixel spacing can confound demographic effects such as BMI and gestational age.

Interaction-aware evaluation reveals persistent biases beyond simple data imbalance.

Abstract

Bias in medical AI is often framed as a problem of representation. However, in image-based tasks such as fetal ultrasound, performance disparities can arise even when representation is adequate, because predictive accuracy depends strongly on image quality. Image quality is shaped by acquisition conditions and operator expertise, as well as patient-dependent factors such as maternal body mass index (BMI), all of which may correlate with sensitive demographic features. Consequently, observed disparities may reflect the combined influence of demographic, clinical, and acquisition-related factors rather than data imbalance alone, and may obscure underlying interaction or confounding effects. We propose a structured framework to explore and detect intersectional bias, combining unsupervised slice discovery, systematic factor-wise analysis, and targeted intersectional evaluation. In a case study of over 94{,}000 ultrasound images for fetal weight estimation, we analyze bias in a state-of-the-art deep learning (DL) model and the clinical standard Hadlock, a regression formula using biometric measurements. Pixel spacing (PS) -- a parameter considered suboptimal in current acquisition protocols -- emerged as a consistent driver of performance differences, with higher PS associated with improvements of up to 24\% in selected subgroups for both models. Because PS is often adapted in cases of high BMI or low gestational age (GA), this effect carries a substantial risk of confounding. Our intersectional analysis revealed that part of the PS-associated signal is explained by GA, while PS-related improvements persist across BMI strata, highlighting the importance of acquisition-aware and interaction-aware evaluation in medical AI fairness research.