RadFusion: Benchmarking Performance and Fairness for Multimodal Pulmonary Embolism Detection from CT and EHR
This addresses fairness and accuracy issues in medical AI for pulmonary embolism detection, though it is incremental as it builds on existing multimodal fusion methods with a new benchmark.
The authors tackled the problem of bias and limited performance in unimodal deep learning for medical imaging by creating RadFusion, a multimodal dataset of 1794 patients with CT scans and EHR data for pulmonary embolism detection. Their results show that integrating imaging and EHR data improves classification performance and robustness while minimizing disparities in true positive rates across subgroups.
Despite the routine use of electronic health record (EHR) data by radiologists to contextualize clinical history and inform image interpretation, the majority of deep learning architectures for medical imaging are unimodal, i.e., they only learn features from pixel-level information. Recent research revealing how race can be recovered from pixel data alone highlights the potential for serious biases in models which fail to account for demographics and other key patient attributes. Yet the lack of imaging datasets which capture clinical context, inclusive of demographics and longitudinal medical history, has left multimodal medical imaging underexplored. To better assess these challenges, we present RadFusion, a multimodal, benchmark dataset of 1794 patients with corresponding EHR data and high-resolution computed tomography (CT) scans labeled for pulmonary embolism. We evaluate several representative multimodal fusion models and benchmark their fairness properties across protected subgroups, e.g., gender, race/ethnicity, age. Our results suggest that integrating imaging and EHR data can improve classification performance and robustness without introducing large disparities in the true positive rate between population groups.