Harmonization with Flow-based Causal Inference
This addresses the issue of data heterogeneity for medical researchers and clinicians, but it is incremental as it builds on existing causal and flow-based methods.
The paper tackled the problem of heterogeneity in medical data, which hinders generalization of machine learning models, by using a normalizing-flow-based method for causal inference to harmonize data, achieving better cross-domain generalization than state-of-the-art algorithms on multiple real-world datasets.
Heterogeneity in medical data, e.g., from data collected at different sites and with different protocols in a clinical study, is a fundamental hurdle for accurate prediction using machine learning models, as such models often fail to generalize well. This paper leverages a recently proposed normalizing-flow-based method to perform counterfactual inference upon a structural causal model (SCM), in order to achieve harmonization of such data. A causal model is used to model observed effects (brain magnetic resonance imaging data) that result from known confounders (site, gender and age) and exogenous noise variables. Our formulation exploits the bijection induced by flow for the purpose of harmonization. We infer the posterior of exogenous variables, intervene on observations, and draw samples from the resultant SCM to obtain counterfactuals. This approach is evaluated extensively on multiple, large, real-world medical datasets and displayed better cross-domain generalization compared to state-of-the-art algorithms. Further experiments that evaluate the quality of confounder-independent data generated by our model using regression and classification tasks are provided.