DUDE: deep unsupervised domain adaptation using variable nEighbors for physiological time series analysis

Objective.Deep learning for continuous physiological signals, such as electrocardiography or oximetry, has achieved remarkable success in supervised learning scenarios where training and testing data are drawn from the same distribution. However, when evaluating real-world applications, models often fail to generalize due to distribution shifts between the source domain on which the model was trained and the target domain where it is deployed. A common and particularly challenging shift often encountered in reality is where the source and target domain supports do not fully overlap. In this paper, we propose a novel framework, named Deep Unsupervised Domain adaptation using variable nEighbors (DUDE), to address this challenge.Approach.We introduce a new type of contrastive loss between the source and target domains using a dynamic neighbor selection strategy, in which the number of neighbors for each sample is adaptively determined based on the density observed in the latent space. We use multiple real-world datasets as source and target domains, with target domains that included demographics, ethnicities, geographies, and comorbidities that were not present in the source domain.Main results.The experimental results demonstrate superior DUDE performance compared to baselines and with an improvement of up to 16% over the original Nearest-Neighbor Contrastive Learning of Visual Representations strategy.Significance.Our contribution provides evidence on the potential of using DUDE to bridge the crucial gap of domain adaptation in medicine, potentially transforming patient care through more precise and adaptable diagnostic tools.