Multimodal Pretraining of Medical Time Series and Notes
This work addresses data scarcity in critical care for healthcare professionals, offering an incremental improvement in self-supervised learning for medical applications.
The paper tackled the challenge of limited labeled data in ICU analysis by proposing a self-supervised pretraining method that aligns clinical measurements and notes, resulting in improved performance in downstream tasks such as in-hospital mortality prediction (AUC-ROC increase of 0.17) and phenotyping (AUC-PR increase of 0.1) with only 1% of labels available.
Within the intensive care unit (ICU), a wealth of patient data, including clinical measurements and clinical notes, is readily available. This data is a valuable resource for comprehending patient health and informing medical decisions, but it also contains many challenges in analysis. Deep learning models show promise in extracting meaningful patterns, but they require extensive labeled data, a challenge in critical care. To address this, we propose a novel approach employing self-supervised pretraining, focusing on the alignment of clinical measurements and notes. Our approach combines contrastive and masked token prediction tasks during pretraining. Semi-supervised experiments on the MIMIC-III dataset demonstrate the effectiveness of our self-supervised pretraining. In downstream tasks, including in-hospital mortality prediction and phenotyping, our pretrained model outperforms baselines in settings where only a fraction of the data is labeled, emphasizing its ability to enhance ICU data analysis. Notably, our method excels in situations where very few labels are available, as evidenced by an increase in the AUC-ROC for in-hospital mortality by 0.17 and in AUC-PR for phenotyping by 0.1 when only 1% of labels are accessible. This work advances self-supervised learning in the healthcare domain, optimizing clinical insights from abundant yet challenging ICU data.