Description:

Abstract: ICU mortality risk stratification may help clinicians take effective interventions to improve patient outcome. Existing machine learning approaches often face challenges in integrating a comprehensive panel of physiologic variables and presenting to clinicians interpretable models. We aim to improve both accuracy and interpretability of prediction models by introducing Subgraph Augmented Non-negative Matrix Factorization (SANMF) on ICU physiologic time series. SANMF converts time series into a graph representation and applies frequent subgraph mining to automatically extract temporal trends. We then apply non-negative matrix factorization to group trends in a way that approximates patient pathophysiologic states. Trend groups are then used as features in training a logistic regression model for mortality risk prediction, and are also ranked according to their contribution to mortality risk. We evaluated SANMF against four empirical models on the task of predicting mortality or survival 30 days after discharge from ICU using the observed physiologic measurements between 12 and 24 hours after admission. SANMF outperforms all comparison models, and in particular, demonstrates an improvement in AUC (0.848 vs. 0.827, p<0.002) compared to a state-of-the-art machine learning method that uses manual feature engineering. Feature analysis was performed to illuminate insights and benefits of subgraph groups in mortality prediction.

Short Bio: Yuan Luo is an Assistant Professor at Department of Preventive Medicine, Division of Health & Biomedical Informatics with courtesy appointments in IEMS and EECS. He earned his PhD degree from MIT EECS. His research interests include machine learning, natural language processing, time series analysis, computational genomics and big data analytics, with a focus on medical applications. He proposed Subgraph Augmented Non-negative Tensor Factorization (SANTF) for building a clinical model that improves both accuracy and interpretability, by turning narrative text into graph representations and applying tensor factorization to mining graph features. This work was awarded the first prize at NLP Doctoral Consortium in 2013 Annual Symposium of the American Medical Informatics Association. He has extended the subgraph mining and factorization models to time series analysis and computational genomics. He was also a member of the Student Editorial Board for Journal of the American Medical Informatics Association.