Clinical decision support that integrates multiple elements of laboratory data could be highly useful in enhancing laboratory diagnosis. Using the analyte ferritin in a proof-of-concept, we extracted clinical laboratory data from patient testing and applied a variety of machine learning algorithms to predict ferritin test results using the results from other tests. We show that predicted ferritin results may sometimes better reflect underlying iron status than measured ferritin. We next integrate temporality into predicting multi-variate analytes. We devise an algorithm termed 3D-MICE alternating between cross-sectional imputation and auto-regressive imputation. We show modest performance improvement of the combined algorithm compared to either component alone. We then integrate Gaussian process with mixture model and introduce individualized mixing weights to handle variance in predictive confidence of Gaussian process components. Experiments show that our best model can provide more accurate imputation than the state-of-the-art including 3D-MICE on both synthetic and real-world datasets.
After participating in this activity, the learner should be better able to:
- Learn a novel machine learning method to impute the lab test results of interest, e.g. imputing values for the analyte ferritin.
- Learn the available methods to impute the lab test results of interest.
Dr. Yuan Luo, PhD
Assistant Professor, Division of Health and Biomedical Informatics, Department of Preventive Medicine, Feinberg School of Medicine
Department of Industrial Engineering and Management Science, McCormick School of Engineering (Courtesy)
Department of Electrical Engineering and Computer Science, McCormick School of Engineering (Courtesy)
Yuan Luo is an Assistant Professor at the Department of Preventive Medicine, Division of Health & Biomedical Informatics at Feinberg School of Medicine in Northwestern University, with courtesy appointments in IEMS and EECS. He earned his PhD degree from MIT EECS with a math minor and a certificate in Graduate Education of Medical Science (GEMS). His research interests include machine learning, natural language processing, time series analysis, computational phenotyping and integrative genomics, with a focus on medical applications. His PhD Thesis was awarded the inaugural Doctoral Dissertation Award Honorable Mention by American Medical Informatics Association (AMIA) in 2017. He won the first prize at the NLP Doctoral Consortium in 2013 at the AMIA Annual Symposium. He is currently an editor with Plus One and JAMIA Open and was on the Student Editorial Board for JAMIA. He co-chairs the eMERGE NLP WG, serves on the AMIA Membership and Outreach Committee, and Bluhm Cardiovascular Institute Research/Innovation and Education Committee. He has served as PC members for top AI and informatics conferences including AAAI, IJCAI, AMIA, AMIA Joint Summits, IEEE ICHI, ACM BCB, CIKM etc. His research is funded by multiple NIH R01 and R21 grants (PI or MPI). He has given multiple plenary talks including at the AMIA Annual Symposium 2017 and at the China National Cancer Center in 2018.