Student Investigator: Alex Larson
Mentored by Professor Tamás Budavári
This project applies machine learning to predict which post-operative patients in the Neurosciences Critical Care Unit (NCCU) will develop complications. Using historical data such as medical conditions and medication records, we’ve developed Bayesian neural network models to estimate complication risk while accounting for uncertainty. Preprocessing of new data includes medication standardization, unit normalization, and feature selection using statistical techniques. Further analysis of medication data—such as therapeutic classes and dosage patterns—is planned to improve predictive power. We plan to incorporate MRI imaging data via neural networks and variational autoencoders to enhance performance. Model interpretability through calibration and SHAP values will support clinical decision-making. The ultimate goal is to build a reliable, interpretable tool to assist NCCU decisions and improve resource allocation.
Alex Larson is a junior at Johns Hopkins University majoring in Applied Mathematics & Statistics with a minor in Physics. He is from New York, also enjoys playing baseball and plans to pursue a PhD in astrophysics, where he hopes to apply data science and statistical modeling to uncover patterns in large-scale datasets.
