Research Project
2107778
DESCRIPTION (adapted from the applicant's abstract): The applicants propose to develop a Patient-Care Advisory (PC Advisor) system utilizing advanced neural network and statistical algorithms to predict post- treatment complications. Neural network can be trained on existing patient records and extract hidden nonlinear high-dimensional relationships between pre-treatment data and post- treatment outcomes. PC Advisor can help physicians decide what tests are needed, what post- operative complications are likely, and help compare the risks of alternative treatments. For maximum utility to health care industry, existing algorithms will be enhanced to handle missing patient data, provide a confidence measure, and explain prediction logic. In Phase I, these algorithms will be developed using real-world coronary artery bypass surgery data.
