Title:
Machine Learning, Biostatistics, and the Challenge of Evaluating AI Models in Healthcare
Abstract:
My PhD research has taken place at the intersection of machine learning, biostatistics, and healthcare, where the evaluation of AI models requires more than the reporting of predictive performance alone. In many healthcare applications, the central questions are not limited to whether a model achieves good accuracy, but also to whether its results are statistically sound, clinically meaningful, and aligned with established research practices. In this talk, lessons drawn from work in Alzheimer’s disease research using UK Biobank data, machine learning model development, and vision-based approaches for biomedical applications will be presented. Attention will be given to the different ways in which evidence is interpreted across these domains, including hypothesis testing, effect estimation, and statistical significance on one hand, and classification, regression, and generalization performance on the other. More broadly, it will be argued that the evaluation of AI in healthcare requires these perspectives to be bridged rather than treated as separate alternatives. These experiences will also be used to reflect on some of the considerations involved in developing machine learning methods for healthcare settings.
Speaker:
Antony Garcia Gonzalez
PhD Candidate, ECE Department, WPI
Bio:
Antony Garcia Gonzalez is a Ph.D. candidate at Worcester Polytechnic Institute (WPI), where he is a member of the ECE Embedded Computing Lab and conducts research under the supervision of Dr. Xinming Huang. He is a Fulbright scholar. He received his B.S. in Electromechanical Engineering and his M.S. in Electrical Power Systems Engineering from the Technological University of Panama. He also serves as a Research Associate in the Faculty of Electrical Engineering at the Technological University of Panama. His research interests include machine learning, deep learning, and artificial intelligence, with a focus on healthcare applications, particularly Alzheimer’s disease research and skin cancer classification.
Host: Professor Alex Wyglinski