Predicting Individual Brain Glucose Metabolism from MRI

Measuring brain glucose metabolism requires PET, which is expensive, involves ionizing radiation, and is far less widely available than MRI. If metabolism could be partly predicted from MRI, metabolic information could in principle be estimated in the many existing cohorts that have MRI but no PET.

I developed a deep learning framework that predicts individual brain glucose metabolism from structural and functional MRI features, demonstrating that glucose metabolism can potentially be inferred from MRI-based representations.

Summer internship at the Cognitive Neurogenetics Lab, Max Planck Institute for Human Cognitive and Brain Sciences, supervised by Dr. Bin Wan and Prof. Sofie Valk.