Duncan and his research team performed seminal work starting in 1987 on the use of parameterized global shape models to incorporate a notion of known prior object shape into the segmentation process using a Bayesian reasoning strategy, helping lead the way towards the use of strategies for automatically finding certain known anatomical structure from any of a variety of medical (e.g. computer tomography (CT), magnetic resonance imaging (MRI), ultrasound) and biological (e.g. confocal microscopy) images.

The strategies he developed have resulted in major advances in bioimaging. He and his research collaborators have applied these strategies to locate the cortical gray matter layer and a variety of co-localized subcortical gray matter structures in the brain as well as to locate the structure near the prostate gland. More recently, Duncan’s team has begun to show that these same techniques will be useful for estimating gray matter-constrained activations from functional MRI data and could help guide the recovery of quantitative biochemical information from MR spectroscopy.