Dr Franco Pestilli, Stanford University
3pm in W342/2
Model-based neuroanatomy: Statistical inference on white-matter connections in the living human brain
Advances in diffusion weighted imaging coupled with computational tractography techniques enable estimates of the network of white-matter fascicles in the living brain. Measuring fascicles in the living human brain allows us to study the relationship between fascicle properties and human behavior, cognition, development and disease. This work opens the possibility of understanding how white matter properties influence human cognition in health and disease.
Tractography algorithms take diffusion measurements as input and produce an estimate of the collection of white-matter fascicles (connectome) as output. Critical to the study of the white matter is quantifying the confidence in the estimated connectome. I will introduce a method, Linear Iterative Fascicle Evaluation (LIFE), to do so. By applying fundamental diffusion equations LIFE takes the connectome as input and produces predicted diffusion signals as output. The algorithm then identifies the fascicles that contribute significantly to predicting the diffusion data and eliminates the rest. Applying the procedure iteratively yields an optimal connectome: the smallest set of fascicles that accurately predicts the diffusion data.
By calculating prediction accuracy the method provides a way to implement simple statistical tests about the network of white matter fascicles. For example, the evidence in favor of the existence of a collection of fascicles can be evaluated by calculating whether these fascicles explain a significant proportion of the diffusion signal. I will illustrate ways to test three types of neuroanatomical hypotheses: (a) the existence of a fascicle within the white matter, (b) the geometric arrangement of multiple fascicles, and (c) the evidence in favor of specific cortico-cortical connections.