Perelman School of Medicine at the University of Pennsylvania

Section for Biomedical Image Analysis (SBIA)

Neurosurgical Planning

Surgical resection, combined with radiation and/or chemotherapy, is the mainstay treatment for human gliomas, with a growing number of studies demonstrating the benefits of maximal safe resection for patient survival. The goal of modern treatment planning is to determine the optimal resection margin that maximizes tumor removal while preserving language, visual and motor function. The characterization of eloquent anatomic and functional areas that are affected by or in the vicinity of a tumor is therefore essential in limiting patient morbidity and hence improving the patient’s quality of life. Since its introduction, diffusion magnetic resonance imaging (dMRI) has provided critical insight into the white matter fiber pathways of the brain. This has led to the growing use of fiber tractography for surgical planning and it forms one of the key translational applications of dMRI. Identifying peritumoral eloquent tracts requires the ability to perform fiber tracking through regions affected by edema and mass effect. However, the performance and accuracy of currently available clinical tools for dMRI are often substantially degraded in these peritumoral regions, leading to frequent failures in reconstruction of fiber pathways that are edematous, displaced or infiltrated by a tumor. Also, technique-related, non-pathologic failures can occur such as in a region of crossing fibers with DTI. This is compounded by the uniqueness of each patient in terms of the variability introduced by tumors and the associated edema. These limitations diminish clinical utility; furthermore, most clinically used tractography tools require manual placement of seed regions of interest (ROIs) for tracking, which is not only time consuming, but also subject to within and across user variability in ROI placement, differences in parameter selection and variability in software implementations. Additionally, these eloquent tracts have not been fully validated in a clinical scenario, especially due to variability across patients. Thus, there remains an unmet need for a customizable treatment planning package that includes the following: 1) tractography that is robust to edema (with and without infiltration) and mass effect, 2) automated extraction of complete tracts and those disrupted by edema and displaced by mass effect, and 3) validated maps of eloquent tracts. We developing TITAN and Confetti to address and alleviate the shortcomings of current surgical and treatment planning.