Nimg-53. Predictive Cytological Topography Maps Regions of Pathologically Confirmed Non-enhancing Hypercellular Tumor in Glioblastoma Patients
Year: 2016
Type: conference-abstract
Abstract: Clinical imaging is important in the detection and treatment of glioblastoma (GBM), where different magnetic resonance imaging (MRI) contrasts to classify tumor and plan treatment. Despite advances in modern MRI technology, differentiation of treatment effects from infiltrative tumor remains a difficult and clinically relevant problem. This study uses a new technique, predictive cytological topography (PiCT), which combines imaging and histology to address this issue. Thirteen patients with GBM were assessed at autopsy. Brains were sliced axially to match pre-mortem MR images and samples were collected from various locations (e.g., center of tumor, surrounding rim, normal white and gray matter). The samples were then digitized and cell nuclei were segmented. Samples were then co-registered to matching MRI slices and cell count was assessed voxel-wise. The MRI values and corresponding cell counts were used to train a machine learning algorithm which was then applied to generate PiCT maps of cell density. To validate these maps, tissue samples were taken from new regions predicted to contain hypercellularity and evaluated. Hypercellularity was defined by our board certified pathologist. Predictive maps correctly identified regions of non-enhancing hypercellularity in all 13 patients. We present a new method for predictively mapping brain tumor cellularity beyond contrast enhancement. This technique could improve treatment planning and disease monitoring.
Source: Neuro-Oncology
Institutions: Medical College of Wisconsin, Froedtert Hospital
Language: English
Subfield: Genetics
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W2588095744