Presented at the 1998 APIII Conference                        Return to 1998 Abstract Index

---

IMAGE-BASED COMPUTERIZED PROSTATE CANCER GRADING

Pittsburgh Supercomputing Center
Pittsburgh, Pennsylvania
Arthur W. Wetzel, PhD

An automated image-based prostate cancer grading system is being developed as the result of a collaborative research effort between the Pittsburgh Supercomputing Center (PSC) and the Pathology Informatics group at University of Pittsburgh Medical Center (UPMC). Currently, the system processes captured image data and assigns a numerical value comparable to a Gleason grade, based on significant image features. One novel component is the use of spanning trees and other computational geometry mechanisms built from the positions of cell nuclei and grandular structure to represent the architectural patterns which distinguish Gleason grades. Based on preliminary results, an accuracy comparable to the differences reported between human pathologists is being achieved. A formal evaluation of the system is underway in cooperation with the Medical Informatics evaluation group at the Center for Biomedical Informatics at the University of Pittsburgh.