Prognostic Reporting of p53 Expression by Image Analysis in Glioblastoma Patients: Detection and Classification

摘要

In this paper, we present a computer aided diagnosis system focusing on one important diagnostic branchpoint in clinical decision-making: prognostic reporting of p53 expression in glioblastoma patients. Studies in other tumor paradigms have shown that the staining intensity correlates with TP53 mutation status, and that gliomas show inter-tumoral heterogeneity in p53 mutation status. Increasing diagnostic accuracy by computer-aided image analysis algorithms would deliver an objective assessment of such prognostic biomark-ers. We proposed a method for the detection and classification of positive and negative cells in digitized p53-stained images by means of a novel adaptive thresholding for the detection, and two-step rule based on weighted color and intensity for the classification. The proposed thresholding technique is able to correctly locate both positive and negative cells by effectively addressing the closely connected cells problem, and records a promising 85% average precision and 88% average recall rate. On the other hand, the proposed two-step rule achieves 81% classification accuracy, which is comparable with neuropathologists' markings.