Automated Tumor Volume Estimation Using Digitized Prostatectomy Specimens.
James P Monaco, John E Tomaszewski, Anant Madabhushi. Rutgers University, Piscataway, NJ; University of Pennsylvania, Philadelphia
Background: Following prostatectomy, the prostate is sliced into histological sections (HSs) which are then analyzed to gather relevant biomarkers such as stage and grade. A promising potential biomarker is tumor volume (TV), which has been shown in some studies to be an independent predictor of recurrence. Unfortunately, the time-consuming nature of assessing TV precludes it from clinical practice. Additionally, the lack of a standard means for measuring TV has resulted in articles both extolling and challenging its prognostic value. For TV to have clinical utility, pathologists require a rapid means for accurately measuring it. Accordingly, we introduce an algorithm that automatically detects the spatial extent of cancerous regions in digitized prostatectomy specimens, providing a means for assessing TV.
Design: The algorithm proceeds as follows: Step 1) The glands are identified and segmented (Figures 1(b) and 1(c)). Step 2) Morphological features are extracted from the segmented boundaries. Step 3) A Markov Random Field based classifier labels each gland as either malignant or benign. The blue dots in Figure 1(d) indicate the centroids of those glands classified as malignant. Step 4) The cancerous glands are consolidated into regions (red polygons in Figure 1(d)). Having determining the cancerous regions, measuring tumor volume becomes trivial.
Results: Figure 2 plots a receiver operator characteristic (ROC) curve indicating the performance (with respect to the area of CaP identified) of our cancer detection system over 40 HSs from 20 patients. Note that at a senstivity of 88% the specificity is 90%. The area under the ROC curve is 0.93.
Conclusions: We introduced an algorithm for automatically measuring TV from prostatectomy specimens. This algorithm could facilitate a large-scale investigation of the prognostic value of TV.
Category: Genitourinary (including renal tumors)
Monday, February 28, 2011 1:00 PM
Poster Session II # 115, Monday Afternoon