Nuclear Structure By Semi-Automated Iiage Analysis Classifies Prostate Cancer Feulgen-Stained Nuclei Captured From Gleason Graded Tissue Patterns.

5291 We assessed the structure of prostate cancer (PCa) nuclei by Quantitative Nuclear Morphometrics(QNM) based on nuclei captured from Gleason grading patterns 3, 4 and 5 of radical prostatectomy (RP) cases. We applied semi-automated image analysis to classify the structure of nuclei in prostate cancer of specific Gleason Grades (GG) from radical prostatectomy (RP) cases. A tissue microarray (TMA) was prepared from RP cases evaluated by a single pathologist (JIE). The TMA cores were prepared from 20 GG-3, 9 GG-4, 10 GG-5 patterns, and 20 benign cancer-adjacent cases (All Gleason scores 3+3) and from normals (no cancer) selected from RP archival paraffin blocks. Feulgen-stained nuclei were captured from all TMA spots for each case of the TMAs (number of objects = 1300 GG-3; 1337 GG-4; 1162 GG-5; and 1300 benign, cancer adjacent nuclei) using the AutoCyteTM Pathology Workstation. Normals (no cancer) were captured from the CPCTR TMA for PCa obtained from the NCI. Multivariate backwards stepwise logistic regression (LR) was employed to calculate a model constant, β-coefficients, predictive indices and predictive probabilities (PP) for each pool of the three GG categories of nuclei compared to the benign cancer-adjacent nuclei and normals (no cancer). QNM was able to calculate models and generate areas of receiver operator characteristic curves (ROC) for differences between benign cancer-adjacent nuclei (same cases) and GG-3 (ROC= 0.82); GG-4 (ROC = 0.87); GG-5 (ROC-AUC = 0.87) using only 8 nuclear morphometric descriptors (NMDs). A comparison of differences between normal (no cancer) nuclei from the CPCTR TMA and GG-3 (ROC= 0.94); GG-4 (ROC= 0.90) and GG-5 (ROC= 0.92) used only 5 NMDs and for both comparisons the variable selection cutoff for Logistic Regression was Pz= 0.000005. Image analysis of PCa nuclear structure comparing normal (no cancer) or benign cancer-adjacent nuclei and expert scored GG patterns from PCa RP cases identified as few as 5-8 NMDs that could differentiate GG 3, 4, and 5 Feulgen-stained nuclei. This is the first step in creating an new objective method for quantifying unique nuclei grading patterns that might be applied to predict disease outcomes.