Abstract 1292: Validation study on quantitative analysis method of interrelationships between immune cells and prostate epithelial and cancer cells

Abstract

Abstract Context based image analysis of immunohistochemically stained sections has become a topic after several studies have demonstrated that T-cell infiltration is important for survival and of prognostic value. Software algorithms for CD3 detection, epithelial gland detection and location analysis were validated using CD3 stained prostate biopsies. Prostate biopsies (n = 96, 38 with and 58 without tumor areas) were stained with anti-CD3-antibody (DAKO) and hematoxylin and were scanned. For validation, epithelial cell (EC) structures were manually annotated in StrataQuest 5.0 (TissueGnostics) using virtual biopsies. Automated enumeration of CD3 positive cells was controlled visually. For image processing, StrataQuests Easy Gland detection algorithm (SQ-GD) was used as provided. Briefly, hematoxylin shade was used for EC nuclei detection. EC structures were segmented using nuclei compactness and area. EC density was used to detect glands. To determine the relationships, distance transform from EC was added as virtual channel. EC proximity for CD3 distribution analysis were grouped into (i) within EC, (ii) 0-20 μm, (iii) 20-40 μm, (iv) 40-60 μm or (v) more from the EC border. Comparison of automated EC detection by SQ-GD with manually annotated EC showed a significant correlation of both methods (Pearson r = 0.74) although SQ-GD detects significantly more EC. The reason for that was that SQ-GD had to be adjusted so that missing EC areas were below 5%. CD3+/mm2 In ECCD3+/mm2 in 0 to 20μm EC proximityCD3+/mm2 in 20 to 40μm EC proximityCD3+/mm2 in 40 to 60μm EC proximityCD3+/mm2 in more than 60μm EC proximitymanualautomanualautomanualAutomanualautomanualautoTumor300±230390±213244±131148±126231±131191±138202±134168±151159±128128±184Non tumor416±586430±240270±209147±127243±168180±143240±143171±140172±128124±277Correlation0.8230.7380.7100.8020.815 In conclusion, SQ-GD algorithm detects EC in more than 95% of EC-glands and tumor areas. EC, tumor EC and CD3+ cells detection was highly sensitive to staining artifacts at biopsy borders and to tissue folding, both of which had to be manually excluded. Interestingly, we observed no differences between the biopsies negative for prostate cancer and the positive ones. It is noteworthy that the T-cell numbers gradually decrease in correlation with the distance to the epithelium suggesting a certain immunogenicity of EC. Note: This abstract was not presented at the meeting. Citation Format: Gero Kramer, Lukas Kenner, Thomas Szekeres, Shahrokh Shariat, Sarah Schwarz. Validation study on quantitative analysis method of interrelationships between immune cells and prostate epithelial and cancer cells. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 1292. doi:10.1158/1538-7445.AM2015-1292