Comparing Prostate Cancer Area Assessment In Radical Prostatectomy Between Visual Assessment By Pathologists And Computer Assisted Tumor Mapping In Digital Slides

Abstract

Background: Prostate cancer reporting guideline of the CAP requires reporting of the percentage of tumor volume in radical prostatectomy. Visual inspection by pathologist has been reported to be the commonest method of assessment, despite its subjectivity and presence of inter-observer variation. Assessment of percentage of tumor area in individual slide is basic to subsequent tumor volume assessment. With recent advance in digital pathology and availability of high-speed slide scanners, this pilot study aims at comparing assessment results of tumor area percentage by visual assessment amongst pathologists versus computer assisted tumor mapping in scanned digital slides. Design: 30 microscopic slides with prostatic acinar adenocarcinoma from 9 radical prostatectomy specimens were selected. Tumor involvement ranged from tiny tumor focus to diffuse tumor infiltration. 4 attending pathologists with a special interest in urologic pathology and 4 residents independently performed visual assessment of tumor area percentage. These slides were scanned by NanoZoomer S210 digital slide scanner (Hamamatsu). Tumor mapping and area assessment of the digital slides were independently performed by 2 attending pathologists using QuPath (Version: 0.1.2; Bankhead, P. et al. QuPath: Open source software for digital pathology image analysis. Scientific Reports (2017)) computer software (Figure 1). The results were statistically analyzed using SPSS 25.0 (IBM). Results: The median tumor area assessment by visual assessment of pathologists was statistically significantly higher than that of tumor mapping of digital slides using computer software (Figure 2, Wilcoxon Signed-Ranks Test, Z = 463.00, p<0.001). The mean tumor area assessment of individual slide was higher by visual assessment amongst pathologists in 29 of the 30 slides (see Table), and was statistically significant in 21 slides (2-tailed t test, p<0.05). 10 slides show mean difference of area assessment of more than 10%. No statistically significant difference in mean visual assessment of tumor area was observed between attending pathologists and residents in all 30 slides (2-tailed t test, p>0.05). Conclusion: In this pilot study, visual tumor area assessment by pathologists in radical prostatectomy results in significantly higher tumor area percentage estimation than tumor mapping of digital slides. This suggests the need for objective measurements including the use of image analysis in prostate cancer quantification in radical prostatectomy.