Identification of Morphological Features of Fibrillar Collagen Indicative of Tumor Progression in Bladder Cancer

Abstract

ObjectiveBladder cancer is the most common malignancy involving the urinary system. As with all types of cancer, accurate monitoring of stage progression is critical in determining proper treatment strategies and predicting clinical outcomes. Collagen fibers are a major component of tumor microenvironment, and their post‐translational modifications during the course of tumor progression are well‐documented in a number of malignancies including bladder cancer. In the present feasibility study we sought to determine whether quantitative morphological features extracted from collagen fibers can be used to distinguish between urothelial carcinoma tumors pre‐muscle‐invasion (stage pT1) and post‐muscle‐invasion (stage pT2).MethodsPicroSirius Red (PSR) staining was used to detect fibrillar collagen in 8 tissue microarray (TMA) cores acquired from Biomax (4 stage pT1, 4 stage pT2). We applied auto‐fluorescence imaging techniques to detect the PSR signal, and employed the CT‐Fire freeware to automatically extract individual collagen fibers. Five morphological features (including length, width, straightness, orientation, and anisotropy) were computed for each detected fiber. Individual features were then combined within each slide to form five corresponding histograms per TMA core. Finally, for each histogram we computed two measures of central tendency (i.e., mean and median), two measures of spread (i.e., range and standard deviation), and two measures of symmetry (i.e., skewness and kurtosis).ResultsTo identify any statistically‐significant differences between the pT1 and pT2 groups, we employed the student’s t‐test (run in Python 3.8.5) to compare the two groups with respect to each of the five morphological features and across each of the six distributional measures mentioned in the Methods section (30 comparisons in total). Of these 30 comparisons, all but one (i.e., median) of the six straightness measures resulted in a p‐value < 0.05. Further, each of the two measures of spread (i.e., range and standard deviation) of straightness returned a p‐value < 0.001. We are currently working on expanding the size of our dataset to further validate these results.ConclusionOur preliminary findings are in line with the hypothesis that collagen remodeling is strongly associated with muscle‐invasive bladder cancer, and suggest that quantitative features measuring the straightness of individual fibers may be used to improve tumor staging. This type of quantitative approach is particularly helpful in distinguishing between stage pT1 and stage pT2 of bladder cancer, a task which can be challenging at times even for experienced pathologists.