Aim: The aim of the study is to evaluate the three main components of the tumor architecture in correlation with two different grading systems of prostate adenocarcinoma (PA) using the fractal dimension (FD) analysis.
Patients, materials and methods: 433 fields with different patterns of PA selected from 83 patients with total prostatectomy according to Gleason and Srigley grading systems were selected. Four serial sections were cut and stained in order to assess the following parameters: tumor grading with Hematoxylin-Eosin (H-E), tumor cells architecture (GÖ) with Gömöri technique, tumor stroma architecture (TC) with Goldner's trichrome, and vascular network (VN) architecture with cluster of differentiation 34 (CD34) immunomarker. Images were binarized with variable user-defined empiric threshold for Goldner's trichrome staining and CD34 immunostaining and k-nearest neighbor approach for GÖ staining. The FD was computed for each binary image using a box-counting algorithm. The three computed values were used for clustering and classification, k-nearest neighbor proving to be a good choice with a classification rate, due to the irregular distribution of cases in different patterns. Values tending to "1" had the meaning of a more "Linear type" distribution and values tending to "2" had the meaning of a more "Area type" distribution.
Results: Tumor cells architecture had a more ordered smooth ascending trend towards "area-like" type of distribution (with FD>1.5) in Srigley system than in Gleason system. Tumor stroma architecture had almost the same type of distribution - between "linear-like" and "area-like" (FD~1.5) - in both grading systems. VN architecture had a more "linear-like" type of distribution (FD<1.5), with a descending trend towards high-grade patterns in both systems. Tumor cells architecture had a direct correlation with tumor stroma architecture and VN architecture (p-value of Pearson's test <0.001), while tumor stroma architecture and VN architecture proved no correlation (p-value of Pearson's test >0.05), irrespective of grading pattern.
Conclusions: Tumor cell population is remodeling and adapting TC and VN in the same way its architectural disposal evolves. TC and VN develop independently of each other, the former towards "Area type" and the latter towards "Linear type" of architectural disposal as the degree of differentiation is decreasing. FD analysis proved that Srigley system is more accurate in grading PA than Gleason system.