3D modelling of radical prostatectomy specimens: Developing a method to quantify tumor morphometry for prostate cancer risk prediction

摘要

Abstract Prostate cancer displays a wide spectrum of clinical behaviour from biological indolence to rapidly lethal disease, but we remain unable to accurately predict an individual tumor’s future clinical course at an early curable stage. Beyond basic dimensions and volume calculations, tumor morphometry is an area that has received little attention, as it requires the analysis of the prostate gland and tumor foci in three-dimensions. Previous efforts to generate three-dimensional prostate models have required specialised graphics units and focused on the spatial distribution of tumors for optimisation of biopsy strategies rather than to generate novel morphometric variables such as tumor surface area. Here, we aimed to develop a method of creating three-dimensional models of a prostate’s pathological state post radical prostatectomy that allowed the derivation of surface areas and volumes of both prostate and tumors, to assess the method’s accuracy to known clinical data, and to perform initial investigation into the utility of morphometric variables in prostate cancer prognostication. Serial histology slides from 21 prostatectomy specimens covering a range of tumor sizes and pathologies were digitised. Computer generated three-dimensional models of tumor and prostate space filling models were reconstructed from these scanned images using Rhinoceros 4.0 spatial reconstruction software. Analysis of three-dimensional modelled prostate volume correlated only moderately with weak concordance to that from the clinical data (r=0.552, θ=0.405), but tumor volume correlated well with strong concordance (r=0.949, θ=0.876). We divided the cohort of 21 patients into those with features of aggressive tumor versus those without and found that larger tumor surface area (32.7 vs 3.4cc, p=0.008) and a lower tumor surface area to volume ratio (4.7 vs 15.4, p=0.008) were associated with aggressive tumor biology.