An inception deep architecture to differentiate close-related Gleason prostate cancer scores

摘要

Histopathological tissue analysis is the most effective and definitive method to prognosis cancer and stratify theaggressiveness of the disease. The Gleason Score (GS) is the most powerful grading system based on architecturaltumor pattern quantification. This score characterizes cancer tumor tissue, such as the level of cell differentiationon histopathological images. The reported GS is described as the sum of two principal grades present in aparticular image, and ranged from 6 (cancer grow slowly) to 10 (cancer cells spread more rapidly). A maindrawback of GS is the pathological dependency on histopathological region stratification, which strongly impactsthe clinical procedure to treat the disease. The agreement among experts has been quantified with a kappa indexof: ～ 0:71. Even worse, a higher uncertainty is reported for intermediate grade stratification. This work presentsa like-inception deep architecture that is able to differentiate between intermediate and close GS grades. Eachimage herein evaluated was split-up into regional patches that correspond to a single GS grade. A set of trainingpatches were augmented according to appearance image variations of each grade. Then, a transfer learningscheme was implemented to adapt a bi-Gleason tumor patterns prediction among close levels. The proposedapproach was evaluated on public set of 886 tissue H&E stained images with different GS grades, achieving anaverage accuracy of 0:73% between grades three and four.