Convolutional Neural Network Quantification of Gleason Pattern 4 and Association with Biochemical Recurrence in Intermediate Grade Prostate Tumors
Recommended Citation
Chen Y, Loveless IM, Nakai T, Newaz R, Abdollah FF, Rogers CG, Hassan O, Chitale D, Arora K, Williamson SR, Gupta NS, Rybicki BA, Sadasivan SM, and Levin AM. Convolutional Neural Network Quantification of Gleason Pattern 4 and Association with Biochemical Recurrence in Intermediate Grade Prostate Tumors. Mod Pathol 2023; 100157.
Document Type
Article
Publication Date
3-14-2023
Publication Title
Modern pathology
Keywords
Male, Humans, Artificial Intelligence, Reproducibility of Results, Prostatic Neoplasms/pathology, Neoplasm Grading, Prostatectomy, Neural Networks, Computer, Neoplasm Recurrence, Local, Gleason grade, artificial intelligence, digital pathology, percent Gleason pattern 4, prostate cancer
Abstract
Differential classification of prostate cancer (CaP) grade group (GG) 2 and 3 tumors remains challenging, likely due to the subjective quantification of percentage of Gleason pattern 4 (%GP4). Artificial intelligence assessment of %GP4 may improve its accuracy and reproducibility and provide information for prognosis prediction. To investigate this potential, a convolutional neural network (CNN) model was trained to objectively identify and quantify Gleason pattern (GP) 3 and 4 areas, estimate %GP4, and assess whether CNN-assessed %GP4 is associated with biochemical recurrence (BCR) risk in intermediate risk GG 2 and 3 tumors. The study was conducted in a radical prostatectomy cohort (1999-2012) of African American men from the Henry Ford Health System (Detroit, Michigan). A CNN model that could discriminate four tissue types (stroma, benign glands, GP3 glands, and GP4 glands) was developed using histopathologic images containing GG 1 (n=45) and 4 (n=20) tumor foci. The CNN model was applied to GG 2 (n=153) and 3 (n=62) for %GP4 estimation, and Cox proportional hazard modeling was used to assess the association of %GP4 and BCR, accounting for other clinicopathologic features including GG. The CNN model achieved an overall accuracy of 86% in distinguishing the four tissue types. Further, CNN-assessed %GP4 was significantly higher in GG 3 compared with GG 2 tumors (p=7.2*10(-11)). %GP4 was associated with an increased risk of BCR (adjusted HR=1.09 per 10% increase in %GP4, p=0.010) in GG 2 and 3 tumors. Within GG 2 tumors specifically, %GP4 was more strongly associated with BCR (adjusted HR=1.12, p=0.006). Our findings demonstrate the feasibility of CNN-assessed %GP4 estimation, which is associated with BCR risk. This objective approach could be added to the standard pathological assessment for patients with GG 2 and 3 tumors and act as a surrogate for specialist genitourinary pathologist evaluation when such consultation is not available.
Medical Subject Headings
Male; Humans; Artificial Intelligence; Reproducibility of Results; Prostatic Neoplasms/pathology; Neoplasm Grading; Prostatectomy; Neural Networks; Computer; Neoplasm Recurrence; Local; Gleason grade; artificial intelligence; digital pathology; percent Gleason pattern 4; prostate cancer
PubMed ID
36925071
ePublication
ePub ahead of print
Volume
36
Issue
7
First Page
100157
