Abstract:Prostate cancer is the most common cancer in men worldwide and the second leading cause of cancer death in the United States. One of the prognostic features in prostate cancer is the Gleason grading of histopathology images. The Gleason grade is assigned based on tumor architecture on Hematoxylin and Eosin (H&E) stained whole slide images (WSI) by the pathologists. This process is time-consuming and has known interobserver variability. In the past few years, deep learning algorithms have been used to analyze histopathology images, delivering promising results for grading prostate cancer. However, most of the algorithms rely on fully annotated datasets which are expensive to generate. To address this limitation, we propose a novel weakly-supervised algorithm for prostate cancer grade classification. Our algorithm comprises three key steps. Firstly, we employ the Multiple Instance Learning (MIL) algorithm, based on Transformers, to extract discriminative areas within the histopathology images. This approach effectively identifies regions containing crucial information for accurate classification. Secondly, we construct a graph representation of the image by connecting the discriminative patches. This graph captures the intricate relationships between patches, facilitating the analysis of contextual information. Lastly, we develop a Graph Convolutional Neural Network (GCN) using a gated attention mechanism to classify the image into its respective Gleason grade. The performance of our algorithm is extensively evaluated using publicly available datasets, including TCGA-PRAD, PANDA, and Gleason 2019 challenge datasets, alongside rigorous cross-validation on an independent dataset. Our results consistently demonstrate that the proposed model surpasses existing methods in Gleason grading, as evidenced by superior accuracy, F1 score, and Cohen-Kappa scores. By leveraging weakly-supervised learning and graph-based analysis, our algorithm presents a promising solution to improve the efficiency and accuracy of prostate cancer grading, ultimately aiding in patient prognosis and treatment decisions. The code for our algorithm is available at https://github.com/NabaviLab/Prostate-Cancer.

Unsupervised Prostate Cancer Detection on H&E using Convolutional Adversarial Autoencoders

Epithelium segmentation using deep learning in H&E-stained prostate specimens with immunohistochemistry as reference standard

Self-supervised learning of a tailored Convolutional Auto Encoder for histopathological prostate grading

High-Resolution Histopathological Image Classification Model Based on Fused Heterogeneous Networks with Self-Supervised Feature Representation

Detection of prostate cancer in whole-slide images through end-to-end training with image-level labels

Prostate cancer grade using self-supervised learning and novel feature aggregator based on weakly-labeled gbit-pixel pathology images

Deep contrastive learning based tissue clustering for annotation-free histopathology image analysis

Automated Prostate Cancer Diagnosis Based on Gleason Grading Using Convolutional Neural Network

Automatic diagnosis and grading of Prostate Cancer with weakly supervised learning on whole slide images

WeGleNet: A Weakly-Supervised Convolutional Neural Network for the Semantic Segmentation of Gleason Grades in Prostate Histology Images

Diagnosis of Helicobacter pylori using AutoEncoders for the Detection of Anomalous Staining Patterns in Immunohistochemistry Images

Weakly-supervised deep learning model for prostate cancer diagnosis and Gleason grading of histopathology images

Gleason Grading of Histology Prostate Images through Semantic Segmentation via Residual U-Net

Segmenting Potentially Cancerous Areas in Prostate Biopsies using Semi-Automatically Annotated Data

Unsupervised Positron Emission Tomography Tumor Segmentation Via GAN Based Adversarial Auto-Encoder

Whole Slide Image Multi-Classification of Cervical Epithelial Lesions Based on Unsupervised Pre-training

Few-Shot Breast Cancer Metastases Classification via Unsupervised Cell Ranking

3D convolutional network with edge detection for prostate gland and tumor segmentation on T2WI and ADC

Comparison of a machine and deep learning model for automated tumor annotation on digitized whole slide prostate cancer histology

Sparse autoencoder for unsupervised nucleus detection and representation in histopathology images

High Accuracy Tumor Diagnoses and Benchmarking of Hematoxylin and Eosin Stained Prostate Core Biopsy Images Generated by Explainable Deep Neural Networks