Detection and Classification of Prostate Cancer Using Dual-Channel Parallel Convolution Neural Network

Prostate cancer (PCa) treatment strategies and optimal diagnostic are constantly changing. The goal of the pathologists is to analyze the whole slide images (WSIs) and distinguish a malignant from a benign tumor. The pathologists must have long-term experience in the detection of biomarkers of a particular disease. Medical imaging is primarily considered as one of the most significant sources to derive information regarding the biomarkers. Therefore, to assist the pathologists and clinicians, we developed Artificial intelligence (AI) based computer-aided detection (CAD) system. This paper presents an AI technique, the dual-channel parallel convolution neural network (DCPCNN) for fine-grained detection and classification. The histological grades of PCa considered for this study are benign, grade 3, grade 4, and grade 5. Therefore, the proposed AI model aims to perform multiclass classification using multiple patterns of images extracted from WSIs of a prostate biopsy. Moreover, stain normalization was performed to standardize the intensity and contrast level of the image. DCPCNN consists of two CNNs whose convolution layer outputs are concatenated for the final classification. The overall accuracy of multiclass classification achieved by DCPCNN is 98.4% indicates that the proposed model predicted the images almost perfectly accurately. Also, the research findings, limitations of the existing method, and future scope have been described distinctly in this paper.