Srivastava, V.Prabhu, A.Sravya, S.Vibha Damodara, K.Lal, S.Kini, J.2026-02-032025International Journal of Imaging Systems and Technology, 2025, 35, 3, pp. -8999457https://doi.org/10.1002/ima.70092https://idr.nitk.ac.in/handle/123456789/20283Prostate cancer is one of the prevalent forms of cancer, posing a significant health concern for men. Accurate detection and classification of prostate cancer are crucial for effective diagnosis and treatment planning. Histopathological images play a pivotal role in identifying prostate cancer by enabling pathologists to identify cellular abnormalities and tumor characteristics. With the rapid advancements in deep learning, Convolutional Neural Networks (CNNs) have emerged as a powerful tool for tackling complex computer vision tasks, including object detection, classification, and segmentation. This paper proposes a Parallel Branch Network (PBN), a CNN architecture specifically designed for the automatic classification of prostate cancer into its subtypes from histopathological images. The paper introduces a novel Efficient Residual (ER) block that enhances feature representation using residual learning and multi-scale feature extraction. By utilizing multiple branches with different filter reduction ratios and dense attention mechanisms, the block captures diverse features while preserving essential information. The proposed PBN model achieved a classification accuracy of 93.16% on the Prostate Gleason dataset, outperforming all other comparison models. © 2025 Wiley Periodicals LLC.Deep neural networksUrologyCellularsConvolutional neural networkDeep learningDiagnosis planningHealth concernsHistopathological imagesMulti-class classificationParallel branchesProstate cancersTreatment planningConvolutional neural networks