INTEGRATION OF U-NET AND MASK R-CNN APPROACH FOR AUTOMATED CLASSIFICATION OF HISTOPATHOLOGY IMAGES OF PROSTATE CANCER.

Abstract

This study proposes a novel approach for automated classification of histopathology images of prostate cancer by integrating U-Net and Mask R-CNN models. The U-Net model is designed to perform segmentation, localizing regions indicative of prostate cancer, while the Mask R-CNN model is optimized for object detection, enhancing classification precision. Our model implementation leverages modified TensorFlow and Mask R-CNN configurations, utilizing custom dataset generation and preprocessing pipelines to handle labeled prostate histopathology images. The dataset is split into training and testing sets, with the U-Net model trained for segmentation tasks, supported by a data generator class for efficient batch processing. After training, the model’s performance is evaluated using metrics such as accuracy, F1 score, recall, and AUC. The initial results demonstrate promising capabilities in accurately segmenting and classifying cancerous regions in histopathology images, indicating the potential for improving diagnostic accuracy in prostate cancer.