Prostate Gland Segmentation in Histology Images via Residual and Multi-resolution U-NET

Abstract

Prostate cancer is one of the most prevalent cancers worldwide. One of the key factors in reducing its mortality is based on early detection. The computer-aided diagnosis systems for this task are based on the glandular structural analysis in histology images. Hence, accurate gland detection and segmentation is crucial for a successful prediction. The methodological basis of this work is a prostate gland segmentation based on U-Net convolutional neural network architectures modified with residual and multi-resolution blocks, trained using data augmentation techniques. The residual configuration outperforms in the test subset the previous state-of-the-art approaches in an image-level comparison, reaching an average Dice Index of 0.77.

A Pathologist-Informed Workflow for Classification of Prostate Glands in Histopathology

“…Typically, a network is trained to recognize and classify structures of interest in the input image, producing pixel-wise probability maps, one per class, with each pixel assigned to the class with the highest probability. For instance, Silva-Rodríguez et al [21] show that segmenting tumor areas through a neural network achieves better results than traditional algorithms, such as [8] and [2]. The U-net ( [20]) is a special form of convolutional-neural-network architecture designed for image segmentation, and many Gleason grading methods such as [3,19,13,18] rely on variants of the U-net to process patches of a whole slide image (WSI) in order to produce a pixelwise Gleason grade classification.…”

Multi-Stage Classification-Based Deep Learning for Gleason System Grading Using Histopathological Images

“…Researchers produced many scientific works on GSG methods [ 8 , 11 , 19 , 20 , 21 ] that depend on variants of U-net. The U-net [ 22 ] is the standard and the most famous deep network for medical image segmentation from both radiology and histopathological image.…”

“…It is the essential part of the algorithm or the main factor in enhancing the experimental results. For example, Silva-Rodríguez et al [ 19 ] developed a computer-aided diagnosis systems (CAD) system to detect and segment the prostate gland in histology images using multi-resolution and residual and U-net. This model is robust for identifying the benign glands and achieves higher accuracy compared with traditional algorithms such as [ 23 ].…”