Quantitative Analysis of Patch-Based Fully Convolutional Neural Networks for Tissue Segmentation on Brain Magnetic Resonance Imaging

Accurate brain tissue segmentation in magnetic resonance imaging (MRI) has attracted the attention of medical doctors and researchers since variations in tissue volume and shape permit diagnosing and monitoring neurological diseases. Several proposals have been designed throughout the years comprising conventional machine learning strategies as well as convolutional neural networks (CNNs) approaches. In particular, in this paper, we analyze a sub-group of deep learning methods producing dense predictions. This branch, referred in the literature as fully CNN (FCNN), is of interest as these architectures can process an input volume in less time than CNNs. Our study focuses on understanding the architectural strengths and weaknesses of literature-like approaches. We implement eight FCNN architectures inspired by robust state-of-the-art methods on brain segmentation related tasks and use them within a standard pipeline. We evaluate them using the IBSR18, MICCAI2012, and iSeg2017 datasets as they contain infant and adult data and exhibit different voxel spacing, image quality, number of scans, and available imaging modalities. The discussion is driven in four directions: comparison between 2D and 3D approaches, the relevance of multiple imaging sequences, the effect of patch size, and the impact of patch overlap as a sampling strategy for training and testing models. Besides the aforementioned analysis, we show that the methods under evaluation can yield top performance on the three data collections. A public version is accessible to download from our research website to encourage other researchers to explore the evaluation framework.