Segmentation of MRI brain scans using spatial constraints and 3D features
- PDF / 2,541,277 Bytes
- 12 Pages / 595.224 x 790.955 pts Page_size
- 93 Downloads / 206 Views
ORIGINAL ARTICLE
Segmentation of MRI brain scans using spatial constraints and 3D features Jonas Grande-Barreto1
1 ´ · Pilar Gomez-Gil
Received: 24 September 2019 / Accepted: 8 September 2020 © International Federation for Medical and Biological Engineering 2020
Abstract This paper presents a novel unsupervised algorithm for brain tissue segmentation in magnetic resonance imaging (MRI). The proposed algorithm, named Gardens2, adopts a clustering approach to segment voxels of a given MRI into three classes: cerebrospinal fluid (CSF), gray matter (GM), and white matter (WM). Using an overlapping criterion, 3D feature descriptors and prior atlas information, Gardens2 generates a segmentation mask per class in order to parcellate the brain tissues. We assessed our method using three neuroimaging datasets: BrainWeb, IBSR18, and IBSR20, the last two provided by the Internet Brain Segmentation Repository. Its performance was compared with eleven well established as well as newly proposed unsupervised segmentation methods. Overall, Gardens2 obtained better segmentation performance than the rest of the methods in two of the three databases and competitive results when its performance was measured by class. Keywords Brain MRI · Atlas · Fuzzy functions · Tissue segmentation · Watershed
1 Introduction Image segmentation is a fundamental issue in image processing and computer vision because it provides further image understanding to tackle complex problems. Segmentation refers to a process of partitioning an image into meaningful non-overlapping regions, in such a way that those elements with the same label share similar characteristics. Image segmentation is an ill-posed inverse problem; thus, there is a broad range of methods trying to solve it [17]. In particular, medical image segmentation helps for surgical planning, post-surgical assessment, and abnormality detection, among others. MRI is a non-invasive medical imaging technique that produces a digital representation of biological samples using a monochromatic scale. One of the main applications of medical image segmentation on MRI is to split the image into CSF, GM, and WM tissues, a mandatory action when studying anatomical structure change and brain quantification. Manual segmentation is the most
Jonas Grande-Barreto
[email protected] 1
National Institute for Astrophysics, Optics and Electronics (INAOE), Puebla, Mexico
popular segmentation method, but inter- and intra-observer variabilities limit its reproducibility [24]. Furthermore, the huge amount of data to analyze leads to a tremendously time-consuming and expensive process. The problem of automatic brain tissue segmentation has been mainly addressed using two different approaches: supervised and unsupervised. In the supervised approach, the model must be provided with a subset of labeled voxel samples from each tissue during a learning phase to calculate the parameters of the method [23, 29]. Then, in the classification phase, unlabeled voxels are classified by the trained model [25]. Supervised metho
Data Loading...
