Mariano Cabezas

A review of atlas-based segmentation for magnetic resonance brain images

Normal and abnormal brains can be segmented by registering the target image with an atlas. Here, an atlas is defined as the combination of an intensity image (template) and its segmented image (the atlas labels). After registering the atlas template and the target image, the atlas labels are propagated to the target image. We define this process as atlas-based segmentation. In recent years, researchers have investigated registration algorithms to match atlases to query subjects and also strategies for atlas construction. In this paper we present a review of the automated approaches for atlas-based segmentation of magnetic resonance brain images. We aim to point out the strengths and weaknesses of atlas-based methods and suggest new research directions. We use two different criteria to present the methods. First, we refer to the algorithms according to their atlas-based strategy: label propagation, multi-atlas methods, and probabilistic techniques. Subsequently, we classify the methods according to their medical target: the brain and its internal structures, tissue segmentation in healthy subjects, tissue segmentation in fetus, neonates and elderly subjects, and segmentation of damaged brains. A quantitative comparison of the results reported in the literature is also presented.

Segmentation of multiple sclerosis lesions in brain MRI: A review of automated approaches

Automatic segmentation of multiple sclerosis (MS) lesions in brain MRI has been widely investigated in recent years with the goal of helping MS diagnosis and patient follow-up. However, the performance of most of the algorithms still falls far below expert expectations. In this paper, we review the main approaches to automated MS lesion segmentation. The main features of the segmentation algorithms are analysed and the most recent important techniques are classified into different strategies according to their main principle, pointing out their strengths and weaknesses and suggesting new research directions. A qualitative and quantitative comparison of the results of the approaches analysed is also presented. Finally, possible future approaches to MS lesion segmentation are discussed.

Comparison of 10 brain tissue segmentation methods using revisited IBSR annotations

Ground‐truth annotations from the well‐known Internet Brain Segmentation Repository (IBSR) datasets consider Sulcal cerebrospinal fluid (SCSF) voxels as gray matter. This can lead to bias when evaluating the performance of tissue segmentation methods. In this work we compare the accuracy of 10 brain tissue segmentation methods analyzing the effects of SCSF ground‐truth voxels on accuracy estimations.

A toolbox for multiple sclerosis lesion segmentation

IntroductionLesion segmentation plays an important role in the diagnosis and follow-up of multiple sclerosis (MS). This task is very time-consuming and subject to intra- and inter-rater variability. In this paper, we present a new tool for automated MS lesion segmentation using T1w and fluid-attenuated inversion recovery (FLAIR) images.MethodsOur approach is based on two main steps, initial brain tissue segmentation according to the gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) performed in T1w images, followed by a second step where the lesions are segmented as outliers to the normal apparent GM brain tissue on the FLAIR image.ResultsThe tool has been validated using data from more than 100 MS patients acquired with different scanners and at different magnetic field strengths. Quantitative evaluation provided a better performance in terms of precision while maintaining similar results on sensitivity and Dice similarity measures compared with those of other approaches.ConclusionOur tool is implemented as a publicly available SPM8/12 extension that can be used by both the medical and research communities.

Improving automated multiple sclerosis lesion segmentation with a cascaded 3D convolutional neural network approach

ABSTRACT In this paper, we present a novel automated method for White Matter (WM) lesion segmentation of Multiple Sclerosis (MS) patient images. Our approach is based on a cascade of two 3D patch‐wise convolutional neural networks (CNN). The first network is trained to be more sensitive revealing possible candidate lesion voxels while the second network is trained to reduce the number of misclassified voxels coming from the first network. This cascaded CNN architecture tends to learn well from a small (Symbol) set of labeled data of the same MRI contrast, which can be very interesting in practice, given the difficulty to obtain manual label annotations and the large amount of available unlabeled Magnetic Resonance Imaging (MRI) data. We evaluate the accuracy of the proposed method on the public MS lesion segmentation challenge MICCAI2008 dataset, comparing it with respect to other state‐of‐the‐art MS lesion segmentation tools. Furthermore, the proposed method is also evaluated on two private MS clinical datasets, where the performance of our method is also compared with different recent public available state‐of‐the‐art MS lesion segmentation methods. At the time of writing this paper, our method is the best ranked approach on the MICCAI2008 challenge, outperforming the rest of 60 participant methods when using all the available input modalities (T1‐w, T2‐w and FLAIR), while still in the top‐rank (3rd position) when using only T1‐w and FLAIR modalities. On clinical MS data, our approach exhibits a significant increase in the accuracy segmenting of WM lesions when compared with the rest of evaluated methods, highly correlating (Symbol) also with the expected lesion volume. Symbol. No caption available. Symbol. No caption available. HIGHLIGHTSWe propose an automated WM lesion segmentation method for MS patient images.The approach relies on a cascade of two 7‐layer convolutional neural networks.We evaluate its accuracy with both the MICCAI2008 challenge and clinical MS data.Our approach is currently the best ranked method of the challenge (1th pos / 60).On MS data, the accuracy is significantly better that state‐of‐the‐art methods.

MARGA: Multispectral Adaptive Region Growing Algorithm for brain extraction on axial MRI

Brain extraction, also known as skull stripping, is one of the most important preprocessing steps for many automatic brain image analysis. In this paper we present a new approach called Multispectral Adaptive Region Growing Algorithm (MARGA) to perform the skull stripping process. MARGA is based on a region growing (RG) algorithm which uses the complementary information provided by conventional magnetic resonance images (MRI) such as T1-weighted and T2-weighted to perform the brain segmentation. MARGA can be seen as an extension of the skull stripping method proposed by Park and Lee (2009) [1], enabling their use in both axial views and low quality images. Following the same idea, we first obtain seed regions that are then spread using a 2D RG algorithm which behaves differently in specific zones of the brain. This adaptation allows to deal with the fact that middle MRI slices have better image contrast between the brain and non-brain regions than superior and inferior brain slices where the contrast is smaller. MARGA is validated using three different databases: 10 simulated brains from the BrainWeb database; 2 data sets from the National Alliance for Medical Image Computing (NAMIC) database, the first one consisting in 10 normal brains and 10 brains of schizophrenic patients acquired with a 3T GE scanner, and the second one consisting in 5 brains from lupus patients acquired with a 3T Siemens scanner; and 10 brains of multiple sclerosis patients acquired with a 1.5T scanner. We have qualitatively and quantitatively compared MARGA with the well-known Brain Extraction Tool (BET), Brain Surface Extractor (BSE) and Statistical Parametric Mapping (SPM) approaches. The obtained results demonstrate the validity of MARGA, outperforming the results of those standard techniques.

Automatic multiple sclerosis lesion detection in brain MRI by FLAIR thresholding

Magnetic resonance imaging (MRI) is frequently used to detect and segment multiple sclerosis lesions due to the detailed and rich information provided. We present a modified expectation-maximisation algorithm to segment brain tissues (white matter, grey matter, and cerebro-spinal fluid) as well as a partial volume class containing fluid and grey matter. This algorithm provides an initial segmentation in which lesions are not separated from tissue, thus a second step is needed to find them. This second step involves the thresholding of the FLAIR image, followed by a regionwise refinement to discard false detections. To evaluate the proposal, we used a database with 45 cases comprising 1.5T imaging data from three different hospitals with different scanner machines and with a variable lesion load per case. The results for our database point out to a higher accuracy when compared to two of the best state-of-the-art approaches.

Intensity Based Methods for Brain MRI Longitudinal Registration. A Study on Multiple Sclerosis Patients

Registration is a key step in many automatic brain Magnetic Resonance Imaging (MRI) applications. In this work we focus on longitudinal registration of brain MRI for Multiple Sclerosis (MS) patients. First of all, we analyze the effect that MS lesions have on registration by synthetically eliminating some of the lesions. Our results show how a widely used method for longitudinal registration such as rigid registration is practically unconcerned by the presence of MS lesions while several non-rigid registration methods produce outputs that are significantly different. We then focus on assessing which is the best registration method for longitudinal MRI images of MS patients. In order to analyze the results obtained for all studied criteria, we use both descriptive statistics and statistical inference: one way ANOVA, pairwise t-tests and permutation tests.

BOOST: A supervised approach for multiple sclerosis lesion segmentation

BACKGROUND Automatic multiple sclerosis lesion segmentation is a challenging task. An extensive analysis of the most recent techniques indicates an improvement of the results obtained when using prior knowledge and contextual information. NEW METHOD We present BOOST, a knowledge-based approach to automatically segment multiple sclerosis lesions through a voxel by voxel classification. We used the Gentleboost classifier and a set of features, including contextual features, registered atlas probability maps and an outlier map. RESULTS Results are computed on a set of 45 cases from three different hospitals (15 of each), obtaining a moderate agreement between the manual annotations and the automatically segmented results. COMPARISON WITH EXISTING METHOD(S) We quantitatively compared our results with three public state-of-the-art approaches obtaining competitive results and a better overlap with manual annotations. Our approach tends to better segment those cases with high lesion load, while cases with small lesion load are more difficult to accurately segment. CONCLUSIONS We believe BOOST has potential applicability in the clinical practice, although it should be improved in those cases with small lesion load.

Improved Automatic Detection of New T2 Lesions in Multiple Sclerosis Using Deformation Fields

BACKGROUND AND PURPOSE: Detection of disease activity, defined as new/enlarging T2 lesions on brain MR imaging, has been proposed as a biomarker in MS. However, detection of new/enlarging T2 lesions can be hindered by several factors that can be overcome with image subtraction. The purpose of this study was to improve automated detection of new T2 lesions and reduce user interaction to eliminate inter- and intraobserver variability. MATERIALS AND METHODS: Multiparametric brain MR imaging was performed at 2 time points in 36 patients with new T2 lesions. Images were registered by using an affine transformation and the Demons algorithm to obtain a deformation field. After affine registration, images were subtracted and a threshold was applied to obtain a lesion mask, which was then refined by using the deformation field, intensity, and local information. This pipeline was compared with only applying a threshold, and with a state-of-the-art approach relying only on image intensities. To assess improvements, we compared the results of the different pipelines with the expert visual detection. RESULTS: The multichannel pipeline based on the deformation field obtained a detection Dice similarity coefficient close to 0.70, with a false-positive detection of 17.8% and a true-positive detection of 70.9%. A statistically significant correlation (r = 0.81, P value = 2.2688e-09) was found between visual detection and automated detection by using our approach. CONCLUSIONS: The deformation field–based approach proposed in this study for detecting new/enlarging T2 lesions resulted in significantly fewer false-positives while maintaining most true-positives and showed a good correlation with visual detection annotations. This approach could reduce user interaction and inter- and intraobserver variability.