Alfiia Galimzianova

Deep Learning for Brain MRI Segmentation: State of the Art and Future Directions

Quantitative analysis of brain MRI is routine for many neurological diseases and conditions and relies on accurate segmentation of structures of interest. Deep learning-based segmentation approaches for brain MRI are gaining interest due to their self-learning and generalization ability over large amounts of data. As the deep learning architectures are becoming more mature, they gradually outperform previous state-of-the-art classical machine learning algorithms. This review aims to provide an overview of current deep learning-based segmentation approaches for quantitative brain MRI. First we review the current deep learning architectures used for segmentation of anatomical brain structures and brain lesions. Next, the performance, speed, and properties of deep learning approaches are summarized and discussed. Finally, we provide a critical assessment of the current state and identify likely future developments and trends.

Robust Estimation of Unbalanced Mixture Models on Samples with Outliers

Mixture models are often used to compactly represent samples from heterogeneous sources. However, in real world, the samples generally contain an unknown fraction of outliers and the sources generate different or unbalanced numbers of observations. Such unbalanced and contaminated samples may, for instance, be obtained by high density data sensors such as imaging devices. Estimation of unbalanced mixture models from samples with outliers requires robust estimation methods. In this paper, we propose a novel robust mixture estimator incorporating trimming of the outliers based on component-wise confidence level ordering of observations. The proposed method is validated and compared to the state-of-the-art FAST-TLE method on two data sets, one consisting of synthetic samples with a varying fraction of outliers and a varying balance between mixture weights, while the other data set contained structural magnetic resonance images of the brain with tumors of varying volumes. The results on both data sets clearly indicate that the proposed method is capable to robustly estimate unbalanced mixtures over a broad range of outlier fractions. As such, it is applicable to real-world samples, in which the outlier fraction cannot be estimated in advance.

Stratified mixture modeling for segmentation of white-matter lesions in brain MR images

Accurate characterization of white-matter lesions from magnetic resonance (MR) images has increasing importance for diagnosis and management of treatment of certain neurological diseases, and can be performed in an objective and effective way by automated lesion segmentation. This usually involves modeling the whole-brain MR intensity distribution, however, capturing various sources of MR intensity variability and lesion heterogeneity results in highly complex whole-brain MR intensity models, thus their robust estimation on a large set of MR images presents a huge challenge. We propose a novel approach employing stratified mixture modeling, where the main premise is that the otherwise complex whole-brain model can be reduced to a tractable parametric form in small brain subregions. We show on MR images of multiple sclerosis (MS) patients with different lesion loads that robust estimators enable accurate mixture modeling of MR intensity in small brain subregions even in the presence of lesions. Recombination of the mixture models across strata provided an accurate whole-brain MR intensity model. Increasing the number of subregions and, thereby, the model complexity, consistently improved the accuracy of whole-brain MR intensity modeling and segmentation of normal structures. The proposed approach was incorporated into three unsupervised lesion segmentation methods and, compared to original and three other state-of-the-art methods, the proposed modeling approach significantly improved lesion segmentation according to increased Dice similarity indices and lower number of false positives on real MR images of 30 patients with MS.

Combining Unsupervised and Supervised Methods for Lesion Segmentation

White-matter lesions are associated to several diseases, which can be characterized by neuroimaging biomarkers through lesion segmentation in MR images. We present a novel automated lesion segmentation method consisting of an unsupervised mixture model based extraction of candidate lesion voxels, which are subsequently classified by a random decision forest (RDF) using simple visual features like multi-sequence MR intensities sourced from connected voxel neighborhoods. The candidate lesion extraction prior to RDF training and classification balanced the number of non-lesion and lesion voxels and the number of non-lesion classes versus a lesion class. Thereby, the RDF established highly discriminating decision rules based on such simple visual features, which have the benefit of no computational overhead and easy extraction from the MR images. On MR images of 18 patients with multiple sclerosis the proposed method achieved the median Dice similarity of 0.73, sensitivity of 0.90 and positive predictive value of 0.61, which indicate accurate segmentation of white-matter lesions.

Automated segmentation of MS lesions in brain MR images using localized trimmed-likelihood estimation

Diagnosis and prognosis of patients with multiple sclerosis (MS) rely on quantitative markers derived from the analysis of magnetic resonance (MR) images. To compute these markers, a segmentation of lesions in the brain tissues, which are characteristic for MS disease, is needed. In this paper, we propose an unsupervised method for segmenting MS lesions that employs localized trimmed-likelihood estimation (TLE) to model the intensity distributions of normal appearing brain tissues (NABT). Compared to the original whole-brain TLE approach, the proposed method employs a set of three-component Gaussian mixture models for each of the spatially localized and non-overlapping subregions of the brain. The subregions were assigned by the customized balanced box decomposition that takes into account the spatial distribution and the cardinality of NABT tissues, as obtained from the initial whole-brain TLE. The proposed method was tested and compared to the original TLE approach on publicly available synthetic BrainWeb datasets. The results indicate a higher average Dice similarity coefficient both for the segmentation of NABT and MS lesions by using the proposed spatially localized TLE as compared to the original whole-brain TLE, which is due to the fact that the proposed method yields a more accurate NABT model and thus detects fewer false NABT outliers.

Robust Mixture-Parameter Estimation for Unsupervised Segmentation of Brain MR Images

Methods for automated segmentation of brain MR images are routinely used in large-scale neurological studies. Automated segmentation is usually performed by unsupervised methods, since these can be used even if different MR sequences or different pathologies are studied. The unsupervised methods model intensity distribution of major brain structures using mixture models, the parameters of which need to be robustly estimated from MR data and in presence of outliers. In this paper, we propose a robust mixture-parameter estimation that detects outliers as samples with low significance level of the corresponding mixture component and iteratively re-estimates the fraction of outliers. Results on synthetic and real brain image datasets demonstrate superior robustness of the proposed method as compared to the popular FAST-TLE method over a broad range of trimming fraction values. The latter is important for segmenting brain structures with pathology, the extent of which is hard to predict in large-scale imaging studies.

A Novel Public MR Image Dataset of Multiple Sclerosis Patients With Lesion Segmentations Based on Multi-rater Consensus

Quantified volume and count of white-matter lesions based on magnetic resonance (MR) images are important biomarkers in several neurodegenerative diseases. For a routine extraction of these biomarkers an accurate and reliable automated lesion segmentation is required. To objectively and reliably determine a standard automated method, however, creation of standard validation datasets is of extremely high importance. Ideally, these datasets should be publicly available in conjunction with standardized evaluation methodology to enable objective validation of novel and existing methods. For validation purposes, we present a novel MR dataset of 30 multiple sclerosis patients and a novel protocol for creating reference white-matter lesion segmentations based on multi-rater consensus. On these datasets three expert raters individually segmented white-matter lesions, using in-house developed semi-automated lesion contouring tools. Later, the raters revised the segmentations in several joint sessions to reach a consensus on segmentation of lesions. To evaluate the variability, and as quality assurance, the protocol was executed twice on the same MR images, with a six months break. The obtained intra-consensus variability was substantially lower compared to the intra- and inter-rater variabilities, showing improved reliability of lesion segmentation by the proposed protocol. Hence, the obtained reference segmentations may represent a more precise target to evaluate, compare against and also train, the automatic segmentations. To encourage further use and research we will publicly disseminate on our website http://lit.fe.uni-lj.si/tools the tools used to create lesion segmentations, the original and preprocessed MR image datasets and the consensus lesion segmentations.

Locally adaptive MR intensity models and MRF-based segmentation of multiple sclerosis lesions

Neuroimaging biomarkers are an important paraclinical tool used to characterize a number of neurological diseases, however, their extraction requires accurate and reliable segmentation of normal and pathological brain structures. For MR images of healthy brains the intensity models of normal-appearing brain tissue (NABT) in combination with Markov random field (MRF) models are known to give reliable and smooth NABT segmentation. However, the presence of pathology, MR intensity bias and natural tissue-dependent intensity variability altogether represent difficult challenges for a reliable estimation of NABT intensity model based on MR images. In this paper, we propose a novel method for segmentation of normal and pathological structures in brain MR images of multiple sclerosis (MS) patients that is based on locally-adaptive NABT model, a robust method for the estimation of model parameters and a MRF-based segmentation framework. Experiments on multi-sequence brain MR images of 27 MS patients show that, compared to whole-brain model and compared to the widely used Expectation-Maximization Segmentation (EMS) method, the locally-adaptive NABT model increases the accuracy of MS lesion segmentation.

Locally adaptive magnetic resonance intensity models for unsupervised segmentation of multiple sclerosis lesions

Abstract. Multiple sclerosis (MS) is a neurological disease characterized by focal lesions and morphological changes in the brain captured on magnetic resonance (MR) images. However, extraction of the corresponding imaging markers requires accurate segmentation of normal-appearing brain structures (NABS) and the lesions in MR images. On MR images of healthy brains, the NABS can be accurately captured by MR intensity mixture models, which, in combination with regularization techniques, such as in Markov random field (MRF) models, are known to give reliable NABS segmentation. However, on MR images that also contain abnormalities such as MS lesions, obtaining an accurate and reliable estimate of NABS intensity models is a challenge. We propose a method for automated segmentation of normal-appearing and abnormal structures in brain MR images that is based on a locally adaptive NABS model, a robust model parameters estimation method, and an MRF-based segmentation framework. Experiments on multisequence brain MR images of 30 MS patients show that, compared to whole-brain MR intensity model and compared to four popular unsupervised lesion segmentation methods, the proposed method increases the accuracy of MS lesion segmentation.