Probabilistic segmentation of white matter lesions in MR imaging

A new method has been developed for fully automated segmentation of white matter lesions (WMLs) in cranial MR imaging. The algorithm uses information from T1-weighted (T1-w), inversion recovery (IR), proton density-weighted (PD), T2-weighted (T2-w) and fluid attenuation inversion recovery (FLAIR) scans. It is based on the K-Nearest Neighbor (KNN) classification technique that builds a feature space from voxel intensities and spatial information. The technique generates images representing the probability per voxel being part of a WML. By application of thresholds on these probability maps, binary segmentations can be obtained. ROC curves show that the segmentations achieve both high sensitivity and specificity. A similarity index (SI), overlap fraction (OF) and extra fraction (EF) are calculated for additional quantitative analysis of the result. The SI is also used for determination of the optimal probability threshold for generation of the binary segmentation. Using probabilistic equivalents of the SI, OF and EF, the probability maps can be evaluated directly, providing a powerful tool for comparison of different classification results. This method for automated WML segmentation reaches an accuracy that is comparable to methods for multiple sclerosis (MS) lesion segmentation and is suitable for detection of WMLs in large and longitudinal population studies.     

Segmentation of age-related white matter changes in a clinical multi-center study

Age-related white matter changes (WMC) are thought to be a marker of vascular pathology, and have been associated with motor and cognitive deficits. In the present study, an optimized artificial neural network was used as an automatic segmentation method to produce probabilistic maps of WMC in a clinical multi-center study. The neural network uses information from T1- and T2-weighted and fluid attenuation inversion recovery (FLAIR) magnetic resonance (MR) scans, neighboring voxels and spatial location. Generalizability of the neural network was optimized by including the Optimal Brain Damage (OBD) pruning method in the training stage. Six optimized neural networks were produced to investigate the impact of different input information on WMC segmentation. The automatic segmentation method was applied to MR scans of 362 non-demented elderly subjects from 11 centers in the European multi-center study Leukoaraiosis And Disability (LADIS). Semi-manually delineated WMC were used for validating the segmentation produced by the neural networks. The neural network segmentation demonstrated high consistency between subjects and centers, making it a promising technique for large studies. For WMC volumes less than 10 ml, an increasing discrepancy between semi-manual and neural network segmentation was observed using the similarity index (SI) measure. The use of all three image modalities significantly improved cross-center generalizability compared to neural networks using the FLAIR image only. Expert knowledge not available to the neural networks was a minor source of discrepancy, while variation in MR scan quality constituted the largest source of error.     

Multi-stage segmentation of white matter hyperintensity, cortical and lacunar infarcts

Cerebral abnormalities such as white matter hyperintensity (WMH), cortical infarct (CI), and lacunar infarct (LI) are of clinical importance and frequently present in patients with stroke and dementia. Up to date, there are limited algorithms available to automatically delineate these cerebral abnormalities partially due to their complex appearance in MR images. In this paper, we describe an automated multi-stage segmentation approach for labeling the WMH, CI, and LI using multi-modal MR images. We first automatically segment brain tissues (white matter, gray matter, and CSF) based on the T1-weighted image and then identify hyperintense voxels based on the fluid attenuated inversion recovery (FLAIR) image. We finally label the WMH, CI, and LI based on the T1-weighted, T2-weighted, and FLAIR images. The segmentation accuracy is evaluated using a community-based sample of 272 old adults. Our results show that the automated segmentation of the WMH, CI, and LI is comparable with manual labeling in terms of spatial location, volume, and the number of lacunes. Additionally, the WMH volume is highly correlated with the visual grading score based on the Age-Related White Matter Changes (ARWMC) protocol. The evaluations against the manual labeling and ARWMC visual grading suggest that our algorithm provides reasonable segmentation accuracy for the WMH, CI, and LI.     

Automatic segmentation of different-sized white matter lesions by voxel probability estimation

A new method for fully automated segmentation of white matter lesions (WMLs) on cranial MR imaging is presented. The algorithm uses five types of regular MRI-scans. It is based on a K-Nearest Neighbor (KNN) classification technique, which builds a feature space from voxel intensity features and spatial information. The technique generates images representing the probability per voxel being part of a WML. By application of thresholds on these probability maps binary segmentations can be produced. ROC-curves show that the segmentations achieve a high sensitivity and specificity. Three similarity measures, the similarity index (SI), the overlap fraction (OF) and the extra fraction (EF), are calculated for evaluation of the results and determination of the optimal threshold on the probability map. Investigation of the relation between the total lesion volume and the similarity measures shows that the method performs well for lesions larger than 2 cc. The maximum SI per patient is correlated to the total WML volume. No significant relation between the lesion volume and the optimal threshold has been observed. The probabilistic equivalents of the SI, OF en EF (PSI, POF and PEF) allow direct evaluation of the probability maps, which provides a strong tool for comparison of different classification results. A significant correlation between the lesion volume and the PSI and the PEF has been noticed. This method for automated WML segmentation is applicable to lesions of different sizes and shapes, and reaches an accuracy that is comparable to existing methods for multiple sclerosis lesion segmentation. Furthermore, it is suitable for detection of WMLs in large and longitudinal population studies.     

Segmentation of MRI brain scans using non-uniform partial volume densities

We present an algorithm that provides a partial volume segmentation of a T1-weighted image of the brain into gray matter, white matter and cerebrospinal fluid. The algorithm incorporates a non-uniform partial volume density that takes the curved nature of the cortex into account. The pure gray and white matter intensities are estimated from the image, using scanner noise and cortical partial volume effects. Expected tissue fractions are subsequently computed in each voxel. The algorithm has been tested for reliability, correct estimation of the pure tissue intensities on both real (repeated) MRI data and on simulated (brain) images. Intra-class correlation coefficients (ICCs) were above 0.93 for all volumes of the three tissue types for repeated scans from the same scanner, as well as for scans with different voxel sizes from different scanners with different field strengths. The implementation of our non-uniform partial volume density provided more reliable volumes and tissue fractions, compared to a uniform partial volume density. Applying the algorithm to simulated images showed that the pure tissue intensities were estimated accurately. Variations in cortical thickness did not influence the accuracy of the volume estimates, which is a valuable property when studying (possible) group differences. In conclusion, we have presented a new partial volume segmentation algorithm that allows for comparisons over scanners and voxel sizes.     

Fully automatic segmentation of white matter hyperintensities in MR images of the elderly

The role of quantitative image analysis in large clinical trials is continuously increasing. Several methods are available for performing white matter hyperintensity (WMH) volume quantification. They vary in the amount of the human interaction involved. In this paper, we describe a fully automatic segmentation that was used to quantify WMHs in a large clinical trial on elderly subjects. Our segmentation method combines information from 3 different MR images: proton density (PD), T2-weighted and fluid-attenuated inversion recovery (FLAIR) images; our method uses an established artificial intelligent technique (fuzzy inference system) and does not require extensive computations. The reproducibility of the segmentation was evaluated in 9 patients who underwent scan-rescan with repositioning; an inter-class correlation coefficient (ICC) of 0.91 was obtained. The effect of differences in image resolution was tested in 44 patients, scanned with 6- and 3-mm slice thickness FLAIR images; we obtained an ICC value of 0.99. The accuracy of the segmentation was evaluated on 100 patients for whom manual delineation of WMHs was available; the obtained ICC was 0.98 and the similarity index was 0.75. Besides the fact that the approach demonstrated very high volumetric and spatial agreement with expert delineation, the software did not require more than 2 min per patient (from loading the images to saving the results) on a Pentium-4 processor (512 MB RAM).     

液体衰减翻转恢复序列（FLAIR）在头颅MR检查中的价值

目的 探讨液体衰减翻转序列(FLAIR序列)在头颅MR检查中的应用价值。方法 对106例病人进行了轴位FSE T2和FLAIR序列的检查,25例做了轴位SE T1增强序列。并分别从不同病种和不同病变部位进行了分析。结果 106例病人中,86例有阳性发现,共显示117个病灶。60例脑梗塞病人,FSE T2和FLAIR序列对病变的敏感性相同,但在FLAIR序列能区分梗塞是否为陈旧性。3例转移瘤病人,FLAIR和SE T1增强序列均比FSE T2序列的敏感性高。3例脑脓肿病人和3例脑膜炎病人,FLAIR和SE T1增强序列能显示FSE T1序列不能显示的累及脑室和脑膜的炎症,且FLAIR序列比SE T1增强序列显示的范围大。13例原发性脑肿瘤病人和2例动静脉畸形病人,FSE T2、FLAIR和SE T1增强序列对病变的敏感性相同。1例癫痫病人,FSE T2、SE T1增强序列均为阴性,而FLAIR序列能显示右颞叶近脑沟的高信号病变。结论 在头颅MR检查中,FLAIR序列对邻近脑室、脑池和脑沟的病变较FSE T2更为敏感。     

An automated tool for detection of FLAIR-hyperintense white-matter lesions in Multiple Sclerosis

In Multiple Sclerosis (MS), detection of T2-hyperintense white matter (WM) lesions on magnetic resonance imaging (MRI) has become a crucial criterion for diagnosis and predicting prognosis in early disease. Automated lesion detection is not only desirable with regard to time and cost effectiveness but also constitutes a prerequisite to minimize user bias. Here, we developed and evaluated an algorithm for automated lesion detection requiring a three-dimensional (3D) gradient echo (GRE) T1-weighted and a FLAIR image at 3 Tesla (T). Our tool determines the three tissue classes of gray matter (GM) and WM as well as cerebrospinal fluid (CSF) from the T1-weighted image, and, then, the FLAIR intensity distribution of each tissue class in order to detect outliers, which are interpreted as lesion beliefs. Next, a conservative lesion belief is expanded toward a liberal lesion belief. To this end, neighboring voxels are analyzed and assigned to lesions under certain conditions. This is done iteratively until no further voxels are assigned to lesions. Herein, the likelihood of belonging to WM or GM is weighed against the likelihood of belonging to lesions. We evaluated our algorithm in 53 MS patients with different lesion volumes, in 10 patients with posterior fossa lesions, and 18 control subjects that were all scanned at the same 3T scanner (Achieva, Philips, Netherlands). We found good agreement with lesions determined by manual tracing (R2 values of over 0.93 independent of FLAIR slice thickness up to 6 mm). These results require validation with data from other protocols based on a conventional FLAIR sequence and a 3D GRE T1-weighted sequence. Yet, we believe that our tool allows fast and reliable segmentation of FLAIR-hyperintense lesions, which might simplify the quantification of lesions in basic research and even clinical trials.     

Brain lesions in septic shock: a magnetic resonance imaging study

Background Understanding of sepsis-induced brain dysfunction remains poor, and relies mainly on data from animals or post-mortem studies in patients. The current study provided findings from magnetic resonance imaging of the brain in septic shock. Methods Nine patients with septic shock and brain dysfunction [7 women, median age 63 years (interquartile range 61–79 years), SAPS II: 48 (44–56), SOFA: 8 (6–10)] underwent brain magnetic resonance imaging including gradient echo T1-weighted, fluid-attenuated inversion recovery (FLAIR), T2-weighted and diffusion isotropic images, and mapping of apparent diffusion coefficient. Results Brain imaging was normal in two patients, showed multiple ischaemic strokes in two patients, and in the remaining patients showed white matter lesions at the level of the centrum semiovale, predominating around Virchow–Robin spaces, ranging from small multiple areas to diffuse lesions, and characterised by hyperintensity on FLAIR images. The main lesions were also characterised by reduced signal on diffusion isotropic images and increased apparent diffusion coefficient. The lesions of the white matter worsened with increasing duration of shock and were correlated with Glasgow Outcome Score. Conclusion This preliminary study showed that sepsis-induced brain lesions can be documented by magnetic resonance imaging. These lesions predominated in the white matter, suggesting increasedblood–brain barrier permeability, and were associated with poor outcome.     

Detailed semiautomated MRI based morphometry of the neonatal brain: preliminary results

In the neonate, regional growth trajectories provide information about the coordinated development of cerebral substructures and help identify regional vulnerability by identifying times of faster growth. Segmentation of magnetic resonance images (MRI) has provided detailed information for the myelinated brain but few reports of regional neonatal brain growth exist. We report the method and preliminary results of detailed semiautomated segmentation of 12 normative neonatal brains (gestational age 31.1-42.6 weeks at time of MRI) using volumetric T1-weighted images. Accuracy was confirmed by expert review of every segmented image. In 5 brains, repeat segmentation resulted in intraclass correlation coefficients >0.9 (except for the right amygdala) and an average percent voxel overlap of 90.0%. Artifacts or image quality limited the number of regions segmented in 9/12 data sets and 1/12 was excluded from volumetric analysis due to ventriculomegaly. Brains were segmented into cerebral exterior (N = 8), cerebral lobes (N = 5), lateral ventricles (N = 8), cerebral cortex (N = 6), white matter (N = 6), corpus callosum (N = 7), deep central gray (N = 8), hippocampi (N = 8), amygdalae (N = 8), cerebellar hemispheres (N = 8), vermis (N = 8), midbrain (N = 8), pons (N = 8) and medulla (N = 8). Linear growth (P < 0.05) was identified in all regions except the cerebral white matter, medulla and ventricles. Striking differences in regional growth rates were noted. These preliminary results are consistent with the heterochronous nature of cerebral development and provide initial estimates of regional brain growth and therefore regional vulnerability in the perinatal time period.     

Gray matter reduction is correlated with white matter hyperintensity volume: a voxel-based morphometric study in a large epidemiological sample

Both brain atrophy and T2-weighted white matter hyperintensities (WMH) are common findings in the brains of asymptomatic elderly individuals as well as in disease-specific brains. The study of the relationship between these two salient features is therefore important. To investigate such a relationship, we performed a brain magnetic resonance imaging (MRI) study on 397 asymptomatic individuals aged between 60 and 64 years, who were recruited randomly from a large community sample. WMH were delineated on T2-weighted fluid attenuation inversion recovery (FLAIR) whole brain scans using an automated procedure. The results showed that gray matter reduction, subarachnoid CSF (SA-CSF) increase and lateral ventricular dilation were significantly correlated with WMH load. Deep white matter hyperintensity (DWMH) had significant correlation with all three global atrophy indices, but periventricular white matter hyperintensity (PVWMH) was correlated only with gray matter volume. Voxel-based morphometric (VBM) analysis showed that regional gray matter reduction correlated more closely with WMH load of the proximate region than with WMH elsewhere. The results suggest that WMH have a relationship with brain atrophy in middle age, although the study cannot determine which process, i.e. the development of WMH or atrophy, is primary. The study also demonstrates that DWMH has a more significant relationship with structural brain changes, and may therefore be more functionally relevant than PVWMH. Further delineation of this relationship needs a longitudinal study of the changes in both WMH and indices of brain atrophy.     

The topography of white matter hyperintensities on brain MRI in healthy 60- to 64-year-old individuals

We report the topography of brain white matter hyperintensities (WMHs) on T2-weighted fluid attenuated inversion recovery (FLAIR) magnetic resonance imaging in 477 healthy subjects aged 60-64 years selected randomly from the community. WMHs were delineated by using a computer algorithm. We found that all subjects had periventricular WMHs and 96.6% subjects also had deep WMHs. The mean volume of WMHs was 4.9 ml, comprising 0.83% of the white matter, of which 1.2 ml was severe in intensity. The deep WMHs were distributed throughout the cerebral hemispheres, with the occipital and frontal white matter bearing the greatest burden. The territory of the lenticulostriate arteries had the greatest WMHs. A white matter region of 4 mm adjacent to the cortex was not affected by hyperintensities. The mean (SD) number of discrete WMHs was 19.6 (7.1) per subject, of which 6.1 (4.4) were severe in intensity. Nearly half (48.6%) of the subjects had at least one large WMH (>12 mm diameter) and one eighth (12.5%) of the subjects had at least one large WMH that appeared to be severe in MRI. The overall load of WMHs was similar in men and women, but the latter had a higher proportion of their white matter so affected. This study provides the first detailed topographic analysis of WMHs in a large representative middle-aged sample, emphasizes their high prevalence in mid-adult life and raises issues about their etiology and significance.     

Automatic segmentation of MR images of the developing newborn brain

This paper describes an automatic tissue segmentation method for newborn brains from magnetic resonance images (MRI). The analysis and study of newborn brain MRI is of great interest due to its potential for studying early growth patterns and morphological changes in neurodevelopmental disorders. Automatic segmentation of newborn MRI is a challenging task mainly due to the low intensity contrast and the growth process of the white matter tissue. Newborn white matter tissue undergoes a rapid myelination process, where the nerves are covered in myelin sheathes. It is necessary to identify the white matter tissue as myelinated or non-myelinated regions. The degree of myelination is a fractional voxel property that represents regional changes of white matter as a function of age. Our method makes use of a registered probabilistic brain atlas. The method first uses robust graph clustering and parameter estimation to find the initial intensity distributions. The distribution estimates are then used together with the spatial priors to perform bias correction. Finally, the method refines the segmentation using training sample pruning and non-parametric kernel density estimation. Our results demonstrate that the method is able to segment the brain tissue and identify myelinated and non-myelinated white matter regions.     

Quantitative analysis of group-specific brain tissue probability map for schizophrenic patients

We developed group-specific tissue probability map (TPM) for gray matter (GM), white matter (WM) and cerebrospinal fluid (CSF) on the common spatial coordinates of an averaged brain atlas derived from normal controls (NC) and from schizophrenic patients (SZ). To identify differences in group-specific TPMs, we used quantitative evaluation methods based on differences in probabilistic distribution as a global criterion, and the mean probability and the similarity index (SI) by lobe as regional criteria. The SZ group showed more spatial variation with a lower mean probability than NC subjects. And, for the right temporal and left parietal lobes, the SI between each group was lower than the other lobes. It can be said that there were significant differences in spatial distribution between controls and schizophrenic patients at those areas. In case of female group, although group differences in the volumes of GM and WM were not significant, global difference in the probabilistic distribution of GM was more prominent and the SI was lower and its descent rate was greater in all lobes, compared with the male group. If these morphological differences caused by disease or group-specific features were not considered in TPM, the accuracy and certainty of specific group studies would be greatly reduced. Therefore, suitable TPM is required as a common framework for functional neuroimaging studies and an a priori knowledge of tissue classification.     

Bayesian construction of geometrically based cortical thickness metrics

This paper describes the construction of cortical metrics quantifying the probabilistic occurrence of gray matter, white matter, and cerebrospinal fluid compartments in their correlation to the geometry of the neocortex as measured in 0.5-1.0 mm magnetic resonance imagery. These cortical profiles represent the density of the tissue types as a function of distance to the cortical surface. These metrics are consistent when generated across multiple brains indicating a fundamental property of the neocortex. Methods are proposed for incorporating such metrics into automated Bayes segmentation.     

Evaluation of automatic neonatal brain segmentation algorithms: The NeoBrainS12 challenge

A number of algorithms for brain segmentation in preterm born infants have been published, but a reliable comparison of their performance is lacking. The NeoBrainS12 study (http://neobrains12.isi.uu.nl), providing three different image sets of preterm born infants, was set up to provide such a comparison. These sets are (i) axial scans acquired at 40 weeks corrected age, (ii) coronal scans acquired at 30 weeks corrected age and (iii) coronal scans acquired at 40 weeks corrected age. Each of these three sets consists of three T1- and T2-weighted MR images of the brain acquired with a 3T MRI scanner. The task was to segment cortical grey matter, non-myelinated and myelinated white matter, brainstem, basal ganglia and thalami, cerebellum, and cerebrospinal fluid in the ventricles and in the extracerebral space separately. Any team could upload the results and all segmentations were evaluated in the same way. This paper presents the results of eight participating teams. The results demonstrate that the participating methods were able to segment all tissue classes well, except myelinated white matter.     

Gliomatosis cerebri: a rare cause of progressive headache

A 49-year-old right-handed woman was admitted for an 8-month history of unusual headache and transient diplopia. Clinical examination and brain CT scan were normal. Two months later, symptoms of raised intracranial pressure developed and a brain CT scan showed small lateral ventricles and sulci without any abnormal contrast enhancement or tumor mass. Brain MRI with T2-weighted spin echo sequences revealed a hyperintense signal in the right temporoparietal region, whereas only a slight enlargement of this region was noted on T1 spin echo. The patient deteriorated rapidly and died with uncontrollable raised cerebrospinal fluid pressure. The diagnosis of gliomatosis cerebri was made at necropsy. Gliomatosis cerebri is a rare intracranial neoplasm of neuroepithelial origin. Spread of this tumor is particularly fast in the white matter compared with the gray matter and nuclei. Clinical symptoms are not specific. The diagnosis can be suspected by MFU showing an isointense or hypointense signal in the deep white matter on T1-weighted images, and largely a hyperitense signal on T2-weighted sequences. The diagnosis is confirmed by stereotactic biopsy or necropsy. No curative treatment is currently available. Radiotherapy can delay the rapidly fatal outcome. Our case illustrates the possible onset of this rare disease by isolated cephalgia with normal early CT scan.     

White matter lesion extension to automatic brain tissue segmentation on MRI

A fully automated brain tissue segmentation method is optimized and extended with white matter lesion segmentation. Cerebrospinal fluid (CSF), gray matter (GM) and white matter (WM) are segmented by an atlas-based k-nearest neighbor classifier on multi-modal magnetic resonance imaging data. This classifier is trained by registering brain atlases to the subject. The resulting GM segmentation is used to automatically find a white matter lesion (WML) threshold in a fluid-attenuated inversion recovery scan. False positive lesions are removed by ensuring that the lesions are within the white matter. The method was visually validated on a set of 209 subjects. No segmentation errors were found in 98% of the brain tissue segmentations and 97% of the WML segmentations. A quantitative evaluation using manual segmentations was performed on a subset of 6 subjects for CSF, GM and WM segmentation and an additional 14 for the WML segmentations. The results indicated that the automatic segmentation accuracy is close to the interobserver variability of manual segmentations.     

Mapping the MRI voxel volume in which thermal noise matches physiological noise--implications for fMRI

This work addresses the choice of the imaging voxel volume in blood oxygen level dependent (BOLD) functional magnetic resonance imaging (fMRI). Noise of physiological origin that is present in the voxel time course is a prohibitive factor in the detection of small activation-induced BOLD signal changes. If the physiological noise contribution dominates over the temporal fluctuation contribution in the imaging voxel, further increases in the voxel signal-to-noise ratio (SNR) will have diminished corresponding increases in temporal signal-to-noise (TSNR), resulting in reduced corresponding increases in the ability to detect activation induced signal changes. On the other hand, if the thermal and system noise dominate (suggesting a relatively low SNR) further decreases in SNR can prohibit detection of activation-induced signal changes. Here we have proposed and called the "suggested" voxel volume for fMRI the volume where thermal plus system-related and physiological noise variances are equal. Based on this condition we have created maps of fMRI suggested voxel volume from our experimental data at 3T, since this value will spatially vary depending on the contribution of physiologic noise in each voxel. Based on our fast EPI segmentation technique we have found that for gray matter (GM), white matter (WM), and cerebral spinal fluid (CSF) brain compartments the mean suggested cubical voxel volume is: (1.8 mm)3, (2.1 mm)3 and (1.4 mm)3, respectively. Serendipitously, (1.8 mm)3 cubical voxel volume for GM approximately matches the cortical thickness, thus optimizing BOLD contrast by minimizing partial volume averaging. The introduced suggested fMRI voxel volume can be a useful parameter for choice of imaging volume for functional studies.     

脑脊液伪影在3.0TMR二维和三维液体衰减反转恢复图像上的表现

目的通过与1.5T MRI比较,探讨3.0T MR上2D和3D FLAIR图像脑脊液伪影的表现。方法 50例受试者行1.5T MR2D液体衰减反转恢复(FLAIR)成像,另外50例行3.0T MR2D和3D FLAIR成像,按4级分类法记录各受试者脑室系统、鞍上池和桥前池内脑脊液伪影:0级,无伪影;1级,伪影信号低于灰质;2级,伪影信号与灰质相似;3级,伪影信号高于灰质。将2级和3级合称为高级别。结果 1.5T MR2D FLAIR图像上,9例未见脑脊液伪影,余41例中共发现94处脑脊液伪影,其中2级最常见(54.3%),最常见于四脑室。3.0T MR2D FLAIR图像上,每例受试者至少有2处脑脊液伪影,其中3级最多见,占总伪影数50.0%。3.0T MR2D FLAIR上各部位高级别脑脊液伪影例数均明显高于1.5T。3D FLAIR图像上均未见脑脊液伪影。结论 2D FLAIR成像时,同1.5T MRI相比,3.0T MRI上高级别脑脊液伪影明显增加;但3D FLAIR图像上脑脊液信号抑制完全、无伪影。