Longitudinal Evaluation of Structural Changes in Frontotemporal Dementia Using Artificial Neural Networks

Segmentation of brain magnetic resonance imaging (MRI) data plays an important role in the computer-aided diagnosis and neuroscience research. Fuzzy c-means (FCM) clustering algorithm is one of the most usually used techniques for brain MRI image segmentation because of its fuzzy nature. However, the conventional FCM method fails to carry out segmentation well enough due to intensity inhomogeneity in MRI data. To overcome this issue, we propose an improved algorithm based on FCM clustering for segmentation of brain MRI data. Specifically, we modify the conventional FCM algorithm to allow for intensity inhomogeneity by introducing the regularization of the neighborhood influence and bias field. Results show that our proposed algorithm obtains reasonable segmentation of white matter (WM), gray matter (GM), and cerebrospinal fluid (CSF) from MRI data, which is superior to the expectation-maximization (EM) and conventional FCM methods.

Commentary on “Diagnosis of Alzheimer's disease: Two decades of progress.” Central role of technology in the treatment and prevention of Alzheimer's disease

Cranial Ultrasonography

The brain volume of chronic drinkers is known to partially recover with abstinence from alcohol. To investigate the relative contribution of grey and white brain matter to this process, magnetic resonance imaging and brain tissue segmentation was used to study brain tissue in acute alcohol withdrawal and abstinence in seven alcohol-dependent men. The patients were studied on three occasions; within 48 h after the last drink and approximately one month and two and a half months later. Total brain tissue class volumes [grey matter (GM) and white matter (WM) and cerebrospinal fluid (CSF)] were measured. Eleven healthy volunteers were scanned twice to serve as a control group. The alcohol-dependent patients were investigated with regard to drinking variables, neuropsychological performance and blood biochemistry. In the alcohol-dependent patients, intracranial volume and total GM volume did not change between scan occasions, except in a single patient who demonstrated a GM increase of 4.8% (4.2% relative volume) between scans 2 and 3. For all patients, the increase in total WM volume ranged between 1.9 and 22.4% (absolute volumes) and 2.1 and 21.2% (relative volumes). Between scans 2 and 3, the increase in total WM volume ranged between 0.3 and 13.2% (absolute volumes), and between 1.5 and 14.0% (relative volumes). One patient resumed drinking and was investigated a second time during acute withdrawal. In this patient, the measured decrease of 8.1 and 8.5% of absolute and relative WM volumes corresponded to the size of the volume increase between scans 1 and 2. CSF, GM and WM volumes in the healthy subjects were constant over time. The results demonstrate that changes in brain volume during short-term abstinence in chronic alcohol-dependent patients are confined to the WM. The time limit of WM volume restitution is variable and continues longer than 3 weeks after withdrawal.

Despite the many neuroimaging studies that suggest gray matter volume reductions in schizophrenia, there is no compelling postmortem evidence to suggest neuronal loss, nor is there a distinctive or specific signature of gray matter abnormalities in schizophrenia. Instead, there appears to be no observed focal lesion that characterizes gray matter pathology in schizophrenia. This state of affairs has led to a movement away from conceptualizing schizophrenia as resulting from a single lesion, to conceptualizing schizophrenia as arising from abnormal communication between brain regions. Abnormal connectivity between brain regions is not a new concept, as Bleuler (1911) postulated such abnormalities. What is new, however, is an appreciation that brain regions that are not spatially proximal may be connected functionally into neural networks and that to understand altered neural connectivity we need to have a better understanding of the connections between brain regions which are made possible by white matter, the main infrastructure in the brain that makes possible long distance communication among neurons. Accordingly, a focus on white matter connections in the brain has become increasingly of interest in schizophrenia, particularly in the areas of imaging studies, postmortem studies, and new animal models of schizophrenia. This special issue focuses on schizophrenia and oligodendrocytes, the latter a class of neuroglia that give rise to myelin. The primary aim of this special issue is to understand and to clarify further white matter pathology in schizophrenia and how it may contribute to disconnectivity among brain regions, which results in the observed cognitive, behavioral, and clinical symptoms in this disorder. The first paper begins with a description of the theory that schizophrenia reflects a disorder of connectivity. The proposition put forth is that schizophrenia is, at least in part, the result of abnormal communication between brain regions that may be spatially distant but nonetheless functionally connected. As white matter is the main infrastructure for brain connectivity, this paper focuses on the use of a magnetic resonance imaging (MRI) technique called diffusion tensor imaging (DTI) to gain insight into the role of white matter abnormalities in schizophrenia. This paper also focuses on the functional implications of white matter abnormalities, particularly with regard to the possible role of myelin in modulating the transmission velocity of neural discharges. The paper ends with a speculative hypothesis about the relationship between gray matter and white matter abnormalities in schizophrenia. The second paper also focuses on altered neuronal connectivity. Here, however, the focus is on altered connectivity and impaired myelination in a postmortem study of schizophrenia and normal controls, where electron microscopy is used to study myelinated fibers and oligodendrocytes. This morphometric study examines myelinated fibers in the prefrontal cortex in both gray and white matter. Six types of abnormal fibers and ultrastructural alterations are described in the schizophrenia sample. The study reveals increased pathological fibers in gray matter in both young and elderly patients that are associated with positive symptoms, while in elderly patients, the frequency of pathological fibers in white matter is increased and these changes are associated with more negative symptoms. The key implication here is that altered myelinated fibers in white matter in schizophrenia likely follow alterations of myelinated fibers in gray matter that occur earlier in the course of illness. The third paper is a postmortem study on age-related changes in the number of oligodendrocytes in prefrontal cortex in schizophrenia, bipolar disorder and major depression. The study reveals an age-related increase in numerical density of oligodendrocytes in layer VI and adjacent white matter of BA9 and 10 that is seen only in normal controls and not in schizophrenia or in mood disorders. The absence of the normal age-related increase in oligodendrocytes in patients suggests that this aspect of normal brain development is dysregulated in schizophrenia and in mood disorders, and confirms postmortem and imaging data, further highlighting the fact that the absence of a normal age-related increase in oligodendrocytes is a key shared feature in these psychiatric populations. The fourth paper reviews white matter abnormalities in schizophrenia, including imaging and postmortem findings on the disconnectivity theory of schizophrenia, a major theme that underlies the three previous papers in this special issue. This paper then reviews some of the white matter diseases that commonly lead to psychotic symptoms and then reviews transgenic and mutant mouse models of schizophrenia. More specifically, genetic mouse models such as Plp1 transgenic mice and mutant mice heterogeneous for either NRG1 or its receptor erbB4, as well as Nogo-A- deficient mice are reviewed. This is followed by a focus on the myelin toxicity model of mice fed cuprizone. In the early stage (the second and third week) of the cuprizone feeding, mice show higher dopamine levels and lower norepinephrine levels in their prefrontal cortex, along with behavioral changes indicative of increased CNS activity. In the late stage (weeks 4–6), when demyelination and oligodendrocyte loss are obvious, mice display cognitive deficits as well as deficits in social interactions that are reminiscent of social withdrawal seen in patients with schizophrenia. Importantly, this cuprizone-feeding mouse may be used as an in vivo platform for psychopharmacological studies testing the effects of antipsychotic drugs on white matter alterations and associated behavioral changes, in addition to providing a new animal model of schizophrenia. The fifth and final paper of this special issue focuses on abnormal behavior and microstructural changes in juvenile mice that are repeatedly exposed to nonneurotoxic levels of amphetamine. These mice are compared with nonamphetamine treated mice. Oligodendrocyte numbers and three proteins expressed in mature oligodendrocytes are investigated in these young male mice at sacrifice. Treated mice showed higher locomotion and impaired spatial working memory, along with lowered Noga-A and GST-pi proteins, and lower MBP proteins, in frontal cortex and hippocampus and fewer mature oligodendrocytes in frontal cortex and corpus callosum, as well as lower MBP staining in frontal cortex, corpus callosum, and hippocampus, than untreated mice. These differences suggest that in wild-type mice late developing white matter is vulnerable to amphetamine and may lead to compromised white matter with increased dopamine in specific brain regions. Not only do these findings explain well the increased locomotion and impaired spatial working memory observed in amphetamine-treated mice, but also may serve as a principle for the amphetamine-treated mouse being used as a novel animal model of schizophrenia. To summarize, compared to the focus on gray matter and neurons, research on white matter and myelin/glial has been sparse in schizophrenia. In this special issue, neuroimaging findings are presented that suggest that myelin abnormalities in schizophrenia may underlie the white matter abnormalities observed using diffusion tensor imaging techniques. Data from postmortem studies also confirm the presence of oligodendrocyte and myelin abnormalities in schizophrenia as well as mood disorders. Oligodendrocytes provide protection and improve communication between brain regions and the oligodendroglia-producing myelin is approaching maturity in the same time frame as the onset of symptoms. Further, normal age-related white matter increases and related gray matter changes that likely predate the white matter changes are dysregulated in schizophrenia and may contribute to the clinical symptomatology of the disease. Finally, the recently developed animal models are not only of help in understanding connectivity abnormalities in schizophrenia but also provide novel platforms to test novel therapeutic approaches for schizophrenic patients. Treatments specifically targeting white matter, for example, may improve on the efficacy of those treatments currently in use. Martha E. Shenton Haiyun Xu Vahram Haroutunian George Bartzokis

Despite extensive efforts to obtain accurate segmentation of magnetic resonance imaging (MRI) scans of a head, it remains challenging primarily due to variations in intensity distribution, which depend on the equipment and parameters used. The goal of this study is to evaluate the effectiveness of an automatic segmentation method for head MRI scans using a multistep Dense U-Net (MDU-Net) architecture. The MDU-Net-based method comprises two steps. The first step is to segment the scalp, skull, and whole brain from head MRI scans using a convolutional neural network (CNN). In the first step, a hybrid network is used to combine 2.5D Dense U-Net and 3D Dense U-Net structure. This hybrid network acquires logits in three orthogonal planes (axial, coronal, and sagittal) using 2.5D Dense U-Nets and fuses them by averaging. The resultant fused probability map with head MRI scans then serves as the input to a 3D Dense U-Net. In this process, different ratios of active contour loss and focal loss are applied. The second step is to segment the cerebrospinal fluid (CSF), white matter, and gray matter from extracted brain MRI scans using CNNs. In the second step, the histogram of the extracted brain MRI scans is standardized and then a 2.5D Dense U-Net is used to further segment the brain's specific tissues using the focal loss. A dataset of 100 head MRI scans from an OASIS-3 dataset was used for training, internal validation, and testing, with ratios of 80%, 10%, and 10%, respectively. Using the proposed approach, we segmented the head MRI scans into five areas (scalp, skull, CSF, white matter, and gray matter) and evaluated the segmentation results using the Dice similarity coefficient (DSC) score, Hausdorff distance (HD), and the average symmetric surface distance (ASSD) as evaluation metrics. We compared these results with those obtained using the Res-U-Net, Dense U-Net, U-Net++, Swin-Unet, and H-Dense U-Net models. The MDU-Net model showed DSC values of 0.933, 0.830, 0.833, 0.953, and 0.917 in the scalp, skull, CSF, white matter, and gray matter, respectively. The corresponding HD values were 2.37, 2.89, 2.13, 1.52, and 1.53mm, respectively. The ASSD values were 0.50, 1.63, 1.28, 0.26, and 0.27mm, respectively. Comparing these results with other models revealed that the MDU-Net model demonstrated the best performance in terms of the DSC values for the scalp, CSF, white matter, and gray matter. When compared with the H-Dense U-Net model, which showed the highest performance among the other models, the MDU-Net model showed substantial improvements in the HD view, particularly in the gray matter region, with a difference of approximately 9%. In addition, in terms of the ASSD, the MDU-Net model outperformed the H-Dense U-Net model, showing an approximately 7% improvements in the white matter and approximately 9% improvements in the gray matter. Compared with existing models in terms of DSC, HD, and ASSD, the proposed MDU-Net model demonstrated the best performance on average and showed its potential to enhance the accuracy of automatic segmentation for head MRI scans.

The correct interpretation of clinical positron emission tomography (PET) data depends largely on the physical limits of the PET scanner. The partial volume effect (PVE) is related to the size of the studied object compared to the spatial resolution. It represents one of the most important limiting factors in quantitative data analysis. This effect is increased in the case of atrophy, as in patients with Alzheimer disease (AD), and it influences measurement of the metabolic reduction generally seen in cerebral degeneration. In this case, interpretation can be biased, because cortical activity will be underestimated due to the atrophy. In general, anatomical images of AD patients have shown diffuse atrophy, while PET studies have found widespread hypometabolism affecting the parietal and temporal lobes. Although hypometabolic areas usually correspond to atrophic regions, they also occur without such changes. Thus, the aim is to differentiate authentic hypometabolism (decrease of glucose consumption per unit volume of gray matter) from that due to PVE from atrophy (cell loss). Consequently, we are using a method for three-dimensional (3D) correction of human PET data with 3D magnetic resonance imaging (MRI). We measured atrophy and metabolism by using both T1-weighted MR images and high and medium resolution PET scans. We injected 12 patients and controls with [18F]fluorodeoxyglucose for glucose consumption measurements. Atrophy was estimated in the following way. We isolated the cerebral structures, using a segmentation technique on the MRI scans, into gray matter (GM), white matter, and cerebrospinal fluid. We superimposed the PET images onto the MR images to obtain anatomo-functional correlations. We degraded the segmented MR images to the resolution of the PET images by a convolution process to create a PET image correction map. We corrected the metabolic PET data for the PVE. We studied the cerebral metabolic rate of glucose in the GM where metabolic variation is the most relevant to AD. By dealing with problems relating to the sensitivity to the segmentation and to the PET-MRI coregistration, computation of MRI convolution processes provided the degree of PVE on a pixel-by-pixel basis, allowing correction of hypometabolisms contained in GM PET values. Global cortical metabolism increased after correction for PVE by, on average, 29 and 24% for tomographs acquired with medium (TTV03 LETI) and high (ECAT 953B CTI/Siemens) resolution, respectively, whereas the cortical metabolism increased by 75 and 65% for the respective tomographs in AD patients. The difference of metabolism between scans after correction for PVE was less than before correction, decreasing from 31 to 17%. This difference was most marked in the frontal and temporal lobes. Fusion imaging allowed correction for PVE in metabolic data using 3D MRI and determination of whether a change in the apparent radiotracer concentration in PET data reflected an alteration in GM volume, a change in radiotracer concentration per unit volume of GM, or both.

Differentiation of absolute metabolite concentrations between gray and white matter in the occipital region of normal human brain was performed by in vivo localized single-voxel 1H magnetic resonance spectroscopy at 1.5 Tesla with long echo time (136 ms). With the combination of image segmentation between white and gray matter and cerebrospinal fluid, signal compensation of T1 and T2 effects, tissue water signal as the internal concentration reference, as well as compensation by different water contents in gray and white matters, it was determined that the levels of N-acetylaspartate (NAA), creatine and/or phosphocreatine (Cr), and choline-containing compounds (Cho) in gray matter were significantly higher than in white matter. The averaged NAA, Cr, and Cho concentrations in gray matter were 11.0, 9.7, and 1.9 mM/liter, respectively, in comparison with 7.5, 5.2, and 1.6 mM/liter in white matter. These results suggest that precise composition of white and gray matter and cerebrospinal fluid is necessary to avoid partial voluming effect in a single voxel and to accurately quantify the metabolite concentrations.

Segmentation of white matter, grey matter and cerebrospinal fluid from brain MR images using a modified FCM based on double estimation

Background: Structural brain damage, especially to white matter, is well documented in chronic alcohol abuse. There is also evidence for brain metabolic abnormalities in this condition. It is unknown, however, to what extent these structural and metabolic changes are present in treated alcohol abusers who achieve long-term abstinence versus treatment-naïve, heavily drinking individuals. Methods: This study compared 12 recovering alcoholics with 8 actively heavily drinking subjects. Participants underwent magnetic resonance (MR) imaging and proton MR spectroscopic imaging of the brain. Semiautomated image segmentation techniques yielded volumes for gray matter, white matter, white matter lesions, and cerebral spinal fluid in multiple brain regions defined by Talairach stereotaxic coordinates. Automated spectral processing methods yielded gray and white matter concentrations of the metabolites N-acetylaspartate, creatine, and choline for the same regions. Results: Recovering alcoholics had greater volumes of frontal white matter, but the opposite was true for white matter in a “remainder” region encompassing the basal frontal and temporal lobes, the cerebellum, and the brainstem. Recovering alcoholics also had smaller volumes of white matter lesions in whole brain, in occipital and mesial parietal regions, and in the remainder region. Recovering alcoholics had greater gray matter volumes in the orbital frontal pole and postcentral gyrus, but smaller gray matter volumes in the anterior cingulate. Whole-brain and regional metabolite concentrations did not differ significantly between the two groups. Conclusions: White and gray matter volumes in different regions of the brain were greater or smaller in recovering, treated alcoholics. The findings suggest region-specific structural recovery from chronic alcohol–induced brain injury, but also region-specific long-term structural damage in abstinent alcoholics. White matter lesions were widespread in active drinkers and may partly resolve during long-term abstinence. Proton MR spectroscopic measures, as applied in this cross-sectional study, were largely ineffective in revealing metabolic effects of abstinence on the alcohol-damaged brain.

Although broadly utilized, computed tomography (CT) has been superseded by magnetic resonance imaging (MRI) for the volumetric assessment of white matter (WM), grey matter (GM) and cerebrospinal fluid (CSF) in the aging human brain. Nevertheless, many scenarios remain where MRI is unavailable or discouraged; furthermore, in developing countries, CT can often be the only accessible imaging method for assessing brain structure in older patients with mild cognitive impairment or Alzheimer’s disease. Thus, there is merit in developing effective approaches for the estimation of brain volumetrics using CT. Here, MRI and CT scans were acquired from 10 older adults [mean (µ) ± standard deviation (σ) of age = 65 ± 7 yrs; 5 females]. To demarcate WM, GM and CSF from head CT, we developed a brain segmentation method based upon probabilistic, atlas-dependent classification. MRI-only segmentation was compared to CT-only segmentation; similarity was calculated through the Dice coefficient (DC). A normal distribution of DCs was found after contrasting the methods [µ ± σ across participants: 85.5% ± 4.6% (WM), 86.7% ± 5.6% (GM) and 91.3% ± 2.8% (CSF)], suggesting a satisfactory capacity of CT to assess brain volumetrics. Sensitivity was adequate: WM, GM and CSF volumes were estimated within ~5%, ~4% and ~3% of their MRI-based values. There was no indication of volume over- or under-estimation with CT [t (9) = 0.89, p > 0.80]. These results facilitate the integration of CT-based brain volumetrics with MRI, thereby offering a wider range of methods for quantifying macroscale brain changes in neurodegenerative diseases.

This thesis focuses on finding ways to differentiate the gray matter (GM) and white matter (WM) in magnetic resonance (MR) images of the human spinal cord (SC). The aim of this project is to quantify tissue loss in these compartments to study their implications on the progression of multiple sclerosis (MS). To this end, we propose segmentation algorithms that we evaluated on MR images of healthy volunteers. Segmentation of GM and WM in MR images can be done manually by human experts, but manual segmentation is tedious and prone to intra- and inter-rater variability. Therefore, a deterministic automation of this task is necessary. On axial 2D images acquired with a recently proposed MR sequence, called AMIRA, we experiment with various automatic segmentation algorithms. We first use variational model-based segmentation approaches combined with appearance models and later directly apply supervised deep learning to train segmentation networks. Evaluation of the proposed methods shows accurate and precise results, which are on par with manual segmentations. We test the developed deep learning approach on images of conventional MR sequences in the context of a GM segmentation challenge, resulting in superior performance compared to the other competing methods. To further assess the quality of the AMIRA sequence, we apply an already published GM segmentation algorithm to our data, yielding higher accuracy than the same algorithm achieves on images of conventional MR sequences. On a different topic, but related to segmentation, we develop a high-order slice interpolation method to address the large slice distances of images acquired with the AMIRA protocol at different vertebral levels, enabling us to resample our data to intermediate slice positions. From the methodical point of view, this work provides an introduction to computer vision, a mathematically focused perspective on variational segmentation approaches and supervised deep learning, as well as a brief overview of the underlying project's anatomical and medical background.

In this paper, we propose an automated approach based on the geometric median shift algorithm over Riemannian manifolds, for the brain tumor detection and segmentation in magnetic resonance images (MRI). This approach is based on the geometric median, geodesic distance. We propose the median shift to overcome the limitation of mean which is not necessary a point in a set. The geodesic distance can describe data points distributed on a manifold, compared to the Euclidean distance, and produce efficient results for image analysis. Coupled with k-means algorithm, the proposed framework can cluster the brain image into tree regions (gray matter, white matter and cerebrospinal fluid) and abnormalities regions. We applied this approach to clustering the brain tissues and brain tumor segmentation, which is validated on a synthetic brain MRI. The obtained results using two datasets show the efficiency of the used algorithm validated qualitatively by the measurement of Dice Similarity Coefficient.

We describe a fully-automatic three-dimensional (3-D)-segmentation technique for brain magnetic resonance (MR) images. By means of Markov random fields (MRF's) the segmentation algorithm captures three features that are of special importance for MR images, i.e., nonparametric distributions of tissue intensities, neighborhood correlations, and signal inhomogeneities. Detailed simulations and real MR images demonstrate the performance of the segmentation algorithm. In particular, the impact of noise, inhomogeneity, smoothing, and structure thickness are analyzed quantitatively. Even single-echo MR images are well classified into gray matter, white matter, cerebrospinal fluid, scalp-bone, and background. A simulated annealing and an iterated conditional modes implementation are presented.

Objective To explore whether gray matter and white matter presented abnormalities in neural circuit in patients with OCD by using multimodal imaging strategies. Methods Magnetic resonance imaging (MRI) scans from 54 OCD patients and 54 healthy controls (HC) were acquired. Gray matter volume was compared using the voxel-based morphometry (VBM) approach based on the SPM software. Fractional anisotropy (FA) difference was also evaluated using tract based spatial statistics (TBSS) based on the FSL software. Results Compared with the HC, Patients with OCD showed decreased gray matter volume in the left middle frontal gyrus, left anterior cingulate and paracingulate gyri, left precental gyrus and right inferior temporal gyrus (P<0.05, Alphasim corrected); meanwhile, decreased FA values in the body and genu of corpus callosum were found. Conclusions Patients with OCD demonstrated widespread gray matter and white matter alterations especially in the incortico-striatal-thalamic-cortical (CSTC)circuit. Therefore, Patients with OCD might present altered gray and white matter structures simultaneously. Key words: Obsessive-compulsive disorder; Magnetic resonance imaging; Diffusion magnetic resonance imaging