Evaluation of feature selection algorithms for detection of depression from brain sMRI scans

Detection of depression from structural MRI (sMRI) scans is relatively new in the mental health diagnosis. Such detection requires processes including image acquisition and pre-processing, feature extraction and selection, and classification. Identification of a suitable feature selection (FS) algorithm will facilitate the enhancement of the detection accuracy by selection of important features. In the field of depression study, there are very limited works that evaluate feature selection algorithms for sMRI data. This paper investigates the performance of four algorithms for FS of volumetric attributes in sMRI scans. The algorithms are One Rule (OneR), Support Vector Machine (SVM), Information Gain (IG) and ReliefF. The performances of the algorithms are determined through a set of experiments on sMRI brain scans. An experimental procedure is developed to measure the performance of the tested algorithms. The result of the evaluation of the FS algorithms is discussed by using a number of analyses.

An empirical comparison of classification algorithms for diagnosis of depression from brain sMRI scans

To be diagnostically effective, structural magnetic resonance imaging (sMRI) must reliably distinguish a depressed individual from a healthy individual at individual scans level. One of the tasks in the automated diagnosis of depression from brain sMRI is the classification. It determines the class to which a sample belongs (i.e., depressed/not depressed, remitted/not-remitted depression) based on the values of its features. Thus far, very limited works have been reported for identification of a suitable classification algorithm for depression detection. In this paper, different types of classification algorithms are compared for effective diagnosis of depression. Ten independent classification schemas are applied and a comparative study is carried out. The algorithms are: Naïve Bayes, Support Vector Machines (SVM) with Radial Basis Function (RBF), SVM Sigmoid, J48, Random Forest, Random Tree, Voting Feature Intervals (VFI), LogitBoost, Simple KMeans Classification Via Clustering (KMeans) and Classification Via Clustering Expectation Minimization (EM) respectively. The performances of the algorithms are determined through a set of experiments on sMRI brain scans. An experimental procedure is developed to measure the performance of the tested algorithms. A classification accuracy evaluation method was employed for evaluation and comparison of the performance of the examined classifiers.

Degree of contribution (DoC) feature selection algorithm for structural brain MRI volumetric features in depression detection

Accurate detection of depression at an individual level using structural magnetic resonance imaging (sMRI) remains a challenge. Brain volumetric changes at a structural level appear to have importance in depression biomarkers studies. An automated algorithm is developed to select brain sMRI volumetric features for the detection of depression.

Classifying Child Maltreatment by Brain Imaging Data

Thesis (Master's)--University of Washington, 2016-08

Mapping genetic influences on brain fiber architecture with high angular resolution diffusion imaging (HARDI)

We report the first 3D maps of genetic effects on brain fiber complexity. We analyzed HARDI brain imaging data from 90 young adult twins using an information-theoretic measure, the Jensen-Shannon divergence (JSD), to gauge the regional complexity of the white matter fiber orientation distribution functions (ODF). HARDI data were fluidly registered using Karcher means and ODF square-roots for interpol ation; each subject's JSD map was computed from the spatial coherence of the ODFs in each voxel's neighborhood. We evaluated the genetic influences on generalized fiber anisotropy (GFA) and complexity (JSD) using structural equation models (SEM). At each voxel, genetic and environmental components of data variation were estimated, and their goodness of fit tested by permutation. Color-coded maps revealed that the optimal models varied for different brain regions. Fiber complexity was predominantly under genetic control, and was higher in more highly anisotropic regions. These methods show promise for discovering factors affecting fiber connectivity in the brain.

Interpretable semantic vectors from a joint model of brain-and text-based meaning

Vector space models (VSMs) represent word meanings as points in a high dimensional space. VSMs are typically created using a large text corpora, and so represent word semantics as observed in text. We present a new algorithm (JNNSE) that can incorporate a measure of semantics not previously used to create VSMs: brain activation data recorded while people read words. The resulting model takes advantage of the complementary strengths and weaknesses of corpus and brain activation data to give a more complete representation of semantics. Evaluations show that the model 1) matches a behavioral measure of semantics more closely, 2) can be used to predict corpus data for unseen words and 3) has predictive power that generalizes across brain imaging technologies and across subjects. We believe that the model is thus a more faithful representation of mental vocabularies.

Polymorphisms in genes involved in neurodevelopment may be associated with altered brain morphology in schizophrenia: Preliminary evidence

An abnormality in neurodevelopment is one of the most robust etiologic hypotheses in schizophrenia (SZ). There is also strong evidence that genetic factors may influence abnormal neurodevelopment in the disease. The present study evaluated in SZ patients, whose brain structural data had been obtained with magnetic resonance imaging (MRI), the possible association between structural brain measures, and 32 DNA polymorphisms,located in 30 genes related to neurogenesis and brain development. DNA was extracted from peripheral blood cells of 25 patients with schizophrenia, genotyping was performed using diverse procedures, and putative associations were evaluated by standard statistical methods (using the software Statistical Package for Social Sciences - SPSS) with a modified Bonferroni adjustment. For reelin (RELN), a protease that guides neurons in the developing brain and underlies neurotransmission and synaptic plasticity in adults, an association was found for a non-synonymous polymorphism (Va1997Leu) with left and right ventricular enlargement. A putative association was also found between protocadherin 12 (PCDH12), a cell adhesion molecule involved in axonal guidance and synaptic specificity, and cortical folding (asymmetry coefficient of gyrification index). Although our results are preliminary, due to the small number of individuals analyzed, such an approach could reveal new candidate genes implicated in anomalous neurodevelopment in schizophrenia. (c) 2007 Elsevier Ireland Ltd. All rights reserved.

Investigating machine learning techniques for detection of depression using structural MRI volumetric features

Structural MRI offers anatomical details and high sensitivity to pathological changes. It can demonstrate certain patterns of brain changes present at a structural level. Research to date has shown that volumetric analysis of brain regions has importance in depression detection. However, such analysis has had very minimal use in depression detection studies at individual level. Optimally combining various brain volumetric features/attributes, and summarizing the data into a distinctive set of variables remain difficult. This study investigates machine learning algorithms that automatically identify relevant data attributes for depression detection. Different machine learning techniques are studied for depression classification based on attributes extracted from structural MRI (sMRI) data. The attributes include volume calculated from whole brain, white matter, grey matter and hippocampus. Attributes subset selection is performed aiming to remove redundant attributes using three filtering methods and one hybrid method, in combination with ranker search algorithms. The highest average classification accuracy, obtained by using a combination of both SVM-EM and IG-Random Tree algorithms, is 85.23%. The classification approach implemented in this study can achieve higher accuracy than most reported studies using sMRI data, specifically for detection of depression.

Using an evidence-based care bundle to improve Thai emergency nurses' knowledge of care for patients with severe traumatic brain injury

There is known variation in Thai nurses' knowledge regarding the best available evidence for care of patients with severe traumatic brain injury. The purpose of this study was to examine the impact of an evidence-based care bundle on Thai emergency nurses' knowledge regarding management of patients with severe traumatic brain injury. A pre-test/post-test design was used. The study intervention was an evidence-based care bundle for initial nursing management of patients with severe traumatic brain injury. Data were collected from 31 Registered Nurses using multiple choice questions. Results revealed a statistically significant improvement in overall knowledge scores after care bundle implementation (p < 0.001). There were statistically significant improvements in five areas of knowledge: understanding of target end-tidal carbon dioxide levels (p < 0.001), implications of hypocapnia in severe traumatic brain injury (p = 0.01), implications of hypercapnia in severe traumatic brain injury (p = 0.02), importance of maintaining head and neck in neutral position (p = 0.05), and administration of sedatives and analgesics in severe traumatic brain injury (p = 0.01). This study suggested that implementation of an evidence-based care bundle improved emergency nurses' knowledge regarding management of patients with severe traumatic brain injury.

Volatile anesthetics influence blood-brain barrier integrity by modulation of tight junction protein expression in traumatic brain injury

Disruption of the blood-brain barrier (BBB) results in cerebral edema formation, which is a major cause for high mortalityrnafter traumatic brain injury (TBI). As anesthetic care is mandatory in patients suffering from severe TBI it may be importantrnto elucidate the effect of different anesthetics on cerebral edema formation. Tight junction proteins (TJ) such as zonularnoccludens-1 (ZO-1) and claudin-5 (cl5) play a central role for BBB stability. First, the influence of the volatile anestheticsrnsevoflurane and isoflurane on in-vitro BBB integrity was investigated by quantification of the electrical resistance (TEER) inrnmurine brain endothelial monolayers and neurovascular co-cultures of the BBB. Secondly brain edema and TJ expression ofrnZO-1 and cl5 were measured in-vivo after exposure towards volatile anesthetics in native mice and after controlled corticalrnimpact (CCI). In in-vitro endothelial monocultures, both anesthetics significantly reduced TEER within 24 hours afterrnexposure. In BBB co-cultures mimicking the neurovascular unit (NVU) volatile anesthetics had no impact on TEER. In healthyrnmice, anesthesia did not influence brain water content and TJ expression, while 24 hours after CCI brain water contentrnincreased significantly stronger with isoflurane compared to sevoflurane. In line with the brain edema data, ZO-1 expressionrnwas significantly higher in sevoflurane compared to isoflurane exposed CCI animals. Immunohistochemical analysesrnrevealed disruption of ZO-1 at the cerebrovascular level, while cl5 was less affected in the pericontusional area. The studyrndemonstrates that anesthetics influence brain edema formation after experimental TBI. This effect may be attributed tornmodulation of BBB permeability by differential TJ protein expression. Therefore, selection of anesthetics may influence thernbarrier function and introduce a strong bias in experimental research on pathophysiology of BBB dysfunction. Futurernresearch is required to investigate adverse or beneficial effects of volatile anesthetics on patients at risk for cerebral edema.

Core networks for visual-concrete and abstract thought content: a brain electric microstate analysis

Commonality of activation of spontaneously forming and stimulus-induced mental representations is an often made but rarely tested assumption in neuroscience. In a conjunction analysis of two earlier studies, brain electric activity during visual-concrete and abstract thoughts was studied. The conditions were: in study 1, spontaneous stimulus-independent thinking (post-hoc, visual imagery or abstract thought were identified); in study 2, reading of single nouns ranking high or low on a visual imagery scale. In both studies, subjects' tasks were similar: when prompted, they had to recall the last thought (study 1) or the last word (study 2). In both studies, subjects had no instruction to classify or to visually imagine their thoughts, and accordingly were not aware of the studies' aim. Brain electric data were analyzed into functional topographic brain images (using LORETA) of the last microstate before the prompt (study 1) and of the word-type discriminating event-related microstate after word onset (study 2). Conjunction analysis across the two studies yielded commonality of activation of core networks for abstract thought content in left anterior superior regions, and for visual-concrete thought content in right temporal-posterior inferior regions. The results suggest that two different core networks are automatedly activated when abstract or visual-concrete information, respectively, enters working memory, without a subject task or instruction about the two classes of information, and regardless of internal or external origin, and of input modality. These core machineries of working memory thus are invariant to source or modality of input when treating the two types of information.

Different duration of at-risk mental state associated with neurofunctional abnormalities. A multimodal imaging study

Objectives: Neurofunctional alterations are correlates of vulnerability to psychosis, as well as of the disorder itself. How these abnormalities relate to different probabilities for later transition to psychosis is unclear. We investigated vulnerability- versus disease-related versus resilience biomarkers of psychosis during working memory (WM) processing in individuals with an at-risk mental state (ARMS). Experimental design: Patients with “first-episode psychosis” (FEP, n = 21), short-term ARMS (ARMS-ST, n = 17), long-term ARMS (ARMS-LT, n = 16), and healthy controls (HC, n = 20) were investigated with an n-back WM task. We examined functional magnetic resonance imaging (fMRI) and structural magnetic resonance imaging (sMRI) data in conjunction using biological parametric mapping (BPM) toolbox. Principal observations: There were no differences in accuracy, but the FEP and the ARMS-ST group had longer reaction times compared with the HC and the ARMS-LT group. With the 2-back > 0-back contrast, we found reduced functional activation in ARMS-ST and FEP compared with the HC group in parietal and middle frontal regions. Relative to ARMS-LT individuals, FEP patients showed decreased activation in the bilateral inferior frontal gyrus and insula, and in the left prefrontal cortex. Compared with the ARMS-LT, the ARMS-ST subjects showed reduced activation in the right inferior frontal gyrus and insula. Reduced insular and prefrontal activation was associated with gray matter volume reduction in the same area in the ARMS-LT group. Conclusions: These findings suggest that vulnerability to psychosis was associated with neurofunctional alterations in fronto-temporo-parietal networks in a WM task. Neurofunctional differences within the ARMS were related to different duration of the prodromal state and resilience factors

FUNCTIONAL PRINCIPAL COMPONENTS MODEL FOR HIGH-DIMENSIONAL BRAIN IMAGING

We establish a fundamental equivalence between singular value decomposition (SVD) and functional principal components analysis (FPCA) models. The constructive relationship allows to deploy the numerical efficiency of SVD to fully estimate the components of FPCA, even for extremely high-dimensional functional objects, such as brain images. As an example, a functional mixed effect model is fitted to high-resolution morphometric (RAVENS) images. The main directions of morphometric variation in brain volumes are identified and discussed.

Effects of minor laboratory procedures, adrenalectomy, social defeat or acute alcohol on regional brain concentrations of corticosterone

Concentrations of corticosterone in brain areas of TO strain mice were measured by radioimmunoassay. The studies examined the effects of routine laboratory maneuvers, variation during the circadian peak, adrenalectomy, social defeat and acute injections of alcohol on these concentrations. Brief handling of mice increased corticosterone levels in plasma but not in striatum and reduced those in the hippocampus. Single injections of isotonic saline raised the plasma concentrations to a similar extent as the handling, but markedly elevated concentrations in the three brain regions. Five minutes exposure to a novel environment increased hippocampal and cerebral cortical corticosterone levels and striatal concentrations showed a larger rise. However, by 30 min in the novel environment, plasma concentrations rose further while those in striatum and cerebral cortex fell to control levels and hippocampal corticosterone remained elevated. Over the period of the circadian peak the hippocampal and striatal concentrations paralleled the plasma concentrations but cerebral cortical concentrations showed only small changes. Adrenalectomy reduced plasma corticosterone concentrations to below detectable levels after 48 h but corticosterone levels were only partially reduced in the hippocampus and striatum and remained unchanged in the cerebral cortex. Single or repeated social defeat increased both brain and plasma concentrations after 1 h. Acute injections of alcohol raised the regional brain levels in parallel with plasma concentrations. The results show that measurements of plasma concentrations do not necessarily reflect the levels in brain. The data also demonstrate that corticosterone levels can change differentially in specific brain regions. These results, and the residual hormone seen in the brain after adrenalectomy, are suggestive evidence for a local origin of central corticosterone.

A deviant EEG brain microstate in acute, neuroleptic-naive schizophrenics at rest

Momentary brain electric field configurations are manifestations of momentary global functional states of the brain. Field configurations tend to persist over some time in the sub-second range (“microstates”) and concentrate within few classes of configurations. Accordingly, brain field data can be reduced efficiently into sequences of re-occurring classes of brain microstates, not overlapping in time. Different configurations must have been caused by different active neural ensembles, and thus different microstates assumedly implement different functions. The question arises whether the aberrant schizophrenic mentation is associated with specific changes in the repertory of microstates. Continuous sequences of brain electric field maps (multichannel EEG resting data) from 9 neuroleptic-naive, first-episode, acute schizophrenics and from 18 matched controls were analyzed. The map series were assigned to four individual microstate classes; these were tested for differences between groups. One microstate class displayed significantly different field configurations and shorter durations in patients than controls; degree of shortening correlated with severity of paranoid symptomatology. The three other microstate classes showed no group differences related to psychopathology. Schizophrenic thinking apparently is not a continuous bias in brain functions, but consists of intermittent occurrences of inappropriate brain microstates that open access to inadequate processing strategies and context information