基于机器学习的阿尔兹海默症分类预测

目的:应用机器学习方法,将脑结构磁共振（sMRI）、年龄、性别、受教育年限和MMSE量表评分作为特征,对阿尔兹海默症进行分类预测。方法:特征选择后,用L1正则Logistic回归、L1正则支持向量机、梯度提升树分别对脑sMRI数据进行分类预测,选出最优模型后引入年龄、性别、受教育年限和MMSE量表评分特征优化模型,用10-折交叉验证评价模型性能。结果:L1正则Logistic回归分类效果最好,加入年龄、性别、受教育年限和MMSE评分后预测准确率提高0.89%~11.42%。结论:L1正则化Logistic回归模型的sMRI+年龄+性别+受教育年限+MMSE评分特征集对阿尔兹海默症有更好的分类效果,可作为辅助诊断阿尔兹海默症的依据。     

Steady-state visual evoked potentials reveal frontally-mediated working memory activity in humans

Steady-state visual evoked potentials (SSVEPs) reflect power changes at the stimulus driving frequency and have been used to assess brain activity reflecting cognitive processing. Only one study has demonstrated SSVEP modulation associated with working memory (WM), and none have compared the spatial localization of SSVEP modulations during WM performance with other brain imaging methods. Here we examined WM-related activity recorded with dense-array SSVEPs, analyzed using low resolution electromagnetic tomography, and compared the results to our previous findings using functional magnetic resonance imaging (fMRI). WM was associated with increased SSVEP activity over the right dorsolateral prefrontal cortex, paralleling our previous fMRI findings. Frontal WM-related SSVEP power correlated selectively with task performance. These results demonstrate the utility of SSVEPs for studying representational aspects of cognition.     

Age,Sex, and Performance Influence the Visuospatial Working Memory Network in Childhood

This study describes the influence of age, sex, and working memory (WM) performance on the visuospatial WM network. Thirty-nine healthy children (7–12 years) completed a dot location functional magnetic resonance imaging (fMRI) task. Percent signal change measured the intensity and laterality indices measured the asymmetry of activation in frontal and parietal brain regions. Old children showed greater intensity of activation in parietal regions than young children but no differences in lateralization were observed. Intensity of activation was similar across sex and WM performance groups. Girls and high WM performers showed more right-sided lateralization of parietal regions than boys and low WM performers.     

Age-specific MRI templates for pediatric neuroimaging

This study created a database of pediatric age-specific magnetic resonance imaging (MRI) brain templates for normalization and segmentation. Participants included children from 4.5 through 19.5 years, totaling 823 scans from 494 subjects. Open-source processing programs (FMRIB Software Library, Statistical Parametric Mapping, Advanced Normalization Tools [ANTS]) constructed head, brain, and segmentation templates in 6-month intervals. The tissue classification (white matter [WM], gray matter [GM], cerebrospinal fluid) showed changes over age similar to previous reports. A volumetric analysis of age-related changes in WM and GM based on these templates showed expected increase/decrease pattern in GM and an increase in WM over the sampled ages. This database is available for use for neuroimaging studies (http://jerlab.psych.sc.edu/neurodevelopmentalmridatabase).     

Gray matter density and white matter integrity in pianists' brain: a combined structural and diffusion tensor MRI study

The current study combined structural magnetic resonance imaging (sMRI) and diffusion tensor MRI (DT-MRI) to investigate both gray matter density (GMD) and white matter integrity (WMI) in 18 pianists and 21 age-matched non-musicians. The pianists began their piano training at a mean age of 12. Voxel-based morphometry of the sMRI data showed that the pianists had higher GMD in the left primary sensorimotor cortex and right cerebellum. Voxel-based analysis of the DT-MRI data showed that pianists had higher fractional anisotropy (FA) (indicating higher WMI) in the right posterior limb of the internal capsule. The sMRI and DT-MRI results indicate that both the GMD and WMI of pianists may exhibit movement-related increases during adolescence or even early adulthood compared with non-musicians.     

基于一般线性模型的机器学习方法在BOLD-fMRI脑胶质瘤患者个体化运动功能区定位中的应用价值

目的 探讨基于一般线性模型（GLM）的机器学习方法在血氧水平依赖功能磁共振成像（BOLD-fMRI）定位脑胶质瘤患者个体化运动功能中的应用价值。方法 前瞻性研究。纳入2017年11月—2021年11月西安交通大学第一附属医院神经外科确诊为脑胶质瘤且病灶位于大脑运动功能区的38例患者作为机器学习模型的验证集（男25例、女13例,年龄24~69岁）,同期招募健康志愿者50例作为模型的训练集（男26例、女24例,年龄22~68岁）。采用独立成分分析法（ICA）,随机提取98例人类连接组计划（HCP）受试者的静息态功能核磁共振（rs-fMRI）特征。依据健康志愿者的rs-fMRI和基于任务的功能磁共振（tb-fMRI）的相关性,训练基于GLM的机器学习模型。观察项目：（1）采用Pearson相关系数（CC）分析比较GLM预测的激活与实际激活的相似度。（2）采用Dice系数（DC）作为模型预测效能的定量指标,比较GLM与ICA方法的预测效能。结果 （1）胶质瘤患者基于GLM的机器学习方法所预测的激活与实际tb-fMRI的功能激活相似度高[（89.47% （34/38）的患者CC值>0.30）]。（2）胶质瘤患者GLM预测任务态运动功能激活的效能,DC为0.34（0.27,0.42）,优于ICA方法的效能DC 0.26（0.16,0.30）,差异有统计学意义（Z=-3.88,P<0.001）;GLM在肿瘤半球的预测效能优于ICA方法,DC分别为0.36（0.17,0.48）和0.34（0.04,0.45）,差异有统计学意义（Z=-2.43,P=0.015）;2种方法在非肿瘤半球的预测效果均显著高于肿瘤半球（Z=-4.33、-3.59,P值均<0.001）。结论 基于GLM的机器学习方法能够很好地在术前利用rs-fMRI数据预测出胶质瘤患者的tb-fMRI运动功能激活,且其预测效果好于ICA方法。     

The effects of acute stress on human prefrontal working memory systems

We examined the relationship between acute stress and prefrontal-cortex (PFC) based working memory (WM) systems using behavioral (Experiment 1) and functional magnetic resonance imaging (fMRI; Experiment 2) paradigms. Subjects performed a delayed-response item-recognition task, with alternating blocks of high and low WM demand trials. During scanning, participants performed this task under three stress conditions: cold stress (induced by cold-water hand-immersion), a room temperature water control (induced by tepid-water hand-immersion), and no-water control (no hand-immersion). Performance was affected by WM demand, but not stress. Cold stress elicited greater salivary cortisol readings in behavioral subjects, and greater PFC signal change in fMRI subjects, than control conditions. These results suggest that, under stress, increases in PFC activity may be necessary to mediate cognitive processes that maintain behavioral organization.     

Age-Specific MRI Templates for Pediatric Neuroimaging

This study created a database of pediatric age-specific magnetic resonance imaging (MRI) brain templates for normalization and segmentation. Participants included children from 4.5 through 19.5 years, totaling 823 scans from 494 subjects. Open-source processing programs (FMRIB Software Library, Statistical Parametric Mapping, Advanced Normalization Tools [ANTS]) constructed head, brain, and segmentation templates in 6-month intervals. The tissue classification (white matter [WM], gray matter [GM], cerebrospinal fluid) showed changes over age similar to previous reports. A volumetric analysis of age-related changes in WM and GM based on these templates showed expected increase/decrease pattern in GM and an increase in WM over the sampled ages. This database is available for use for neuroimaging studies (http://jerlab.psych.sc.edu/neurodevelopmentalmridatabase).     

Neocortical and hippocampal volume loss in a human ciliopathy: A quantitative MRI study in Bardet-Biedl syndrome

Cilia are ubiquitous cell surface organelles with diverse roles from embryogenesis to adult life. The neurodevelopmental functions of the cilium are currently under investigation in animal systems, but relevance to human brain development remains uncertain. We present the first systematic investigation of structural neuroanatomy in a ciliopathy-Bardet-Biedl syndrome (BBS). Qualitative and quantitative aspects of brain structure were evaluated via magnetic resonance imaging in 10 patients with BBS (ages 14-28 years). In comparison to age and gender-matched healthy controls, BBS patients had significantly reduced total gray matter (GM) volume but no total white matter (WM) or cerebrospinal fluid volume changes. Voxel-based morphometric analysis indicated regional GM volume loss bilaterally in the anterior temporal lobes and in the medial orbitofrontal cortex, and WM volume loss in the right inferior longitudinal fasciculus. Region-of-interest measurements revealed reduced volume of the hippocampus. Two patients were found to have ventriculomegaly. Global GM reduction and regional volume reductions in the temporal lobe may underlie the learning disabilities and behavioral problems experienced by some patients with BBS. These findings are consistent with previous observations in mouse models of BBS, and further implicate the cilium in neurodevelopmental processes relevant to human cognitive function.     

Early and late onset Alzheimer's disease patients have distinct patterns of white matter damage

We investigated patterns of white matter (WM) loss in 18 early onset (EO) and 24 late onset (LO) Alzheimer's disease (AD) patients compared with 42 healthy controls (HC), and explored relationships of WM atrophy and apolipoprotein E (ApoE) genotype. Subjects underwent magnetic resonance imaging (MRI). Patterns of WM were assessed using voxel-based morphometry. Compared with healthy controls, LOAD patients had a selective parahippocampal WM loss, while EOAD patients experienced a more widespread pattern of posterior WM atrophy. The distinct regional distribution of WM atrophy reflected the topography of gray matter (GM) loss. ApoE ε4 status was associated with a greater parahippocampal WM loss in both AD groups. The greater WM atrophy in EOAD than LOAD fits with the evidence that EOAD is a more aggressive form of the disease. The ApoE ε4 effect on WM damage in AD is restricted to specific WM regions and does not seem to be related to age of onset.     

Age and gender effects on human brain anatomy: a voxel-based morphometric study in healthy elderly

The adult human brain shrinks slowly with age, but the regional specificity and tissue class specificity of this loss is unclear. Subjects (n=122) were healthy aged participants in a longitudinal cohort who undergo periodic standardized cognitive and clinical examination. Multi-spectral segmentation of magnetic resonance images into grey matter (GM), white matter (WM) and CSF was performed on cross-sectional image data using a custom template and calculated prior probability maps. Global differences were evaluated by fitting a regression model for absolute and normalized subject GM, WM, and CSF values. Global and regional patterns of GM, WM and CSF differences were assessed using optimized voxel-based morphometry (VBM). GM volume decreased with age at a rate of 2.4 cm(3)/year (-0.18%/year); CSF increased by 2.5 cm(3)/year (0.20%/year). Regression analyses showed no significant decrease in WM volume, but a focal WM decrease with age was detected in the anterior corpus callosum using VBM. Diffuse reductions of GM volume were seen with age in the frontal, parietal, and temporal cortex, cerebellum and basal ganglia. Relative regional differences in cortical GM volume with age occurred in the frontal, parietal and temporal lobes, but not in medial temporal lobe or in posterior cingulate. We did not observe significant gender effects. These findings establish a baseline for comparison with pathologic changes in human brain volume between ages 58 and 95 years.     

Temporal modeling demonstrates preserved overlearning processes in schizophrenia: an fMRI study

Working memory (WM) deficits are a core feature of schizophrenia. However, it has not been examined whether these deficits are related to altered temporal dynamics of information acquisition and changes in executive cognitive control. Therefore, the present study intended to quantify and model the dynamic process of information acquisition during continuous overlearning of WM information. It also aimed at investigating the relation between overlearning-associated change in behavioral performance and brain activity. Thirteen schizophrenic patients and 13 healthy volunteers were studied with functional magnetic resonance imaging (fMRI) while performing a recently developed overlearning paradigm [Koch K, Wagner G, von Consbruch K, Nenadic I, Schultz C, Ehle C, Reichenbach J, Sauer H, Schlösser R (2006) Temporal changes in neural activation during practice of information retrieval from short-term memory: An fMRI study. Brain Res 1107:140–150]. Consistent with the earlier study, short-term learning of stimulus material was associated with significant performance improvements and exponential signal decreases in a fronto-parieto-cerebellar network both in schizophrenic patients and in healthy volunteers. Against expectation patients exhibited stronger signal decreases relative to controls in anterior cingulate (Brodmann area (BA) 32), middle and superior temporal (BA 37, BA 22), superior frontal (BA 8/9, BA 6) and posterior parietal regions (BA 40). Furthermore, the individually modeled exponential decay rate of the blood oxygenation level-dependent signal in the right dorsolateral prefrontal cortex was significantly correlated with exponential decrease in mean behavioral response times in healthy controls while a statistical trend emerged in patients. A relative hyperactivation in the patient group was observable only at the start of the learning process and diminished with continued overlearning. This effect might indicate a gradual reduction of recruited neuronal resources and a practice-associated activation normalization in patients with schizophrenia. Our data suggest that in subacute patients learning and associated decreases in cerebral activation brought about by short-term practice are left unimpaired.     

Crossed cerebral lateralization for verbal and visuo-spatial function in a pair of handedness discordant monozygotic twins: MRI and fMRI brain imaging

To examine the nature of hemispheric lateralization for neural processes underlying verbal fluency and visuo-spatial attention, we investigated a single pair of handedness discordant monozygotic (MzHd) twins. Imaging of the brain was undertaken using magnetic resonance imaging (MRI) and functional magnetic resonance imaging (fMRI) in combination with manual performance tasks. The twins were discordant for MRI anatomical asymmetries of the pars triangularis and planum temporale, whose asymmetry was consistent with verbal laterality on fMRI. Thus, the right-handed twin had left lateralized verbal with right lateralized visuo-spatial attention, while the left-handed twin had right lateralized verbal with left lateralized visuo-spatial activation; these data lend further support for to the conclusions of Sommer et al.     

Changes in grey matter development in autism spectrum disorder

Results on grey matter (GM) structural alterations in autism spectrum disorder (ASD) are inconclusive. Moreover, little is known about age effects on brain-structure abnormalities in ASD beyond childhood. Here, we aimed to examine regional GM volumes in a large sample of children, adolescents, and adults with ASD. Magnetic resonance imaging scans were obtained in 47 male ASD subjects and 51 matched healthy controls aged 8–50 years. We used whole-brain voxel-based morphometry to first assess group differences in regional GM volume across age. Moreover, taking a cross-sectional approach, group differences in age effects on regional GM volume were investigated. Compared to controls, ASD subjects showed reduced GM volumes in the anterior cingulate cortex, posterior superior temporal sulcus, and middle temporal gyrus. Investigation of group differences in age effects on regional GM volume revealed complex, region-specific alterations in ASD. While GM volumes in the amygdala, temporoparietal junction, septal nucleus and middle cingulate cortex increased in a negative quadratic fashion in both groups, data indicated that GM volume curves in ASD subjects were shifted to the left along the age axis. Moreover, while GM volume in the right precentral gyrus decreased linearly with age in ASD individuals, GM volume development in controls followed a U-shaped pattern. Based on a large sample, our voxel-based morphometry results on group differences in regional GM volumes help to resolve inconclusive findings from previous studies in ASD. Results on age-related changes of regional GM volumes suggest that ASD is characterized by complex alterations in lifetime trajectories of several brain regions that underpin social-cognitive and motor functions.     

Comparison of SNR and CNR for in vivo mouse brain imaging at 3 and 7 T using well matched scanner configurations

Signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) for magnetic resonance microimaging were measured using two nearly identical magnetic resonance imaging (MRI) scanners operating at field strengths of 3 and 7 T. Six mice were scanned using two imaging protocols commonly applied for in vivo imaging of small animal brain: RARE and FLASH. An accounting was made of the field dependence of relaxation times as well as a small number of hardware disparities between scanner systems. Standard methods for relaxometry were utilized to measure T1 and T2 for two white matter (WM) and two gray matter (GM) regions in the mouse brain. An average increase in T1 between 3 and 7 T of 28% was observed in the brain. T2 was found to decrease by 27% at 7 T in agreement with theoretical models. The SNR was found to be uniform throughout the mouse brain, increasing at higher field by a factor statistically indistinguishable from the ratio of Larmor frequencies when imaging with either method. The CNR between GM and WM structures was found to adhere to the expected field dependence for the RARE imaging sequence. Improvement in the CNR for the FLASH imaging sequence between 3 and 7 T was observed to be greater than the Larmor ratio, reflecting a greater susceptibility to partial volume effects at the lower SNR values at 3 T. Imaging at 7 T versus 3 T in small animals clearly provides advantages with respect to the CNR, even beyond the Larmor ratio, especially in lower SNR regimes. This careful multifaceted assessment of the benefits of higher static field is instructive for those newly embarking on small animal imaging. Currently the number of 7 T MRI scanners in use for research in human subjects is increasing at a rapid pace with approximately 30 systems deployed worldwide in 2008. The data presented in this article verify that if system performance and radio frequency uniformity is optimized at 7 T, it should be possible to realize the expected improvements in the CNR and SNR compared with MRI at 3 T.     

Aging-related changes of neural mechanisms underlying visual-spatial working memory

Capacities of the prefrontal cortex (PFC) such as working memory (WM) are known to decline with increasing age. However, it is unclear which neurofunctional mechanisms may underlie this aging-related cognitive decline. The finding that PFC activity tends to be less lateralized in older subjects has led to the assumption of a hemispheric asymmetry reduction in the PFC associated with aging (HAROLD). Using functional magnetic resonance imaging (fMRI), we here investigated aging-related neurofunctional alterations during the performance of a visual-spatial WM task with differential levels of difficulty. Older volunteers activated dorsolateral PFC regions bilaterally while young subjects recruited these areas only in the left hemisphere. These data corroborate the hemispheric asymmetry reduction in the PFC associated with aging (HAROLD) account. However, we also observed functional reorganizations in parieto-occipital areas, and with increasing WM demands, an aging-related reversed hemispheric asymmetry of prefrontal activations. Importantly, neither PFC nor parieto-occipital reorganizations prevented older participants from showing worse WM performance than young volunteers. We conclude that frontal-parietal functional reorganizations may reflect compensational mechanisms related to aging, but do not obviate diminished visual-spatial WM performance in older people.     

Regional homogeneity of fMRI time series in autism spectrum disorders

Functional magnetic resonance imaging (fMRI) and functional connectivity MRI (fcMRI) studies of autism spectrum disorders (ASD) have suggested atypical patterns of activation and long-distance connectivity for diverse tasks and networks in ASD. We explored the regional homogeneity (ReHo) approach in ASD, which is analogous to conventional fcMRI, but focuses on local connectivity. FMRI data of 26 children with ASD and 29 typically developing (TD) children were acquired during continuous task performance (visual search). Effects of motion and task were removed and Kendall's coefficient of concordance (KCC) was computed, based on the correlation of the blood oxygen level dependent (BOLD) time series for each voxel and its six nearest neighbors. ReHo was lower in the ASD than the TD group in superior parietal and anterior prefrontal regions. Inverse effects of greater ReHo in the ASD group were detected in lateral and medial temporal regions, predominantly in the right hemisphere. Our findings suggest that ReHo is a sensitive measure for detecting cortical abnormalities in autism. However, impact of methodological factors (such as spatial resolution) on ReHo require further investigation.     

A discriminative model-constrained EM approach to 3D MRI brain tissue classification and intensity non-uniformity correction

We describe a fully automated method for tissue classification, which is the segmentation into cerebral gray matter (GM), cerebral white matter (WM), and cerebral spinal fluid (CSF), and intensity non-uniformity (INU) correction in brain magnetic resonance imaging (MRI) volumes. It combines supervised MRI modality-specific discriminative modeling and unsupervised statistical expectation maximization (EM) segmentation into an integrated Bayesian framework. While both the parametric observation models and the non-parametrically modeled INUs are estimated via EM during segmentation itself, a Markov random field (MRF) prior model regularizes segmentation and parameter estimation. Firstly, the regularization takes into account knowledge about spatial and appearance-related homogeneity of segments in terms of pairwise clique potentials of adjacent voxels. Secondly and more importantly, patient-specific knowledge about the global spatial distribution of brain tissue is incorporated into the segmentation process via unary clique potentials. They are based on a strong discriminative model provided by a probabilistic boosting tree (PBT) for classifying image voxels. It relies on the surrounding context and alignment-based features derived from a probabilistic anatomical atlas. The context considered is encoded by 3D Haar-like features of reduced INU sensitivity. Alignment is carried out fully automatically by means of an affine registration algorithm minimizing cross-correlation. Both types of features do not immediately use the observed intensities provided by the MRI modality but instead rely on specifically transformed features, which are less sensitive to MRI artifacts. Detailed quantitative evaluations on standard phantom scans and standard real-world data show the accuracy and robustness of the proposed method. They also demonstrate relative superiority in comparison to other state-of-the-art approaches to this kind of computational task: our method achieves average Dice coefficients of 0.93 ± 0.03 (WM) and 0.90 ± 0.05 (GM) on simulated mono-spectral and 0.94 ± 0.02 (WM) and 0.92 ± 0.04 (GM) on simulated multi-spectral data from the BrainWeb repository. The scores are 0.81 ± 0.09 (WM) and 0.82 ± 0.06 (GM) and 0.87 ± 0.05 (WM) and 0.83 ± 0.12 (GM) for the two collections of real-world data sets-consisting of 20 and 18 volumes, respectively-provided by the Internet Brain Segmentation Repository.     

磁共振脑功能成像信号——噪声及分析处理研究

随着磁共振成像技术的发展,其在人脑功能研究中的应用越来越广泛。深入了解磁共振成像信号的产生与采集过程对于数据的获取和分析及取得所需结果有很大的帮助。首先介绍了磁共振成像信号及其噪声,在其基础上对实验数据的分析进行讨论,并介绍了脑功能磁共振成像数据分析的新进展。