Qi Zhou

An optimal decisional space for the classification of alzheimer's disease and mild cognitive impairment

This paper proposes to combine MRI data with a neuropsychological test, mini-mental state examination (MMSE), as input to a multi-dimensional space for the classification of Alzheimers disease (AD) and its prodromal stages-mild cognitive impairment (MCI) including amnestic MCI (aMCI) and nonamnestic MCI (naMCI). The decisional space is constructed using those features deemed statistically significant through an elaborate feature selection and ranking mechanism. FreeSurfer was used to calculate 55 volumetric variables, which were then adjusted for intracranial volume, age and education. The classification results obtained using support vector machines are based on twofold cross validation of 50 independent and randomized runs. The study included 59 AD, 67 aMCI, 56 naMCI, and 127 cognitively normal (CN) subjects. The study shows that MMSE scores contain the most discriminative power of AD, aMCI, and naMCI. For AD versus CN, the two most discriminative volumetric variables (right hippocampus and left inferior lateral ventricle), when combined with MMSE scores, provided an average accuracy of 92.4% (sensitivity: 84.0%; specificity: 96.1%). MMSE scores are found to improve all classifications with accuracy increments of 8.2% and 12% for aMCI versus CN and naMCI versus CN, respectively. Results also show that brain atrophy is almost evenly seen on both sides of the brain for AD subjects, which is different from right-side dominance for aMCI and left-side dominance for naMCI. Furthermore, hippocampal atrophy is seen to be the most significant for aMCI, while Accumbens area and ventricle are most significant for naMCI.

Significance of Normalization on Anatomical MRI Measures in Predicting Alzheimer’s Disease

This study establishes a new approach for combining neuroimaging and neuropsychological measures for an optimal decisional space to classify subjects with Alzheimers disease (AD). This approach relies on a multivariate feature selection method with different MRI normalization techniques. Subcortical volume, cortical thickness, and surface area measures are obtained using MRIs from 189 participants (129 normal controls and 60 AD patients). Statistically significant variables were selected for each combination model to construct a multidimensional space for classification. Different normalization approaches were explored to gauge the effect on classification performance using a support vector machine classifier. Results indicate that the Mini-mental state examination (MMSE) measure is most discriminative among single-measure models, while subcortical volume combined with MMSE is the most effective multivariate model for AD classification. The study demonstrates that subcortical volumes need not be normalized, whereas cortical thickness should be normalized either by intracranial volume or mean thickness, and surface area is a weak indicator of AD with and without normalization. On the significant brain regions, a nearly perfect symmetry is observed for subcortical volumes and cortical thickness, and a significant reduction in thickness is particularly seen in the temporal lobe, which is associated with brain deficits characterizing AD.

Inclusion of neuropsychological scores in atrophy models improves diagnostic classification of alzheimer’s disease and mild cognitive impairment

Brain atrophy in mild cognitive impairment (MCI) and Alzheimers disease (AD) are difficult to demarcate to assess the progression of AD. This study presents a statistical framework on the basis of MRI volumes and neuropsychological scores. A feature selection technique using backward stepwise linear regression together with linear discriminant analysis is designed to classify cognitive normal (CN) subjects, early MCI (EMCI), late MCI (LMCI), and AD subjects in an exhaustive two-group classification process. Results show a dominance of the neuropsychological parameters like MMSE and RAVLT. Cortical volumetric measures of the temporal, parietal, and cingulate regions are found to be significant classification factors. Moreover, an asymmetrical distribution of the volumetric measures across hemispheres is seen for CN versus EMCI and EMCI versus AD, showing dominance of the right hemisphere; whereas CN versus LMCI and EMCI versus LMCI show dominance of the left hemisphere. A 2-fold cross-validation showed an average accuracy of 93.9%, 90.8%, and 94.5%, for the CN versus AD, CN versus LMCI, and EMCI versus AD, respectively. The accuracy for groups that are difficult to differentiate like EMCI versus LMCI was 73.6%. With the inclusion of the neuropsychological scores, a significant improvement (24.59%) was obtained over using MRI measures alone.

A practical guideline for intracranial volume estimation in patients with Alzheimer's disease

BackgroundIntracranial volume (ICV) is an important normalization measure used in morphometric analyses to correct for head size in studies of Alzheimer Disease (AD). Inaccurate ICV estimation could introduce bias in the outcome. The current study provides a decision aid in defining protocols for ICV estimation in patients with Alzheimer disease in terms of sampling frequencies that can be optimally used on the volumetric MRI data, and the type of software most suitable for use in estimating the ICV measure.MethodsTwo groups of 22 subjects are considered, including adult controls (AC) and patients with Alzheimer Disease (AD). Reference measurements were calculated for each subject by manually tracing intracranial cavity by the means of visual inspection. The reliability of reference measurements were assured through intra- and inter- variation analyses. Three publicly well-known software packages (Freesurfer, FSL, and SPM) were examined in their ability to automatically estimate ICV across the groups.ResultsAnalysis of the results supported the significant effect of estimation method, gender, cognitive condition of the subject and the interaction among method and cognitive condition factors in the measured ICV. Results on sub-sampling studies with a 95% confidence showed that in order to keep the accuracy of the interleaved slice sampling protocol above 99%, the sampling period cannot exceed 20 millimeters for AC and 15 millimeters for AD. Freesurfer showed promising estimates for both adult groups. However SPM showed more consistency in its ICV estimation over the different phases of the study.ConclusionsThis study emphasized the importance in selecting the appropriate protocol, the choice of the sampling period in the manual estimation of ICV and selection of suitable software for the automated estimation of ICV. The current study serves as an initial framework for establishing an appropriate protocol in both manual and automatic ICV estimations with different subject populations.

Insights into cognitive aging and Alzheimer’s disease using amyloid PET and structural MRI scans

Positron emission tomography using amyloid binding ligands, labeled with carbon-11 and fluorine-18, (Amyloid PET) has been used to understand the relationship between amyloid deposition in the brain, neurodegeneration and the development of mild cognitive impairment and dementia. Structural MRI has been used to identify morphological changes in the brain which may relate to the cause(s) of cognitive impairment, including infarcts, space-occupying lesions, hydrocephalus and the patterns of atrophy which are characteristic of various neurodegenerative diseases. These two imaging biomarkers have also played an important role in revealing the sequence of cerebral amyloid deposition, neurodegeneration and cognitive impairment. Although there may not be a direct relationship between amyloid deposition and brain atrophy or cognitive deficits, the presence of both amyloid deposition on PET and neurodegeneration on MRI has been associated with accelerated cognitive decline. The main focus of this article is to summarize some of the insights gained using these two imaging methods individually and in combination to better understand the biological bases of normal aging and age-associated cognitive impairment.

Apolipoprotein-E4 (ApoE4) carriers show altered small-world properties in the default mode network of the brain

Apolipoprotein-E (ApoE), specifically the e4 allele, has been identified as a risk factor for Alzheimers disease (AD). The prevalence of the gene in 25% to 30% of the population necessitates careful examination of its role in neuroregulation and its impact on distributed brain networks. In this study, large-scale and small-world anatomical cortical networks are assessed in cognitive normal (CN) subjects with differing apolipoprotein-E4 (ApoE4) gene expression. The large-scale anatomical networks are estimated from cortical thickness measurements derived from magnetic resonance imaging in 147 CN subjects examined in relation to the ApoE4 genotype (e4+ carriers (n = 41) versus e4− non-carriers (n = 106)). Brain networks were constructed by thresholding anatomical cortical thickness correlation matrices of 68 bilateral brain regions analyzed using well-established graph theoretical approaches. Compared to non-carriers, carriers showed increased interregional correlation coefficients in regions such as the precentral, superior frontal and inferior temporal lobes, and most of the altered connections were intra-hemispheric and limited primarily to the right hemisphere. The number of long-length anatomic connections in carriers (26.1%) was significantly higher compared to non-carriers (17.9%). Carriers showed a reduced amount of short connections in the frontal lobe as compared to non-carriers, but an increased presence of long-range connections was observed in the frontal lobe. Furthermore, ApoE4 carriers demonstrated abnormal small-world architecture in the cortical networks with increased clustering coefficient and path lengths as compared to non-carriers, suggesting a less optimal topological organization of their brain networks. Finally, as compared to non-carriers, ApoE4 carriers demonstrated higher betweenness in regions such as middle temporal, parahippocampal gyrus, posterior cingulate and insula of the default mode network, also seen in subjects with AD and mild cognitive impairment. The results suggest that the complex morphological cortical connectivity patterns are altered in ApoE4 carriers, providing evidence for disruption of integrity in large-scale anatomical brain networks in asymptomatic ApoE4 carriers.

Altered small-world anatomical networks in Apolipoprotein-E4 (ApoE4) carriers using MRI.

Apolipoprotein E (ApoE) gene and primarily its allele e4 have been identified as a risk factor for Alzheimers disease (AD). The prevalence of the gene in 25-30% in the population makes it essential to estimate its role in neuroregulation and its impact on distributed brain networks. In this study, we provide computational neuroanatomy based interpretation of large-scale and small-world cortical networks in cognitive normal (CN) subjects with differing Apolipoprotein-E4 (ApoE4) gene expression. We estimated large-scale anatomical networks from cortical thickness measurements derived from magnetic resonance imaging in 147 CN subjects explored in relation to ApoE4 genotype (e4+ carriers (n=41) versus e4- non-carriers (n=106)). Brain networks were constructed by thresholding cortical thickness correlation matrices of 68 bilateral regions of the brain analyzed using well-established graph theoretical approaches. Compared to ApoE4 non-carriers, carriers showed increased interregional correlation coefficients in regions like precentral, superior frontal and inferior temporal regions. Interestingly most of the altered connections were intra-hemispheric limited primarily to the right hemisphere. Furthermore, ApoE4 carriers demonstrated abnormal small-world architecture in the cortical networks with increased clustering coefficient and path lengths as compared to non-carrier, suggesting a less optimal topological organization. Additionally non-carriers demonstrated higher betweenness in regions such as middle temporal, para-hippocampal gyrus, posterior cingulate and insula of the default mode network (DMN), also seen in subjects with AD and mild cognitive impairment (MCI). The results suggest that the complex morphological cortical connectivity patterns are altered in ApoE4 carriers as compared to non-carriers, providing evidence for disruption of integrity in large-scale anatomical brain networks.

Multivariate Analysis of structural MRI and PET (FDG and 18F-AV-45) for Alzheimer's disease and its prodromal stages

A multivariate analysis method, orthogonal partial least squares to latent structures (OPLS), was used to discriminate Alzheimers disease (AD), early and late mild cognitive impairment (EMCI and LMCI) from cognitively normal control (CN) using MRI and PET measures. FreeSurfer 5.1 generated 271 MRI features including 49 subcortical volumes, 68 cortical volumes, 68 cortical thicknesses, 70 surface areas and 16 hippocampus subfields. Subjects with all aforementioned MRI measures passing quality control and valid Fludeoxyglucose (18F) (FDG) and Florbetapir (18F) PET scans were selected from ADNI database, resulting in a total of 524 participants (137 CN, 214 EMCI, 103 LMCI and 70 AD) for the study. Altogether 286 features including 15 significant PET uptake features (7 for FDG and 8 for AV-45) were utilized for OPLS analysis. Predictive power was evaluated by Q2(Y), a quantifier of the statistical significance for class separation. The results show that MRI features (Q2(Y) =0.645), and PET features (Q2(Y) = 0.636) has comparable predictive power in separating AD from CN, and MRI features are better predictor of LMCI (Q2(Y) = 0.282) than PET (Q2(Y) = 0.294). Combination of PET and MRI has the most predictive power for LMCI and AD with Q2(Y) of 0.294 and 0.721, respectively. While for EMCI, cortical thickness was found to be the best predictor with a Q2(Y) of 0.108, suggesting cortical thickness may be the first structural change ahead of others and should be prioritized in prediction of very mild cognitive impairment.

Discriminating regional functional networks in pediatric epilepsy

This study introduces a correlation-based threshold analysis on region-of-interest connectivity networks calculated from functional MRI and based on the auditory description decision task (ADDT) language paradigm. The correlation threshold using the proposed method is determined to ensure that optimal connectivity is obtained with a minimum number of edges. Empirical evaluations on patients with epilepsy and control groups demonstrate that such correlation thresholds lead to significantly different connectivity networks of left and right temporal lobe when comparing the control group with patients and in ascertaining these networks relative to the location of the seizure focus. Interestingly, such thresholds did not show similar trends when using the AAL 90 template for the whole brain, which emphasize the need for the proposed regional connectivity concept. Also, most of the demographic and clinical parameters, such as handedness, age, gender, and seizure onset duration are shown not to have any effect on these thresholds.

Regional MRI measures and neuropsychological test for multi-dimensional analysis in Alzheimer's disease

Regional brain atrophy is a typical structural symptom of Alzheimers disease (AD). Magnetic resonance imaging (MRI) scans capture brain structure with high resolution and are often processed with automated segmentation and parcellation algorithm (e.g. Freesurfer) to generate regional measures, like cortical volume, cortical thickness and surface area, which are widely used as inputs in classification algorithms. This study aims to find out which combination of MRI measures and neuropsychological test coupled with different normalization techniques can best predict AD using a proposed multivariate feature selection and classification method. A total of 189 subjects with 60 Alzheimers disease (AD) and 129 cognitively normal (CN) are included in this study. Freesurfer was used to obtain 34 cortical thickness measures and 35 surface area measures for each hemisphere and 55 regional volumes across the brain. Statistically significant variables selected for each model were used to construct the multi-dimensional space for further classification using a support vector machine (SVM) classifier. Different normalization approaches were explored to gauge if the classification performance could be improved. Results indicate neuropsychological test score contains the most discriminative information among single measure models, and out of the three MRI measures, cortical volume is a better predicator than the other two. Normalization approaches are seen not to enhance the performance much if any. Hierarchical model of neuropsychological test and cortical volumes without normalization yield the best classification accuracy for this study.