A support vector machine-based method to identify mild cognitive impairment with multi-level characteristics of magnetic resonance imaging

Long, Zhuqing; Jing, Bin; H, Yan; Dong, J Z; Liu, Han; Mo, Xiao; Han, Ying; Li, Haiyun

doi:10.1016/j.neuroscience.2016.06.025

Cited by 49 publications

(25 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The systematic literature search yielded 305 titles and abstracts. Of these results, we identified 10 studies with 11 datasets suitable for final inclusion (378 patients, 435 controls) [ 10 , 12 – 14 , 27 – 32 ]. The detailed study selection process for the meta-analysis is shown in the flowchart Fig.…”

Section: Resultsmentioning

confidence: 99%

“…Over the past decade, many rs-fMRI studies have reported aberrant ReHo in patients with aMCI relative to healthy controls. ReHo can be applied to distinguish patients with aMCI from healthy controls with an accuracy of 90.32% (sensitivity 86.21% and specificity 93.94%) by employing a support vector machine-based approach [ 10 ]. In addition, ReHo alterations have been observed to correlate with cognitive and memory impairment in patients with aMCI [ 11 – 14 ].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Alterations of brain local functional connectivity in amnestic mild cognitive impairment

Zhen

Xia

et al. 2018

Transl Neurodegener

View full text Add to dashboard Cite

BackgroundResting-state functional magnetic resonance imaging studies using a regional homogeneity (ReHo) method have reported that amnestic mild cognitive impairment (aMCI) was associated with abnormalities in local functional connectivity. However, their results were not conclusive.MethodsSeed-based d Mapping was used to conduct a coordinate-based meta-analysis to identify consistent ReHo alterations in aMCI.ResultsWe identified 10 studies with 11 datasets suitable for inclusion, including 378 patients with aMCI and 435 healthy controls. This meta-analysis identified significant ReHo alterations in patients with aMCI relative to healthy controls, mainly within the default mode network (DMN) (bilateral posterior cingulate cortex [PCC], right angular gyrus, bilateral middle temporal gyri, and left parahippocampal gyrus/hippocampus), executive control network (right superior parietal lobule and dorsolateral prefrontal cortex), visual network (right lingual gyrus and left middle occipital gyrus), and sensorimotor network (right paracentral lobule/supplementary motor area, right postcentral gyrus and left posterior insula). Significant heterogeneity of ReHo alterations in the bilateral PCC, left parahippocampal gyrus/hippocampus, and right superior parietal lobule/angular gyrus was observed. Exploratory meta-regression analyses indicated that general cognitive function, gender distribution, age, and education level partially contributed to this heterogeneity.ConclusionsThis study provides provisional evidence that aMCI is associated with abnormal ReHo within the DMN, executive control network, visual network, and sensorimotor network. These local functional connectivity alterations suggest coexistence of functional deficits and compensation in these networks. These findings contribute to the modeling of brain functional connectomes and to a better understanding of the neural substrates of aMCI. Confounding factors merit much attention and warrant future investigations.Electronic supplementary materialThe online version of this article (10.1186/s40035-018-0134-8) contains supplementary material, which is available to authorized users.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Alterations of brain local functional connectivity in amnestic mild cognitive impairment

Zhen

Xia

et al. 2018

Transl Neurodegener

View full text Add to dashboard Cite

show abstract

“…These metrics exhibit high test-retest reliably (Li et al, 2012; Zuo et al, 2013), rendering them advantageous for clinical imaging. Both measures can reveal aberrant activity underlying cognitive changes, since they relate to cognitive abilities in health and disease (Dai et al, 2012; Tian et al, 2012; Zhang et al, 2012; Zou et al, 2013; Long et al, 2016; Ren et al, 2016). …”

Section: Introductionmentioning

confidence: 99%

Aberrant Intrinsic Activity and Connectivity in Cognitively Normal Parkinson’s Disease

Harrington

Shen

Castillo

et al. 2017

Front. Aging Neurosci.

View full text Add to dashboard Cite

Disturbances in intrinsic activity during resting-state functional MRI (rsfMRI) are common in Parkinson’s disease (PD), but have largely been studied in a priori defined subnetworks. The cognitive significance of abnormal intrinsic activity is also poorly understood, as are abnormalities that precede the onset of mild cognitive impairment. To address these limitations, we leveraged three different analytic approaches to identify disturbances in rsfMRI metrics in 31 cognitively normal PD patients (PD-CN) and 30 healthy adults. Subjects were screened for mild cognitive impairment using the Movement Disorders Society Task Force Level II criteria. Whole-brain data-driven analytic approaches first analyzed the amplitude of low-frequency intrinsic fluctuations (ALFF) and regional homogeneity (ReHo), a measure of local connectivity amongst functionally similar regions. We then examined if regional disturbances in these metrics altered functional connectivity with other brain regions. We also investigated if abnormal rsfMRI metrics in PD-CN were related to brain atrophy and executive, visual organization, and episodic memory functioning. The results revealed abnormally increased and decreased ALFF and ReHo in PD-CN patients within the default mode network (posterior cingulate, inferior parietal cortex, parahippocampus, entorhinal cortex), sensorimotor cortex (primary motor, pre/post-central gyrus), basal ganglia (putamen, caudate), and posterior cerebellar lobule VII, which mediates cognition. For default mode network regions, we also observed a compound profile of altered ALFF and ReHo. Most regional disturbances in ALFF and ReHo were associated with strengthened long-range interactions in PD-CN, notably with regions in different networks. Stronger long-range functional connectivity in PD-CN was also partly expanded to connections that were outside the networks of the control group. Abnormally increased activity and functional connectivity appeared to have a pathological, rather than compensatory influence on cognitive abilities tested in this study. Receiver operating curve analyses demonstrated excellent sensitivity (≥90%) of rsfMRI variables in distinguishing patients from controls, but poor accuracy for brain volume and cognitive variables. Altogether these results provide new insights into the topology, cognitive relevance, and sensitivity of aberrant intrinsic activity and connectivity that precedes clinically significant cognitive impairment. Longitudinal studies are needed to determine if these neurocognitive associations presage the development of future mild cognitive impairment or dementia.

show abstract

“…A grid search algorithm was used to optimize the two parameters of SVM: γ, width of the RBF, and C, an input parameter for the SVM algorithm. The detailed description of the application of RBF kernel SVM in MRI data has been introduced in previous studies [ 21 – 22 ]. During the training phase, the SVM uses data was categorized into groups to determine the largest margin “hyperplane” to optimally separate the two groups.…”

Section: Methodsmentioning

confidence: 99%

Resting-state functional connectivity of the dorsal frontal cortex predicts subcortical vascular cognition impairment

Zhou

Zhang

et al. 2017

Oncotarget

View full text Add to dashboard Cite

Functional magnetic resonance imaging (fMRI) studies have revealed group differences in the frontal area between the subcortical vascular cognition impairment (SVCI) patients and the controls. However, most of the existing research focused on average differences between the two groups, and therefore had limited clinical applicability. The aim of our study was to investigate whether inter-regions functional connectivity of the dorsal frontal cortex (DFC) can be used to discriminate the SVCI from the controls at the level of the individual. Thirty-two SVCI patients and 32 demographically similar healthy individuals underwent resting-state functional magnetic resonance imaging. The DFC, derived from a prior atlas, was divided into 10 clusters. Features based on DFC were obtained through functional connectivity analysis between pairs of DFC. A nonlinear kernel support vector machine was used for classification and validated using 8-fold cross validation. An excellent classification accuracy was obtained from both the left and the right DFC functional connectivity (accuracy=75.07%, sensitivity=81.57% and specificity=61.71%; accuracy=45.38%, sensitivity=60.74% and specificity=39.48%; P<0.001). These findings shed further light on the pathogenesis of SVCI and showed promising classification performance using machine learning analysis based on DFC fMRI data, which may be useful for the differentiation of SVCI.

show abstract

A support vector machine-based method to identify mild cognitive impairment with multi-level characteristics of magnetic resonance imaging

Cited by 49 publications

References 39 publications

Alterations of brain local functional connectivity in amnestic mild cognitive impairment

Alterations of brain local functional connectivity in amnestic mild cognitive impairment

Aberrant Intrinsic Activity and Connectivity in Cognitively Normal Parkinson’s Disease

Resting-state functional connectivity of the dorsal frontal cortex predicts subcortical vascular cognition impairment

Contact Info

Product

Resources

About