Elsevier

Neurocomputing

Volume 269, 20 December 2017, Pages 199-205
Neurocomputing

Whole-brain functional connectome-based multivariate classification of post-stroke aphasia

https://doi.org/10.1016/j.neucom.2016.10.094Get rights and content

Highlights

  • Accuracy of classification of patients with PAS and controls reached to 86.5%.

  • Consensus connections were located in fronto-parietal and auditory networks.

  • Region with the highest weight of classification was the right rolandic operculum.

Abstract

Patients with post-stroke aphasia (PSA) show abnormalities of intrinsic functional connectivity. However, whether the whole-brain functional connectome can be used as a feature to distinguish patients with PSA from healthy controls is poorly understood. We aim to distinguish PSA patients from controls using whole-brain functional connectivity-based multivariate pattern analysis. These features would be helpful in the understanding of the pathophysiology of PSA. In the present study, resting-state functional magnetic resonance images (fMRI) were acquired in 17 patients with PSA and 20 age- and sex-matched healthy controls. We used functional connectivity pattern and linear support vector machine to classify two groups. The results showed that the accuracy of classification reached to 86.5%, sensitivity reached to 76.5%, and specificity reached to 95.0%. In addition, consensus connections were mainly located in the fronto-parietal, auditory, sensory-motor, and visual networks. Furthermore, the right rolandic operculum contributed the highest weight. We suggest that whole-brain functional connectivity could be used as a potential neuromarker to distinguish PSA patients from controls.

Introduction

Aphasia, mostly caused by left hemisphere lesions, results in disturbances in the comprehension and formulation of language [1]. Post-stroke aphasia (PSA) is related to higher mortality, higher costs, and occupational problems [2]. Although there are a number of structural and functional researches exploring the language architecture in PSA [3–5], its pathophysiology and potential clinical application remains unknown.

Resting-state functional magnetic resonance imaging (fMRI) techniques have become a focus in clinical studies [6]. The use of local brain activity and remote connectivity-based resting-state fMRI can help us understand the pathophysiology of PSA in detail. Remote functional connectivity has been used to investigate the underlying mechanism of language processing in aphasia [4], [7], [8], [9]. It was found that patients with PSA showed alterations of resting-state functional connectivity in the default-mode network [10] and the frontoparietal network [11]. Moreover, changes in local brain activity and associated remote functional connectivity patterns were examined to enhance the understanding of the pathophysiological mechanisms of aphasia [3].

Multivariate pattern analysis (MVPA) is a data-driven technique that is useful in distinguishing patients from controls [12]. The method plays a critical role in understanding neuroimaging data and is able to identify core features contributing to the classification or prediction of PSA [6], [13], [14]. Brain connectome-based multivariate classifications have been reliably identified for patients and controls using individual-level information [15]. Furthermore, whole-brain functional connectivity can be used as potential feature to distinguish patients from controls [16], [17], [18], which would be helpful in the understanding of the pathophysiology of PSA.

To this end, we utilized the MVPA method based on whole-brain functional connectivity to distinguish PSA patients from controls. We hypothesized that whole-brain functional connectivity could be used as feature to classify successfully. This study may contribute to the understanding of the pathophysiology and clinical application of PSA.

Section snippets

Subjects

Seventeen patients with PSA (all right-handed, six females and 11 males, aged, 53.53 ± 14.06 years) were recruited from Fuzhou Hospital, China. A subset of these patients had participated in one of our earlier studies [3–5]. Patients were recruited according to the following criteria: (i) first stroke occurred in the left hemisphere; (ii) age of > 18 and < 85 years; (iii) native Chinese speaker; (iv) aphasia persistent at day 1 post-stroke; and (v) right-handed. Participants were excluded if they

Demographics and clinical characteristics of participants

Patients with PSA and HCs did not significantly differ in age (two-sample t-test, P = 0.98), gender (χ2-test, P = 0.90), or years of education (Mann–Whitney U test, P = 0.58; see Table 1 in [3]). Stroke-related clinical characteristics of patients were tested using the ABC [31], [32]. The ABC provides the following information: aphasia quotient, which includes spontaneous speech, auditory comprehension, repetition, and naming scores; performance quotient, which includes reading/writing, praxis, and

Discussion

To discriminate patients with PSA from healthy controls, we combined linear SVM and LOOCV methods using whole-brain functional connectome. We found that consensus connections were distributed in FPN, SMN, AN, and VN, and the highest weight of classification was the right ROL.

Our findings showed that abnormal functional connections were associated within the FPN, AN, and SMN. Most regions in these networks are involved in language processing (e.g., MTG and STG) [37]. The dorsal pathway lies

Conclusion

In sum, we combined linear SVM and whole-brain functional connectome-based classification to distinguish PSA patients from controls with high accuracy. Altered functional connections were identified in the SMN, FPN, AN, and VN. We suggest that whole-brain functional connectivity could be used as a feature to explore the pathophysiology in PSA patients.

Declaration of interest

All authors disclosed no relevant conflict of interest.

Acknowledgments

This work was supported by the 863 project (2015AA020505), Natural Science Foundation of China (61533006, and 81471653), China Postdoctoral Science Foundation (2013M532229), Fundamental Research Funds for the Central Universities (ZYGX2013Z004), and Sichuan provincial health and family planning commission research project (16PJ051).

Mi Yang, Dental, Deputy director of the physician, Department of Stomatology, the Fourth people's Hospital of Chengdu. She is also the Ph.D. student of Biomedical Engineering at University of Electronic Science Technology of China, Her main research direction includes: Multi- model imaging method and application in Post-stroke Aphasia and Orthodontic tooth. She have published over 10 articles including scientific reports, Brain topography and Medicine etc.

References (57)

  • F. Geranmayeh et al.

    Semantic retrieval during overt picture description: left anterior temporal or the parietal lobe?

    Neuropsychologia

    (2015)
  • D. Lopez-Barroso et al.

    Multiple brain networks underpinning word learning from fluent speech revealed by independent component analysis

    NeuroImage

    (2015)
  • R. Dalla Volta et al.

    Spatiotemporal dynamics during processing of abstract and concrete verbs: an ERP study

    Neuropsychologia

    (2014)
  • A. Moriai-Izawa et al.

    Multichannel fNIRS assessment of overt and covert confrontation naming

    Brain Lang.

    (2012)
  • H. Duffau et al.

    A re-examination of neural basis of language processing: proposal of a dynamic hodotopical model from data provided by brain stimulation mapping during picture naming

    Brain Lang.

    (2014)
  • A.E. Hillis et al.

    Reperfusion of specific brain regions by raising blood pressure restores selective language functions in subacute stroke

    Brain Lang.

    (2001)
  • M.V. Bonini et al.

    Cognitive deficits in post-stroke aphasia

    Arq. Neuro Psiquiatria

    (2015)
  • YangM. et al.

    Altered intrinsic regional activity and interregional functional connectivity in post-stroke aphasia

    Sci. Rep.

    (2016)
  • YangM. et al.

    Disrupted intrinsic local synchronization in poststroke aphasia

    Medicine (Baltimore)

    (2016)
  • YangM. et al.

    Beyond the arcuate fasciculus: damage to ventral and dorsal language pathways in aphasia

    Brain Topogr.

    (2017)
  • LiuF. et al.

    Multivariate classification of social anxiety disorder using whole brain functional connectivity

    Brain Struct. Funct.

    (2015)
  • S. van Hees et al.

    A functional MRI study of the relationship between naming treatment outcomes and resting state functional connectivity in post-stroke aphasia

    Hum. Brain Mapp.

    (2014)
  • S. van Hees et al.

    Changes in white matter connectivity following therapy for anomia post stroke

    Neurorehabil. Neural Repair

    (2014)
  • V.A. Nair et al.

    Functional connectivity changes in the language network during stroke recovery

    Ann. Clin. Transl. Neurol.

    (2015)
  • WangX. et al.

    Resting state brain default network in patients with motor aphasia resulting from cerebral infarction

    Chin. Sci. Bull.

    (2014)
  • ZhuD. et al.

    Changes of functional connectivity in the left frontoparietal network following aphasic stroke

    Front. Behav. Neurosci.

    (2014)
  • J.S. Anderson et al.

    Functional connectivity magnetic resonance imaging classification of autism

    Brain J. Neurol.

    (2011)
  • SuL. et al.

    Influence of Resting-state network on lateralization of functional connectivity in mesial temporal lobe epilepsy

    Am. J. Neuroradiol.

    (2015)
  • Cited by (0)

    Mi Yang, Dental, Deputy director of the physician, Department of Stomatology, the Fourth people's Hospital of Chengdu. She is also the Ph.D. student of Biomedical Engineering at University of Electronic Science Technology of China, Her main research direction includes: Multi- model imaging method and application in Post-stroke Aphasia and Orthodontic tooth. She have published over 10 articles including scientific reports, Brain topography and Medicine etc.

    Jiao Li received the B.S. degree from University of Electronic Science and Technology of China (UESTC), in 2015. She is currently working toward the M.S. degree in UESTC. Her main research interest is data analysis of functional magnetic resonance imaging.

    Zhiqiang Li received the B.S. and the M.S. degrees from University Electronic of Science Technology of China (UESTC), in 2013 and 2016, respectively. Her main research direction includes: Brain Imaging method, fMRI data analysis and pattern recognition.

    Dezhong Yao, Ph.D., Professor of Biomedical Engineering at University of Electronic Science Technology of China, fellow of Natural Science Foundation for Distinguished Young Scholars, and Distinguished Professor of changjiang river scholars. Prof. Chen devotes himself to the research of EEG and fMRI technique and cognitive applications; Brain-machine interface, Multi- model imaging method and application in Nervous disease and mental disease etc. He have published over 150 SCI articles including NeuroImage, Human Brian Mapping, IEEE Trans BME etc.

    Wei Liao received the B.S., M.S. and Ph.D. degrees from the University of Electronic Science and Technology of China (UESTC), Chengdu, China, in 2004, 2008, and 2011 respectively. He is currently a Professor at the School of Life Science and Technology, UESTC. His research interests include functional magnetic resonance imaging and image processing.

    Huafu Chen, Ph.D., Professor of Biomedical Engineering at University of Electronic Science Technology of China (UESTC), fellow of Natural Science Foundation for Distinguished Young Scholars, and Distinguished Professor of Changjiang River scholars. Prof. Chen devotes himself to the research of brain pattern recognition technique and cognitive applications; Dynamic Brain network; Multi-model imaging method and application in Nervous disease and mental disease etc. He have published over 120 SCI articles including Brain, NeuroImage, Human Brian Mapping, IEEE Trans MI/BME etc.

    View full text