AUTHOR=Zhang Jiahao , Lu Haifeng , Zhu Lin , Ren Huixia , Dang Ge , Su Xiaolin , Lan Xiaoyong , Jiang Xin , Zhang Xu , Feng Jiansong , Shi Xue , Wang Taihong , Hu Xiping , Guo Yi 

TITLE=Classification of Cognitive Impairment and Healthy Controls Based on Transcranial Magnetic Stimulation Evoked Potentials

JOURNAL=Frontiers in Aging Neuroscience

VOLUME=Volume 13 - 2021

YEAR=2021

URL=https://www.frontiersin.org/journals/aging-neuroscience/articles/10.3389/fnagi.2021.804384

DOI=10.3389/fnagi.2021.804384

ISSN=1663-4365

ABSTRACT=Backgrounds: Nowadays, risks of Cognitive Impairment (CI) (highly suspected Alzheimer's disease in this study) threaten the quality of life for more older adults as population ageing. The emergence of Transcranial Magnetic Stimulation-Electroencephalogram (TMS-EEG) enables noninvasive neurophysiological investigation of the human cortex, which is potential for CI detection.
Objectives: The aim of this study is to explore the spatiotemporal features extraction method of TMS Evoked Potentials (TEPs), and classify CI and healthy controls through machine learning.
Methods: Twenty-one CI patients and twenty-two healthy controls underwent a single-pulse TMS-EEG protocol in which the pulses were delivered to the left dorsolateral prefrontal cortex (left DLPFC). After data preprocessing, seven regions of interest and two most reliable TEPs' components: N100 and P200 were selected. Next, seven simple and interpretable linear features of TEPs were extracted, three common machine learning algorithms including Support Vector Machine, Random Forest and K-Nearest Neighbor were used to detect CI. Meanwhile, data augmentation and voting strategy were used for more robust model. Finally, the performance differences of features in classifiers and their contribution were investigated.
Results: 1. In the time domain, the features of N100 had the best performance in the SVM classifier, with the accuracy of 88.37%. 2. In the aspect of spatiality, the features of the right frontal region and left parietal region had the best performance in the SVM classifier, with the accuracy of 83.72%. 3. The Local Mean Field Power, Average Value, Latency and Amplitude contributed most in classification.
Conclusions: The TEPs induced by TMS over the left DLPFC has significant differences between CI and healthy controls. The machine learning based on the spatiotemporal features of TEPs is effective in detecting CI. Further, TEPs' features has the potential to become non-invasive biomarkers for CI diagnosis.