Search for: All records

Transcranial magnetic stimulation (TMS) is gaining increasing attention for therapeutic treatment of mental illnesses. However, a clear understanding of its impact to the underlying brain mechanisms is critical for its effective application. For this, we analyze electroencephalography (EEG) response to TMS subthreshold pulse at the left motor cortex from 6 healthy controls and 6 schizophrenia patients. We use principal component analysis (PCA) along sparse nonnegative matrix factorization (NMF), an unsupervised machine learning technique, on brain connectivity data established by sliding window coherence of EEG based source localized data. The source localization was achieved by using the sLORETA algorithm on our EEG data after artifact removal. This, hence, provides high temporal and spatial resolution in the connectivity analysis results, giving advantage over other neuroimaging modalities. PCA aids in establishing the number of common underlying connected subnetworks (say k) across subjects whereas NMF is employed to derive these k spatial and temporal signature subnetwork response to the stimulus. Within these signatures, we studied motor cortical connectivity and found that schizophrenia patients exhibited sensory gating deficits as compared to controls. These findings can act as potential biomarkers to monitor TMS for clinical therapeutic techniques in the future. 

Studying electroencephalography (EEG) in response to transcranial magnetic stimulation (TMS) is gaining popularity for investigating the dynamics of complex neural architecture in the brain. For example, the primary motor cortex (M1) executes voluntary movements by complex connections with other associated subnetworks. To understand these connections better, we analyzed EEG signal response to TMS at left M1 from schizophrenia patients and healthy controls and in contrast with resting state EEG recording. After removing artifacts from EEG, we conducted 2D to 3D sLORETA conversion, a well-established source localization method, for estimating signal strength of 68 source dipoles or cortical regions inside the brain. Next, we studied dynamic connectivity by computing time-evolving spatial coherence of 2278 (=68*(68-1)/2) pairs of cortical regions, with sliding window technique of 200ms window size and 20ms shift over 1sec long data. Pairs with consistent coherence (coherence>0.8 during 200+ sliding windows of patients and controls combined) were chosen for identifying stable networks. For example, we found that during the resting state, precuneus was steadily coherent with middle and superior temporal gyrus in the left hemisphere in both patient and controls. Their connectivity pattern over the sliding windows significantly differed between patients and controls (pvalue<0.05). Whereas for M1, the same was true for two other coherent pairs namely, superamarginal gyrus with lateral occipital gyrus in right hemisphere and medial orbitofrontal gyrus with fusiform in left hemisphere. The TMS-EEG dynamic connectivity results can help to differentiate patient and normal subjects and also help to better understand the brain architecture and mechanisms.