Klepachevskyi, Dmytro

Introduction: Parkinson’s disease (PD) is a neurodegenerative disorder that affects motor function, causing tremor and impaired coordination. This study explores changes in neural oscillations in magnetoencephalography (MEG) data to identify potential biomarkers linked to depression, apathy, and cognitive impairment in PD patients. Methods: MEG data from 20 PD patients were preprocessed and source-reconstructed to 116 time-series. Power spectral density (PSD), its ratios, and spectral slopes were calculated in 5 bands: Alpha (8-12 Hz), Beta (12-30 Hz), Gamma (30-100 Hz), Delta (0.5-4 Hz), and Theta (4-8 Hz). Spearman’s correlation and odds ratios of binary logistic regression for PSD and slopes were calculated for MEG features and clinical measures: depression assessed by Beck Depression Inventory (BDI), apathy by apathy evaluation scale, and cognitive function by Montreal Cognitive Assessment (MoCA). Using Minimum Redundancy Maximum Relevance method to avoid overfitting, 6 most significant features were selected, and Machine Learning (ML) algorithms were applied to predict cognitive impairment and depression. Results: The strongest PSD Spearman’s correlation for BDI was in Alpha activity channel 3, corresponding to the left frontal superior orbital region (r=0.74, p<0.05), and in Alpha activity channel 42 on the right side (r=0.71, p<0.05). For apathy, Alpha and Beta bands in channel 52 (paracentral lobule right side) were most significant (Alpha r=0.64, p<0.05; Beta r=0.60, p<0.05). A negative correlation was found between gamma, alpha band slopes and MoCA scores (r= -0.66 to -0.45, p<0.05), and between PSD of various bands and MoCA scores (r= -0.63 to -0.44, p<0.05). The best predictive model was the logistic regression model for BDI with a 90%±9.24 accuracy. The most significant features were PSD power in Alpha band (channels 40, 46), Theta (channels 1, 2), Delta (channel 97), Gamma/Beta ratio (channel 3). ML methods showed a lower accuracy in prediction of apathy (50.86%) and MoCA levels (64.29%). Conclusion: The findings suggest that altered Alpha and Beta activity in the frontal superior orbital region is related to depressive symptoms, while activity in the paracentral lobule is linked to apathy in PD. Steeper gamma spectral slopes and elevated power in Alpha and Theta frequency bands may be markers of cognitive impairment. The ML models showed strong predictive capabilities, revealing the value of MEG data in identifying depressive symptoms in PD.