Open Access

Tauqeer Anjum, Ali Seyfizadeh, Hao Ding, Tahir Mahmood, Rüdiger Pryss, Jens Volkmann, Dumitru Ciolac, Muthuraman Muthuraman

• The propagation of neural signals across various brain regions requires us to understand directional connectivity in functional magnetic resonance imaging (fMRI) data. We employ temporal partial directed coherence (TPDC), a data driven method to explore directional connectivity in young and old healthy individuals, manifest PD and prodromal PD patients. TPDC provides comprehensive analysis of neural signal propagation compared to traditional methods like Dynamic Causal Modeling, Granger Causality and Transfer Entropy.Methods: We analyzed preprocessed fMRI data from the young and old groups of healthy individuals and PD patients at different disease stages. Time series were extracted by reducing the voxel data (by averaging) into 7 networks of the Yeo-atlas. TPDC was applied in the frequency range of 0.009–0.08 Hz. Statistical significance of connections was determined via bootstrapping, followed by thresholding using permutation testing. Finally, machine learning classifiers wereThe propagation of neural signals across various brain regions requires us to understand directional connectivity in functional magnetic resonance imaging (fMRI) data. We employ temporal partial directed coherence (TPDC), a data driven method to explore directional connectivity in young and old healthy individuals, manifest PD and prodromal PD patients. TPDC provides comprehensive analysis of neural signal propagation compared to traditional methods like Dynamic Causal Modeling, Granger Causality and Transfer Entropy.Methods: We analyzed preprocessed fMRI data from the young and old groups of healthy individuals and PD patients at different disease stages. Time series were extracted by reducing the voxel data (by averaging) into 7 networks of the Yeo-atlas. TPDC was applied in the frequency range of 0.009–0.08 Hz. Statistical significance of connections was determined via bootstrapping, followed by thresholding using permutation testing. Finally, machine learning classifiers were trained to distinguish prodromal PD from PD patients.Results: In young healthy individuals, the somatomotor network regulates control and attention systems, indicating cognitive and motor flexibility. Older healthy controls show lack of significant connections from control to somatomotor networks, suggesting a cognitive decline related to age. The somatomotor network becomes secluded in the prodromal PD patients. A compensatory mechanism is visible in groups of PD patients. Additionally, machine learning classifiers achieved high accuracy in distinguishing between prodromal and PD groups based on directed connectivity patterns.Conclusion: The study highlights the gradual loss of the significant directed causal connections between the control and motor networks in different stages of PD. The governing influence of control network over the motor and attentional networks diminishes, leading to the isolation of the somatomotor network. The ability of TPDC-derived features to distinguish prodromal from Parkinson’s patients underscores its value for identifying potential biomarkers of disease onset and progression.…