| 摘要: | 帕金森氏症（Parkinson’s disease, PD）是一種神經退化性疾病，影響了中腦和腦幹的色素神經核，導致錐體外系症狀。除了動作障礙症狀外，帕金森氏症也會引起認知問題，可能發生在晚期階段或是與動作障礙症狀同時出現。其主要病理變化涉及皮層下結構，以及基底神經節、丘腦與額葉皮層之間的多巴胺能有強的連結，當這些連結受損時，會導致帕金森氏症之失智症（Parkinson′s disease dementia, PDD），而影響其執行功能、注意力和記憶力。腦電圖（electroencephalogram, EEG）能夠以高時間精確度測量大腦活動，但傳統時間-頻率分析中的線性計算在捕捉非線性和非穩態的神經活動過程時存在局限性。因此，我們採用創新的全息希爾伯特頻譜分析（Holo-Hilbert Spectral Analysis, HHSA）和希爾伯特-黃變換（Hilbert-Huang Transform, HHT），並結合集成經驗模態分解（Ensemble Empirical Mode Decomposition, EEMD），以分析非線性腦波訊號，檢測載頻（ƒc）與振幅調制頻率（ƒam），以區分帕金森氏症患者與健康受試者的神經訊號。HHSA分析結果揭示帕金森氏症患者與健康受試者之間靜息態腦電波的非線性動態特徵，顯示患者在額葉與中央區域的β-fc減弱，以及在中央、頂葉和顳葉區域的γ-fc減少。而與早期階段的患者相比，晚期階段的患者在後中央區域的β-fc下降，並且在左頂葉區域的θ-和δ-fc增強。此外，全腦θ-和β-fc與漢米爾頓憂鬱量表的分數呈正相關。更進一步，基於三個優先選擇的腦波特徵，機器學習演算法顯示「Bag」的分類準確率最高（0.90），其次是「LogitBoost」（0.89）。此外，旁側夾擊作業（Flanker task）通過強化注意區域的信息並抑制非注意區域的信息來測試視空間注意力。在帕金森氏症的患者中，額葉-皮層下迴路功能障礙影響該注意力網絡，進而影響反應選擇與抑制能力。HHT分析顯示，患者在一致性效應（congruency effect）期間的中額θ（FMθ）活動減弱。而HHSA分析結果進一步顯示，在一致性效應期間，θ範圍內的低頻fam減少。此外，基於全息希爾伯特交頻相位聚類分析（Holo-Hilbert cross-frequency phase clustering, HHCFPC）的連結性分析顯示，在患者的右前額葉皮層內θ-γ耦合減弱，而枕葉與右額葉區域內的θ-β耦合也受到損害。這些研究結果為進一步的大規模研究奠定了基礎，以增強特徵提取能力並改進模擬技術，以區分帕金森氏症患者的不同階段以及帕金森氏症相關的失智症，不僅如此，患者視空間注意力缺陷，透過HHSA和HHCFPC的分析，可用於識別新的神經生理生物標記，並指導介入措施，如神經調控像是連續θ脈衝刺激、視覺搜尋任務，或結合兩者以改善前額葉和枕葉區域的連結性。;Parkinson′s disease (PD) is a neurodegenerative disorder affecting pigmented nuclei in the midbrain and brainstem, leading to extrapyramidal symptoms. In addition to motor symptoms, PD also causes cognitive issues, whether in later stages or alongside motor symptoms. The primary pathology involves subcortical structures and dopaminergic connections between the basal ganglia (BG), thalamus, and frontal cortex, which, when impaired, contribute to Parkinson′s disease dementia (PDD), affecting executive function, attention, and memory. Electroencephalography (EEG) measures brain activity with precise timing and reliability. However, traditional linear computations in time-frequency analyses pose limitations when capturing nonlinear and non-stationary neural activity processes. Thus, we utilize the innovative Holo-Hilbert Spectral Analysis (HHSA) and Hilbert-Huang Transform (HHT) subsequent to Ensemble Empirical Mode Decomposition (EEMD) to analyse nonlinear EEG signals, examining carrier frequencies (ƒc) and amplitude modulation frequencies (ƒam) to differentiate neural signals in PD and healthy normal controls (NCs).
In PD, abnormal communication between the thalamus and cortical areas characterizes abnormal oscillations named thalamocortical dysrhythmia. The HHSA revealed dynamic non-linear features of the rsEEG in PD patients and NCs, where PD patients revealed decline of β-fc in frontal and central expanses, and lessening of γ-fc in central, parietal, and temporal expanses. Paralleled to early-stage PD (EPD), late-stage PD (LPD) patients revealed decrement of β-fc in the posterior central area, and augmented θ- and δ2-fc in the left parietal region. θ- and β-fc in the whole brain was positively correlated with Hamilton depression rating scale scores. Machine learning algorithms expending three prioritized HHSA features demonstrated “Bag” with the best accuracy of 0.90, shadowed by “LogitBoost” with an accuracy of 0.89. Additionally, the flanker task tests visuospatial attention by enhancing information in attended areas and suppressing it in unattended ones (Kopp et al., 1996a). In PD, frontal-subcortical circuit dysfunction impairs this attentional network, affecting response selection and inhibition. Holo-Hilbert transform (HHT) analysis revealed reduced midfrontal theta (FMθ) activity in PD patients during the congruency effect, confirming previous findings. HHSA showed a loss of low-frequency fam in theta activity during the congruency effect. Connectivity analysis using Holo-Hilbert cross-frequency phase clustering (HHCFPC) revealed reduced theta-gamma coupling in the right prefrontal cortex (PFC), with theta-beta coupling impairment in occipital and right frontal areas.
The rsEEG findings highlight the effectiveness of the HHSA method in distinguishing between PD patients and NCs. HHSA also detected depression predispositions linked to hyper-stable arousal regulation. Features from HHSA allowed clear separation of PD from NCs. These results provide a foundation for further studies with larger groups to enhance feature extraction and improve simulations to discriminate PD stages and PD-related dementia. The visuospatial attention deficits shown in the flanker task, analysed through the HHSA and HHCFPC can identify new neurophysiological biomarkers and guide interventions like neuromodulation (continuous theta burst stimulation or visual search tasks, or a combination of both to improve connectivity in the PFC and occipital regions. |
