以線性及非線性分析研究失智患者之腦波及心電圖

Abstract

緒論 失智症已是現今醫療及社會福利機構的ㄧ大負擔，預測到2040年全世界將有八千萬失智人口。因為目前有效的治療都強調早期使用的好處。人們積極的尋找疾病早期或是與預後有關的標記。兩大失智症病因為阿茲海默症以及血管型失智症，再其次則為路易氏體型失智症以及額顳葉型失智症。病理學上的研究仍未能解釋各型的成因，不過兩大病因皆與血管方面的危險因子有關。 阿茲海默症由基底額葉到大腦皮質及海馬迴的乙烯膽鹼系統是受損的。乙烯膽鹼系統被認為在大腦各項功能扮演重要腳色，如清醒及睡眠周期、視覺訊號處理、學習、以及記憶等功能。心與腦之間有密切關係，”neurovisceral integration model” 描述出從大腦前額葉皮質到達心臟的竇房結的整個中樞自主神經網路，並強調中樞自主神經中心以右邊的大腦前額葉皮質為主。 本研究應用各種線性及非線性分析法，來研究非侵犯性檢查-腦波及心電圖之訊號。生理訊號都受制於複雜的調控系統，其特性為既非穩定亦非線性，故分析方法須考慮或克服這些障礙。傳統的線性分析，多根基於傅立葉轉換，無法很細微及即時的提供訊息。目前常用方法中，線性但適合非穩定訊號的方法有小波分析及Wigner分佈等；非線性但穩定的分析法則多基於混沌理論，有如碎形及熵的各種運算；非線性且非穩定訊號的分析法則首推Hilbert-Huang transform (HHT)，其應用已在生物訊號的領域裡有很好的結果。 在心電圖方面，心跳兩兩之間的時間 (RR interval, RRI)，取決於竇房結去極化的速度，此則受制於自主神經系統的調控，於是研究心率變異便能代表心臟的自主神經系統調控。而當副交感神經作用較強時，HF升高、LF/HF ratio降低。LF則可能受交感及副交感神經兩者共同作用，並受壓力感受器系統調控，VLF則代表較慢的調控，可能代表一些血管運動、腎素-血管緊張素-醛固酮系統、以及人體熱調節系統。ULF則代表更長時間的影響。心率變異分析，用於失智症方面的報告結果不一。睡眠中的心律變異比清醒時高已在多個研究中獲得證實，而且不管是在REM sleep或Non-REM sleep皆然。睡眠呼吸中止症對於兩大型失智症皆是危險因子，這睡眠呼吸中止可以新的分析方法，由心電圖分析得知。 腦內的各項功能，皆需要各個不同區域的神經元做功能性聯結。這可以腦波上不同區域間的波在統計學上有互相依賴 (interdependence) 關係的狀況來代表，稱為耦合。腦波依頻段可分為alpha、beta、theta、delta、及gamma波，其來源不明。在失智的情況下，alpha波的強度變小、分佈前移、耦合變小。 Cross correlation and spectral coherence 以及 synchronization likelihood 是兩種近年來最具代表性的方法，前者為線性後者為非線性分析，皆是用以探討腦波各個channels 間關係的數學運算法，可以量化大腦皮質細胞間的同步化以及功能性耦合。如果大腦因為老化、藥物或是病變而使各區域間的聯繫受阻的話，同步化以及功能性耦合應該下降。 動脈硬化與血管型失智症及阿茲海默症皆有密切關係，有許多測量動脈硬化的參數可用來預測中風的危險性。生活型態例如社交參與、運動習慣、菸、酒等都與認知能力的下降有關。喝水量則尚無報告。 研究方法及材料 本研究設定為一個以老人科門診患者為對象來源、觀察性、病例對照、並且前瞻性追蹤的形態進行。通過署立台南醫院的臨床試驗同意 (IRB-2008004)。疾病組為65歲及以上的失智老人，第一次被診斷，其簡易認知功能測試經調整是否識字後分數小於等於26者。總計疾病組60人 (age: 80.52±5.6 , range:67-93, AD/VD: 24 /36 , male/female: 30/30, MMSE=19.8±6.9)，對照組29人(age: 75.28±6.5, range:65-87, male/female: 16/13, MMSE=28.4±0.9, MMSE=28.4±0.9)。檢查項目包括常規性腦波、24小時心電圖數位記錄、腦血管超音波、認知功能測驗 (mini-mental status examination (MMSE), clock-drawing test, clock-completion test, number transcoding task, trail making test, world list from CERAD-Plus, Boston naming test)、行動及活動量表 (Barthel index, timed up-and-go, Tinneti test)、憂鬱指數 (geriatric depressive scale)。疾病組並有腦部影像學檢查。 數據分析以MATLAB進行，運算腦波的spectral coherence、cross correlation coefficient of IMFs、multiscale entropy (MSE)，以及RRI的time domain and frequency domain data、multiscale entropy (MSE)。 結果與討論 比較特別的發現是: 1. 由HHT解構後的訊號所做的功能性耦合分析, 腦部各區的變數廣泛的與臨床心理測驗分數在校正性別年齡後成正比, 這意味大腦智能的運作是以全面的方式（holistic manner）進行。相較於線性方法，則只有在少數以額葉及顳葉為主的 channels看到此現象。 2. 失智患者比起對照組，有較顯著的vagal dominance during sleep 的現象。 3. MSE of EEG在各種尺度、各個channel下，皆與智能測驗分數成正比。 4. MSE of EKG 某些特定尺度的值與sum of MSEs of EEG 在廣泛channels的值，皆有很強的線性負相關。可由與線性參數的值對照(MSE 值與LF/HF ratio成正比)得到失智的腦有較低的副交感神經輸出（parasympathetic output）的結論, 或者可直接懷疑心律與腦的律動存在相同或極相關的成分。此外清醒的心電圖只與eye-closed resting 的腦波相關連 (Fz、T4、F3、Pz、O2、F8、F4、P4、C3、P3、F2)，而睡眠時的心電圖則與photic stimulation的腦波相關連(F2、T5、F4、T3、F1、T4、Cz、C3、O2、Pz、O1、Fz、F3、F7、C4)。清醒心電圖關連右腦的channels比較多，而睡眠心電圖則關連兩腦。 5. 比較三種情況下的腦波 (休息閉眼、閃光刺激、以及快速呼吸) 所得的結果，以閃光刺激所得結果最為顯著。 6. 各種不同的智能測驗，對照腦波的結果看來，word list、clock drawing、trail making、number transcoding鑑別力可說比MMSE還好。 7. 有Glabellar sign 者清醒的 LF/HF 較低。 8. 糖尿病患者睡眠的HF、LF、及清醒的 HF較低。 9. 血管超音波沒有發現顯著參數。 展望 本研究已顯示出腦波及心電圖的確藏有標記智能程度的參數。下一步計畫以社區中的老人為對象，收集腦波心電圖以及智能測驗分數，做長期的追蹤，然後運用所找到的較合適的分析方法，來尋找可以當為早期診斷或危險因子的標記。 另外對藥物治療之監測也很可行，對藥物有反應者如果腦波的參數有其特徵性，那麼也許可以早期投藥以預防疾病，或者用於預告無效而尋求其他的治療。 心與腦的關係，以腦波與心電圖的關係來繼續探討，應是值得投入的領域。 關鍵字: 失智症、心電圖、腦波、線性/非線性分析、耦合、熵、 Hilbert-Huang 轉換

Introduction The load of caring for demented patients is increasing globally very fast as eighty million demented population is expected in 2040. Alzheimer’s disease (AD) and vascular dementia (VD) are the two major causes of dementia. Since all current therapies for dementia depend on early diagnoses, risk and predicative factors for dementia are crucial. AD and VD share common risk factors as aging and vascular risks such as diabetes, hypertension, metabolic syndrome, homocystinemia, atrial fibrillation, and smoking. There are bidirectional connections between the heart and the brain. A neurovisceral integration model with laterality on the right prefrontal cortex was proposed to describe the pathways. EEG and EKG are nonstationary and nonlinear signals. The traditional Fourier spectrum is too coarse and fails to represent instantaneous changes. Nonstaionary but linear methods such as Wigner-Ville distribution and Wavelet have other drawbacks that hinder its adaptivity to ever-changing signals. Methods based on theories of chaos, fractal and entropy, quantifying either similarity, disorder, or stability, are suitable for nonlinear but stationary data. The Hilbert–Huang transform(HHT) on the other hand, is adaptive to nonlinear and nonstationary signals. The advantages of the HHT over traditional Fourier-based methods have been appreciated in many studies of different physiological systems. The study of heart rate variability (HRV), namely the variability of RR intervals (RRI), which reflects depolarization of the sinoatrial node, can monitor the autonomic system. The cholinergic deficits in the brain of dementia may affect the central autonomic network. Yet the HRV changes in dementia in previous reports were not congruent. A higher risk of dementia was shown in people with obstructive sleep apnea, which could be indicated by some newly developed methods of HRV analysis. Understanding how functional interactions among different brain regions are crucial to the study of higher cortical functions. The cross correlation and coherence analysis are two of the classical methodologies of linear approach. While the synchronization likelihood, which calculates the probability of similarity between two signals in phase space is a nonlinear approach. From synchronization to the execution of particular tasks of the brain, there hide still many puzzles such as ‘binding problem’. Social participation, exercise, smoking, and alcohol drinking may affect cognitive performance, while water intake insufficiency has yet to be proved. Methods This is a hospital-based, case control, and observational study with prospective follow-up of two groups (dementia and control). Various neuro-psychological and motility tests were performed in all subjects. As vascular risks are important in both types of dementia, carotid echosonography was also taken for each subject. Life style and eating behaviors were also compared. Both linear and nonlinear methods such as short time Fourier transform, spectral coherence, Hilbert Hung transform (HHT), multiscale entropy (MSE), and synchronization likelihood (SL) were performed for EEG signals. Heart rate variability was calculated individually in both awake and sleep EKG signals with linear analysis and MSE. Results and Discussions The demented group consists of 60 subjects (female/male=30/30, age 80.5±5.6, VD/AD=37/23, MMSE=19.8±6.9), while the control group consists of 29 subjects (female/male=13/16. age 75.3±6.4, MMSE=28.4±0.9). Significant findings are as following: 1. The cross correlation coefficients of data decomposed by HHT suggest that the brain functions in a more holistic manner. 2. The phenomenon of ‘Vagal dominance during sleep’ was only shown in the demented group. 3. The values of all scales of MSE from a wide range of electrodes are positively correlated with scores of mental abilities or mobility. 4. The MSE of RRI showed no correlation to mental capacities, but it had significantly negative correlations to the MSE of EEG in multiple area. Interestingly, the EEGs of closed-eye resting were associated to the RRIs during the awake state, while the EEGs of photic stimulation were mostly associated to the RRIs during sleep. 5. The photic stimulation yielded the most copious results. 6. Word list, clock drawing, trail making, and number transcoding tests had better differentiating power than MMSE. 7. Subjects with positive Glabellar signs had a lower LF/HF during awake state. 8. Diabetic subjects had lower HF and LF during sleep, and HF during awake state. 9. There was no correlation among the severity of carotid atherosclerosis to either mental capacities, parameters of EEG or parameters of EKG. 10. EEG of Dementia showed loss of coupling, complexity and stationarity Prospect A community based design with long time following is the next plan. It aims at risk factors and markers for early diagnosis by the analysis of EEG. Therapy monitoring by EEG is an ongoing study with some promising primitive results. The cross-talk between the brain and the heart could be further explored noninvasively by the information hidden in EKG and EEG. Key words: dementia, EKG, EEG, functional couplings, linear /nonlinear analysis, Multiscale Entropy, Hilbert- Huang Transformation