植入式神經(jīng)電信號(hào)檢測及其在麻醉深度評(píng)估中的應(yīng)用
發(fā)布時(shí)間:2018-06-04 04:18
本文選題:植入式神經(jīng)電信號(hào) + 近似熵 ; 參考:《杭州電子科技大學(xué)》2014年碩士論文
【摘要】:基于多電極陣列的植入式神經(jīng)電信號(hào)采集手段,提供給研究者一種在神經(jīng)元層次上觀察信息流產(chǎn)生、突觸傳遞以及編碼的途徑,在神經(jīng)工程的研究中扮演了越來越重要的角色。但其在具體應(yīng)用中,仍存在若干個(gè)急需解決的問題,例如多電極陣列各通道采集到的信號(hào)一般并非是單個(gè)神經(jīng)元的電活動(dòng),而是采集系統(tǒng)所引入的噪聲、神經(jīng)系統(tǒng)的背景噪聲,以及電極附近多個(gè)神經(jīng)元電活動(dòng)的時(shí)空疊加。因此首先需要從含噪的植入式神經(jīng)電信號(hào)中,檢測出有效的動(dòng)作電位(spike)信號(hào),并按發(fā)放源的不同對(duì)其進(jìn)行模式分類,這是探索神經(jīng)編碼機(jī)制的基礎(chǔ)和前提。又例如,當(dāng)檢測獲得特定皮層神經(jīng)元的脈沖發(fā)放序列后,用脈沖發(fā)放率這類粗略的傳統(tǒng)特征來表達(dá)它與外界激勵(lì)模式之間的關(guān)聯(lián)性,,可能并非是一種高效可行的編碼方式。因此本文針對(duì)植入式神經(jīng)電信號(hào)的檢測方法展開研究,并以麻醉深度評(píng)估為具體應(yīng)用,來探索神經(jīng)電信號(hào)的編碼技術(shù)。 由于神經(jīng)元電活動(dòng)是細(xì)胞膜上鉀鈉等離子通道開合的外在反映,其興奮時(shí)所產(chǎn)生的動(dòng)作電位信號(hào)不可避免具有一定的非線性時(shí)變特征,因此傳統(tǒng)的模板匹配法或者PCA等線性特征提取方法,在植入式神經(jīng)電信號(hào)的檢測中性能并不理想。本文利用近似熵在描述非線性信號(hào)復(fù)雜度的有效性,提出了基于近似熵的spike信號(hào)特征提取方法,并結(jié)合K均值聚類方法完成對(duì)信號(hào)的模式分類;同時(shí)考慮到神經(jīng)系統(tǒng)作為一類非線性動(dòng)力系統(tǒng),其看似無序的放電活動(dòng)在高維空間中存在著有序的成分。因此本文提出利用相空間重構(gòu)方法構(gòu)建spike信號(hào)的高維相空間,并結(jié)合QR分解方法提取空間中的結(jié)構(gòu)特征值,然后運(yùn)用信息熵進(jìn)一步量化其信號(hào)特征,最后實(shí)現(xiàn)了DBSCAN聚類方法在spike信號(hào)盲源分離中的應(yīng)用,克服了K均值對(duì)類簇形狀條件上的局限性。由于麻醉深度評(píng)估在醫(yī)學(xué)臨床中具有突出的研究意義,本文設(shè)計(jì)并開展了麻醉狀態(tài)下植入式神經(jīng)電信號(hào)采集的動(dòng)物實(shí)驗(yàn),并嘗試?yán)锰囟ㄆ由窠?jīng)元的脈沖發(fā)放序列,分別從放電頻率、放電間隔時(shí)間以及放電模式的復(fù)雜性等非線性角度出發(fā),研究神經(jīng)編碼在麻醉深度評(píng)估中的應(yīng)用。 本文主要工作和研究成果如下: (1)本文提出了一種基于近似熵的動(dòng)作電位特征提取方法。首先利用近似熵得到動(dòng)作電位的多維非線性特征,然后利用KS檢驗(yàn)進(jìn)行特征降維,選出最具可分性的特征,并結(jié)合K均值聚類實(shí)現(xiàn)spike信號(hào)的分類。針對(duì)仿真和真實(shí)實(shí)驗(yàn)數(shù)據(jù),兩類數(shù)據(jù)的模式分類都取得了較好的結(jié)果,本文新方法對(duì)于非同源spike信號(hào)分類具有一定的優(yōu)勢。 (2)本文提出了一種基于相空間重構(gòu)和QR分解的動(dòng)作電位特征提取方法。針對(duì)動(dòng)作電位發(fā)放過程表現(xiàn)的動(dòng)態(tài)性,本文提出用相空間重構(gòu)的方法表征其動(dòng)態(tài)信息,然后在重構(gòu)的高維相空間中利用QR分解提取出其特征值,最后結(jié)合DBSCAN聚類方法實(shí)現(xiàn)spike信號(hào)的無監(jiān)督分類。仿真數(shù)據(jù)的分類準(zhǔn)確率達(dá)到98%以上,并通過與傳統(tǒng)方法PCA分類結(jié)果比較,進(jìn)一步說明相空間重構(gòu)結(jié)合密度聚類方法所刻畫的特征能較好的區(qū)分出非同源信號(hào)間的差異。實(shí)驗(yàn)數(shù)據(jù)的分類結(jié)果也表明本文新方法可作為spike模式分類的依據(jù)。 (3)本文提出了一種基于復(fù)雜度和多尺度熵的脈沖序列編碼方法,彌補(bǔ)了放電頻率和放電時(shí)間間隔編碼時(shí)對(duì)放電信息的大量丟失,從放電模式的非線性特征上表達(dá)了一種可行的神經(jīng)編碼方法。本文針對(duì)醫(yī)學(xué)臨床中的麻醉深度評(píng)估應(yīng)用,設(shè)計(jì)并開展了對(duì)應(yīng)的動(dòng)物實(shí)驗(yàn),采集了不同麻醉狀態(tài)下的植入式神經(jīng)電活動(dòng),分別從神經(jīng)元放電信號(hào)的平均發(fā)放率、基于發(fā)放時(shí)間間隔的三種時(shí)間編碼方式、以及Lempel-Ziv復(fù)雜度和多尺度熵等放電模式的角度,研究神經(jīng)編碼在麻醉深度評(píng)估應(yīng)用中的可行性,結(jié)果表明,本文提出的復(fù)雜度和多尺度熵方法,在深度麻醉和淺度麻醉狀態(tài)的模式識(shí)別中,具有一定的區(qū)分性,驗(yàn)證了它們作為一種神經(jīng)編碼方式的可行性。
[Abstract]:In this paper , it is necessary to detect effective action potential ( spike ) signals in neural engineering , and to classify them according to different distribution sources , which is not an efficient and feasible coding method .
Because the neuronal activity is an external reflection of ionic channels such as potassium and sodium on the cell membrane , the action potential signal generated during excitation inevitably has a certain non - linear time - varying characteristic , so the traditional template matching method or PCA and other linear feature extraction methods are not ideal in the detection of implanted neural signals .
In this paper , we put forward the application of the method of phase space reconstruction to construct the high - dimensional phase space of spike signal , and combine the QR decomposition method to extract the characteristic value of the structure in space , and then use the information entropy to further quantify its signal characteristics . Finally , the application of DBSCAN clustering method in the blind source separation of spike signal is realized .
The main work and research results are as follows :
( 1 ) A feature extraction method based on approximate entropy is presented in this paper . First , using approximate entropy to obtain multi - dimensional non - linear characteristic of action potential , and then using KS test to carry out the feature reduction , the most distinguishing feature is selected , and the classification of spike signal is realized by combining K - means clustering . For the simulation and real experimental data , the pattern classification of two types of data has got better results . The new method has some advantages for non - homologous spike signal classification .
( 2 ) A feature extraction method based on phase space reconstruction and QR decomposition is presented in this paper . The dynamic information is characterized by the method of phase space reconstruction .
( 3 ) In this paper , a kind of pulse sequence coding method based on complexity and multi - scale entropy is put forward , which makes up the great loss of discharge information when the discharge frequency and the discharge time interval are coded . The feasibility of the nerve coding in the anesthesia depth evaluation is studied .
【學(xué)位授予單位】:杭州電子科技大學(xué)
【學(xué)位級(jí)別】:碩士
【學(xué)位授予年份】:2014
【分類號(hào)】:R614
【參考文獻(xiàn)】
相關(guān)期刊論文 前10條
1 侯澍e
本文編號(hào):1975876
本文鏈接:http://sikaile.net/yixuelunwen/mazuiyixuelunwen/1975876.html
最近更新
教材專著