CN106037759A - Cerebral self-adjusting index detection method for sleep apnea - Google Patents

Abstract

The invention discloses a cerebral self-adjusting index detection method for sleep apnea. The detection method comprises five steps: 1, conducting such pre-processing as denoising, down-sampling and the like on tissue blood oxygen and pulse blood oxygen which are acquired by virtue of a near-infrared spectrum recorder; 2, calculating a moving correlation coefficient r of the tissue blood oxygen and the pulse blood oxygen; 3, conducting segmented weighting on the obtained moving correlation coefficient, so that a weighted moving correlation coefficient wr is obtained; 4, calculating a weighted average of the weighted moving correlation coefficient, so that a cerebral self-adjusting index CARindex is obtained; and 5, assessing the influence of the sleep apnea on a cerebral self-adjusting function in accordance with the cerebral self-adjusting index. The detection method disclosed by the invention can effectively avoid shortcomings of an existing method which is complex in data acquisition, difficult to operate and the like.

Description

A kind of brain self-regulation index detection method towards sleep apnea

Technical field

The invention belongs to processing of biomedical signals technical field, particularly to utilize near-infrared spectrum technique calculate based on The brain self-regulation index of brain oximetry data evaluates a kind of method that sleep apnea is dangerous.

Background technology

Low blood oxygen and high carbon acid can be caused when suffering from the patient respiratory time-out recurrent exerbation of sleep apnea syndrome, can Cause maladjusted nervous system, endocrine regulation and hemodynamic change, cause the damage of whole body multiple organ, serious threat people Healthy.Clinical research confirmation, it is closely related with multiple fatal disease, including apoplexy, hypertension, coronary artery disease Sick etc..

Sleep analysis monitor (Polysomnography, PSG) is the gold mark of current diagnosis sleep apnea syndrome Accurate.By the record patient's multinomial physiological parameter in sleep procedure, including electroencephalogram, electrocardiogram, electromyogram, mouth and nose respiratory gas Stream, the sound of snoring, blood oxygen saturation, heart rate, position, chest and abdomen breathing etc., utilize professional software to be analyzed these parameters, slept The final reports such as dormancy structure, disordered breathing, sleep monitor.But sleep analysis monitor has following defects that

1, monitoring process need to be carried out at the laboratory of specialty；

2, inspection needs synchronization to carry out long-time multinomial physiology monitoring, and patient needs whole body to stick big quantity sensor, and this is to trouble The sleep quality of person has certain impact, thus causes monitoring of environmental and true environment difference relatively big, affects the accurate of monitoring result Property；

3, check expensive；

4, the order of severity of sleep apnea syndrome is only diagnosed by sleep analysis monitor, can not directly assess The degree of injury of brain in patients self-regulating function.

In the case of the brain self-regulating function of people is int, apneic generation can cause pulse blood oxygen to reduce, for Maintaining the normal level of providing brain with oxygen, tissue oxygenation can be stablized at certain model under brain self-regulating negative feedback mechanism effect Enclose.But the apneic recurrent exerbation of extended sleep can make brain self-regulating function impaired, between pulse blood oxygen and tissue oxygenation Negative feedback will gradually weaken and even disappear, and the power of negative feedback and brain damage degree have substantial connection.

The method that brain self-regulating function is affected by assessment sleep apnea syndrome at present is mainly: use many through cranium Pu Le (trancranial Doppler, TCD) technology, by calculating average cerebral blood flow velocity (cerebral blood flow Velocity, CBFV) and the mobile correlation coefficient of arteriotony (arterial blood pressure, ABP), obtain one The brain index from main regulation can be assessed.

But there is following defect in the method:

1, transcranial doppler need doctor at other operating instrument, and the region that middle cerebral artery to be guaranteed is irradiated by sound wave Constant diameter, so this acquisition method can not obtain the long-time data during sleep apnea syndrome patient sleeps.

2, arteriotony measurement generally requires employing invasive mode, and for sleep-apnea, patient uses invasive blood pressure Metering system is unpractical.

Summary of the invention

In order to overcome above-mentioned existing methodical defect, it is an object of the invention to provide a kind of towards sleep apnea Brain self-regulation index detection method, utilizes near-infrared spectrum technique, is calculated from the brain near infrared light spectrum signal gathered Tissue oxygenation and pulse blood oxygen signal, between period of sleep apnea, pulse blood oxygen has and to a certain degree reduces, and now brain is certainly Regulating power will directly be reflected in tissue oxygenation signal, tissue oxygenation signal moves weighting to pulse blood oxygen signal relevant The calculating of coefficient, finally gives brain self-regulation index, to assess sleep apnea syndrome to brain self-regulating function Impact, utilizes the method can be prevented effectively from the defects such as existing method data acquisition complexity, operating difficulties.

In order to achieve the above object, the technical solution used in the present invention is:

A kind of brain self-regulation index detection method towards sleep apnea, comprises the following steps:

Step one: near infrared spectrum recorder is gathered detection object night tissue oxygenation with pulse blood oxygen number According to, carrying out denoising, down-sampled pretreatment, pretreated tissue oxygenation signal is expressed as a length of with pulse blood oxygen signal Discrete-time series rso2 (n) of N and spo2 (n)；Afterwards the data handled well are sent to processing unit；

Step 2: processing unit computation organization blood oxygen and the mobile correlation coefficient of pulse blood oxygen

(1), determine the movable length mLen of mobile segmentation, segment length sLen and segments nesg, see formula (1)～ (3):

$mLen=\frac{f_s}{f_0}---\left(1\right)$

SLen=3*mLen (2)

Wherein,

f_sSignal sampling rate；

f₀Required minimum frequency resolution；

Rounding operation downwards；

(2), calculate the mobile correlation coefficient r (i) that each signal subsection is corresponding, see formula (4):

$r\left(i\right)=\frac{l\left(spo{2}_i\right(n),rso{2}_i(n\left)\right)}{\sqrt{l\left(spo{2}_i\right(n),spo{2}_i(n\left)\right)*l\left(rso{2}_i\right(n),rso{2}_i(n\left)\right)}},1\≤i\≤nseg---\left(4\right)$

Wherein,

spo2_i(n) and rso2_iN () spo2 (n) and i-th segmentation of rso2 (n), be expressed as follows:

spo2_i(n)=spo2 (n), mLen* (i-1)+1≤n≤mLen* (i-1)+sLen (5)

rso2_i(n)=rso2 (n), mLen* (i-1)+1≤n≤mLen* (i-1)+sLen (6)

L (... ...) sum of products of mean deviations (or sum of sguares of deviation from mean) computing, computing formula is as follows:

$l(x,y)=\underset{i=1}{\overset{m}{\Σ}}(x_i-\stackrel{\‾}{x})*(y_i-\stackrel{\‾}{y})---\left(7\right)$

Wherein,

The length of m x Yu y；

WithThe average of x Yu y；

(3), the weight w that each signal subsection is corresponding is calculated_p(i) and w_r(i), weight computing in step 4, see formula (8) with (9):

$w_p\left(i\right)=\sqrt{\frac{\underset{n=mLen*(i-1)+1}{\overset{mLen*(i-1)+sLen}{\Σ}}\left(spo2\right(n)-\stackrel{\‾}{spo{2}_i\left(n\right)})}{sLen-1}}---\left(8\right)$

$w_r\left(i\right)=\sqrt{\frac{\underset{n=mLen*(i-1)+1}{\overset{mLen*(i-1)+sLen}{\Σ}}\left(spo2\right(n)-\stackrel{\‾}{spo{2}_i\left(n\right)})}{sLen-1}}---\left(9\right)$

Wherein,

WithThe i-th block signal average of spo2 (n) and rso2 (n), is calculated as follows:

$\stackrel{\‾}{spo{2}_i\left(n\right)}=\frac{\underset{n=mLen*(i-1)+1}{\overset{mLen*(i-1)+sLen}{\Σ}}spo2\left(n\right)}{sLen-1}---\left(10\right)$

$\stackrel{\‾}{rso{2}_i\left(n\right)}=\frac{\underset{n=mLen*(i-1)+1}{\overset{mLen*(i-1)+sLen}{\Σ}}rso2\left(n\right)}{sLen-1}---\left(11\right)$

Step 3: step 2 obtain mobile correlation coefficient be r=[r (1), r (2) ..., r (nseg)], corresponding two groups Weights are w_p=[w_p(1),w_p(2),...,w_p] and w (nseg)_r=[w_r(1),w_r(2),...,w_r(nseg)], weights pair are used Mobile correlation coefficient weighted calculates and moves weighted correlation coefficient wr:

(1), to mobile correlation coefficient segmentation, determine segments nseg ', see formula (12) and (13):

SLen '=300*f₀ (12)

Wherein,

Every section of segment length of sLen ' segmentation；

(2), calculate every section of corresponding weighting and move correlation coefficient wr (j), see formula (14):

$wr\left(j\right)=w\left(j\right)*\frac{\underset{i={\mathrm{sLen}}^{\′}*j+1}{\overset{{\mathrm{sLen}}^{\′}*(j+1)}{\Σ}}r\left(i\right)}{{\mathrm{sLen}}^{\′}-1},1\≤j\≤{\mathrm{nseg}}^{\′}---\left(14\right)$

Wherein,

The correlation coefficient weights that w (j) jth segmentation is corresponding, computing formula is as follows:

$w\left(j\right)=max(\frac{\underset{i={\mathrm{sLen}}^{\′}*j+1}{\overset{{\mathrm{sLen}}^{\′}*(j+1)}{\Σ}}{w}_p\left(i\right)}{{\mathrm{sLen}}^{\′}-1},\frac{\underset{i={\mathrm{sLen}}^{\′}*j+1}{\overset{{\mathrm{sLen}}^{\′}*(j+1)}{\Σ}}{w}_r\left(i\right)}{{\mathrm{sLen}}^{\′}-1}),1\≤j\≤{\mathrm{nseg}}^{\′}---\left(15\right)$

Wherein,

Max (... ...) take higher value computing；

Step 4: previous step obtains mobile weighted correlation coefficient wr, and it is asked weighted average calculation brain self-regulation index CARindex, is shown in formula (16):

$CARindex=\frac{\underset{j=1}{\overset{{\mathrm{nseg}}^{\′}}{\Σ}}wr\left(j\right)}{\underset{j=1}{\overset{{\mathrm{nseg}}^{\′}}{\Σ}}w\left(j\right)}---\left(16\right)$

If CARindex > 1, then CARindex=1；If CARindex <-1, then CARindex=-1；

Step 5: processing unit sends signals below prompting, CARindex ∈ [-1,1]: work as CARindex to display unit When≤0, send brain self-regulating function and be without damage；As CARindex > 0 time, more level off to 1, send sleep apnea and combine Close the harm of disease regulatory function autonomous to brain the biggest；More level off to 0, send sleep apnea syndrome to brain from main regulation Function harm is the least.

The invention have the advantage that and the detecting sphygmus and blood oxygen saturation that cause because of sleep apnea is declined, with by brain from The change of the tissue oxygenation saturation of regulation connects, and method easy when using noinvasive long obtains initial data, it is to avoid Traditional method can only gather and the defect of complexity the short time.

Move correlation coefficient by calculating the weighting of pulse blood oxygen and tissue oxygenation, obtain reflecting that sleep apnea is to greatly The brain self-regulation index of brain self-regulating function influence degree, thus utilize brain self-regulation index to evaluate sleep apnea Dangerous.

Accompanying drawing explanation

Fig. 1-a is the time domain beamformer of pulse blood oxygen and tissue oxygenation primary signal.

Fig. 1-b is the part-time sectional drawing of pulse blood oxygen and tissue oxygenation primary signal.

Fig. 2-a is through denoising and down-sampled pretreated time domain beamformer.

Fig. 2-b is through denoising and down-sampled pretreated part-time sectional drawing.

Fig. 3-a is the mobile correlation coefficient figure of pulse blood oxygen and tissue oxygenation.

Fig. 3-b is the part-time sectional drawing of pulse blood oxygen and tissue oxygenation.

Fig. 4 is the mobile weighted correlation coefficient figure of pulse blood oxygen and tissue oxygenation.

Fig. 5 is the flow chart of the present invention.

Specific embodiment

Below in conjunction with the accompanying drawings and the present invention is described in detail by example.

Suffer from sleep apnea syndrome patient as data sample with one, gather near infrared light spectrum signal at night, number According to duration 8 hours 8 points 56 seconds (from 19:53:00 04:01:56), sample rate 10Hz, 293360 points of data length.According to closely Infrared spectroscopy signals calculates pulse blood oxygen and tissue oxygenation data.Primary signal time domain beamformer and part sectional drawing such as Fig. 1-a with Shown in Fig. 1-b.In order to assess the impact of the sleep apnea syndrome of this patient regulatory function autonomous on its brain, use this Data are analyzed by invention.

A kind of brain self-regulation index detection method towards sleep apnea, with reference to Fig. 5, comprises the following steps:

Step one: the signal own situation gathered according near infrared spectrum recorder, carries out denoising and down-sampled to it Deng pretreatment, after pretreatment, signal sampling rate is 0.2Hz, and data length is 5268 points.Time domain plethysmographic signal figure after pretreatment And part sectional drawing is as shown in Fig. 2-a and Fig. 2-b, afterwards the data handled well is sent to processing unit；

Step 2: processing unit computation organization blood oxygen and the mobile correlation coefficient r of pulse blood oxygen

Required minimum frequency resolution f₀=0.1Hz, obtains according to formula (1)～(3), the movable length of mobile segmentation MLen=2, segment length sLen=6 and segments nesg=2632；The movement that each signal subsection is corresponding is calculated according to formula (4) Correlation coefficient r (i).Finally, mobile correlation coefficient r is obtained, as shown in Figure 3.

Step 3: calculate mobile weighted correlation coefficient wr.

According to formula (12) and (13) mobile correlation coefficient carried out segmentation, segment length sLen '=30, segments nseg '= 85；Calculate corresponding weight value according to formula (15), finally obtain mobile weighted correlation coefficient wr according to formula (14), as shown in Figure 4.

Step 4: correlation coefficient wr mobile to weighting asks weighted average to obtain brain self-regulation index CARindex.

The brain self-regulation index CARindex=0.7773 of this patient can be calculated according to formula (16).

Step 5: processing unit sends signals below prompting, CARindex ∈ [-1,1]: work as CARindex to display unit When≤0, send brain self-regulating function and be without damage；As CARindex > 0 time, more level off to 1, send sleep apnea and combine Close the harm of disease regulatory function autonomous to brain the biggest；More level off to 0, send sleep apnea syndrome to brain from main regulation Function harm is the least.The sample data brain self-regulation index of this patient is 0.7773, shows sleep apnea syndrome pair The damage of its brain self-regulating function is the biggest.

Claims (1)

1. the brain self-regulation index detection method towards sleep apnea, it is characterised in that comprise the following steps:

Step one: near infrared spectrum recorder is gathered detection object night tissue oxygenation with pulse blood oxygen data, enter Row denoising, down-sampled pretreatment, pretreated tissue oxygenation signal and pulse blood oxygen signal be expressed as a length of N from Dissipate time series rso2 (n) and spo2 (n)；Afterwards the data handled well are sent to processing unit；

Step 2: processing unit computation organization blood oxygen and the mobile correlation coefficient of pulse blood oxygen

(1), determine the movable length mLen of mobile segmentation, segment length sLen and segments nesg, see formula (1)～(3):

$mLen=\frac{f_s}{f_0}---\left(1\right)$

SLen=3*mLen (2)

Wherein,

f_sSignal sampling rate；

f₀Required minimum frequency resolution；

Rounding operation downwards；

(2), calculate the mobile correlation coefficient r (i) that each signal subsection is corresponding, see formula (4):

$r\left(i\right)=\frac{l\left(spo{2}_i\right(n),rso{2}_i(n\left)\right)}{\sqrt{l\left(spo{2}_i\right(n),spo{2}_i(n\left)\right)*l\left(rso{2}_i\right(n),rso{2}_i(n\left)\right)}},1\&le;i\&le;nseg---\left(4\right)$

Wherein,

spo2_i(n) and rso2_iN () spo2 (n) and i-th segmentation of rso2 (n), be expressed as follows:

spo2_i(n)=spo2 (n), mLen* (i-1)+1≤n≤mLen* (i-1)+sLen (5)

rso2_i(n)=rso2 (n), mLen* (i-1)+1≤n≤mLen* (i-1)+sLen (6)

L (... ...) sum of products of mean deviations (or sum of sguares of deviation from mean) computing, computing formula is as follows:

$l(x,y)=\underset{i=1}{\overset{m}{\&Sigma;}}(x_i-\stackrel{\&OverBar;}{x})*(y_i-\stackrel{\&OverBar;}{y})---\left(7\right)$

Wherein,

The length of m x Yu y；

WithThe average of x Yu y；

(3), the weight w that each signal subsection is corresponding is calculated_p(i) and w_r(i), weight computing in step 4, see formula (8) with (9):

$w_p\left(i\right)=\sqrt{\frac{\underset{n=mLen*(i-1)+1}{\overset{mLen*(i-1)+sLen}{\&Sigma;}}\left(spo2\right(n)-\stackrel{\&OverBar;}{spo{2}_i\left(n\right)})}{sLen-1}}---\left(8\right)$

$w_r\left(i\right)=\sqrt{\frac{\underset{n=mLen*(i-1)+1}{\overset{mLen*(i-1)+sLen}{\&Sigma;}}\left(spo2\right(n)-\stackrel{\&OverBar;}{spo{2}_i\left(n\right)})}{sLen-1}}---\left(9\right)$

Wherein,

WithThe i-th block signal average of spo2 (n) and rso2 (n), is calculated as follows:

$\stackrel{\&OverBar;}{spo{2}_i\left(n\right)}=\frac{\underset{n=mLen*(i-1)+1}{\overset{mLen*(i-1)+sLen}{\&Sigma;}}spo2\left(n\right)}{sLen-1}---\left(10\right)$

$\stackrel{\&OverBar;}{rso{2}_i\left(n\right)}=\frac{\underset{n=mLen*(i-1)+1}{\overset{mLen*(i-1)+sLen}{\&Sigma;}}rso2\left(n\right)}{sLen-1}---\left(11\right)$

Step 3: step 2 obtain mobile correlation coefficient be r=[r (1), r (2) ..., r (nseg)], two groups of corresponding weights For w_p=[w_p(1),w_p(2),...,w_p] and w (nseg)_r=[w_r(1),w_r(2),...,w_r(nseg)], use weights to movement Correlation coefficient weighted calculates mobile weighted correlation coefficient wr；

(1), to mobile correlation coefficient segmentation, determine segments nseg ', see formula (12) and (13):

SLen '=300*f₀ (12)

Wherein,

Every section of segment length of sLen ' segmentation；

(2), calculate every section of corresponding weighting and move correlation coefficient wr (j), see formula (14):

$wr\left(j\right)=w\left(j\right)*\frac{\underset{i={\mathrm{sLen}}^{\&prime;}*j+1}{\overset{{\mathrm{sLen}}^{\&prime;}*(j+1)}{\&Sigma;}}r\left(i\right)}{{\mathrm{sLen}}^{\&prime;}-1},1\&le;j\&le;{\mathrm{nseg}}^{\&prime;}---\left(14\right)$

Wherein,

The correlation coefficient weights that w (j) jth segmentation is corresponding, computing formula is as follows:

$w\left(j\right)=max(\frac{\underset{i={\mathrm{sLen}}^{\&prime;}*j+1}{\overset{{\mathrm{sLen}}^{\&prime;}*(j+1)}{\&Sigma;}}{w}_p\left(i\right)}{{\mathrm{sLen}}^{\&prime;}-1},\frac{\underset{i={\mathrm{sLen}}^{\&prime;}*j+1}{\overset{{\mathrm{sLen}}^{\&prime;}*(j+1)}{\&Sigma;}}{w}_r\left(i\right)}{{\mathrm{sLen}}^{\&prime;}-1}),1\&le;j\&le;{\mathrm{nseg}}^{\&prime;}---\left(15\right)$

Wherein,

Max (... ...) take higher value computing；

Step 4: previous step obtains mobile weighted correlation coefficient wr, and it is asked weighted average calculation brain self-regulation index CARindex, is shown in formula (16):

$CARindex=\frac{\underset{j=1}{\overset{{\mathrm{nseg}}^{\&prime;}}{\&Sigma;}}wr\left(j\right)}{\underset{j=1}{\overset{{\mathrm{nseg}}^{\&prime;}}{\&Sigma;}}w\left(j\right)}---\left(16\right)$

If CARindex > 1, then CARindex=1；If CARindex <-1, then CARindex=-1；

Step 5: processing unit sends signals below prompting, CARindex ∈ [-1,1]: when CARindex≤0 to display unit Time, send brain self-regulating function and be without damage；As CARindex > 0 time, more level off to 1, send sleep apnea syndrome Regulatory function autonomous to brain harm is the biggest；More level off to 0, send sleep apnea syndrome regulatory function autonomous to brain Endanger the least.