KR20060079444A - Separating apparatus and method of fetal ecg from maternal abdomen ecg - Google Patents

Abstract

The present invention relates to an apparatus and method for separating fetal ECG from a maternal abdominal electrocardiogram, and more particularly to an apparatus and method for separating fetal ECG from a maternal ECG detected in a single channel from the maternal abdomen.

A fetal ECG separation method from a single channel maternal abdominal electrocardiogram according to the present invention includes a data input step of detecting maternal abdominal electrocardiogram from a maternal abdomen as a single channel and starting abdominal electrocardiogram data input through an A / D converter; A determination step of determining whether a separation vector exists; If the separation vector does not exist, the independent vector is reconstructed into a multi-dimensional signal in the time-frequency domain by applying the smooth pseudo Wigner-Ville distribution (SPWVD) function from the signal received from the A / D converter for a predetermined time. A separation vector setting step of applying a component analysis (ICA) to set a separation vector; If the separation vector exists, the multi-dimensional signal reconstruction step of reconstructing the multi-dimensional signal in the time-frequency domain by applying the smooth pseudo Wigner-Ville distribution (SPWVD) function from the signal received from the A / D converter for a predetermined time ; A bit detection step of detecting a bit (R point, heart rate) of the fetal ECG by multiplying the separation vector by the multidimensional signal reconstructed in the multidimensional signal reconstruction step; Calculating a fetal heart rate rate and configuring the fetal ECG template; Characterized in having a.

 Single Channel Abdominal Electrocardiogram, Fetal ECG Isolation, Independent Factor Analysis, Isolation Vector, Fetal Health Monitoring

Description

Separating apparatus and method of fetal ECG from maternal abdomen ECG

1 is an example of a signal recorded in the mother's abdomen.

Figure 2 is a block diagram illustrating the configuration of fetal ECG separation apparatus from a single channel abdominal electrocardiogram according to an embodiment of the present invention.

3 is a flowchart illustrating an example of a first time-frequency domain converter of FIG. 2.

4 is a connection diagram illustrating an application of multidimensional signals and independent element analysis from a single channel signal.

FIG. 5 is a flowchart for describing the blind signal separation unit of FIG. 2.

6 is a flowchart illustrating the fetal heartbeat extraction unit of FIG. 2.

FIG. 7 is a flowchart illustrating the fetal ECG complex calculation unit of FIG. 2.

8 is a flowchart schematically illustrating a method of separating fetal ECG from a maternal abdominal electrocardiogram according to an embodiment of the present invention.

FIG. 9A is an example of maternal abdominal electrocardiogram and smooth pseudo-Wigner-Ville distribution (SPWVD) function application results when using a fetal ECG separation device from the single channel abdominal electrocardiogram of FIG. 2.

FIG. 9B is an example of an estimated signal source obtained by applying an independent element analysis (ICA) to the result of applying the smooth pseudo-Wigner-Ville distribution (SPWVD) function of FIG. 9A.

FIG. 10A illustrates an example of maternal abdominal electrocardiogram and fetal ECG bit detection results using a separation vector when using the fetal ECG separation apparatus from the single-channel abdominal electrocardiogram of FIG. 2.

FIG. 10B illustrates an example in which an ECG of a mother and an ECG of a fetus overlap with each other in FIG. 10A.

11A is an example of fetal heart rate change rate calculated from detected fetal ECG bits when using fetal ECG separation device from the single channel abdominal electrocardiogram of FIG. 2.

FIG. 11B is an example of a representative template obtained from FIG. 11A.

The present invention relates to an apparatus and method for separating fetal ECG from a maternal abdominal electrocardiogram, and more particularly to an apparatus and method for separating fetal ECG from a maternal ECG detected in a single channel from the maternal abdomen.

In general, the monitoring of the fetal condition is essential for mothers associated with gestational hypertension, gestational diabetes mellitus, and low birth weight infants, and it is necessary to continuously examine the condition of the fetus from prenatal to analgesia.

In order to examine the condition of the fetus is a fetal ECG. The fetal ECG contains important information indicating the health of the fetus, which is an important diagnostic tool.

One method of acquiring a fetal ECG is through an ECG acquisition system in the mother's abdomen.

1 is an example of a signal recorded in the mother's abdomen. In general, the signal recorded in the mother's abdomen is complicated by the mother's ECG and random noise. The main peak that appears periodically is maternal electrocardiogram due to maternal cardiac activity, and peaks with relatively smaller and faster periods are due to fetal cardiac activity. As in adults, the fetal ECG complex shape and heart rate variability (HRV) are important health indicators available from fetal ECG.

Recently, various studies for detecting fetal ECG have been conducted, and most of these studies have been conducted when a multi-channel signal is acquired.

In the study of separating ECG signals recorded in multiple channels, the most direct method is to subtract ECG signals obtained from the mother's chest from mixed signals measured in the mother's abdomen, which is called singular value decomposition (SVD). ), Adaptive filter, wavelet transform method, etc., have been studied by applying various signal processing techniques. In this case, in addition to the electrode for measuring the maternal abdominal electrocardiogram signal, there is an inconvenience that the electrode for measuring the ECG of the maternal chest should be separately installed.

On the other hand, relatively few studies have been published to isolate ECG signals recorded on a single channel. For example, Kanjilal et al. First detect the maternal and fetal heartbeats from a single-channel signal, then divide the signal into maternal and fetal intervals, and matrix them to apply singular value decomposition (SVD). By suggesting a method for obtaining an ECG complex.

In particular, there is an increasing demand to check the condition of the fetus 24 hours in normal mothers as well as mothers associated with gestational hypertension, gestational diabetes mellitus, and low birth weight infants. It is desirable to develop a fetal monitoring device that can monitor the fetal ECG by acquiring abdominal ECG data without measuring the mother's chest ECG.

Without measuring maternal chest electrocardiogram, it is difficult to use most of the existing techniques to measure fetal electrocardiogram by acquiring single channel or double channel maternal abdominal electrocardiogram data.

Accordingly, the present invention provides a fetal ECG separation method for detecting a fetal heartbeat by using a single channel signal obtained from a mother's abdomen and obtaining a representative complex form of the fetus through the detected heartbeat information. In particular, the present invention can calculate the fetal heart rate rate (FHRV) and the morphology of the beat by separating the fetal ECG from a single channel ECG measured in the mother's abdomen, from which it is possible to provide fetal health information .

In addition, the present invention is a fetal ECG separation apparatus from maternal abdominal electrocardiogram for applying a smooth pseudo Wigner-Ville distribution (SPWVD) to reconstruct the multi-dimensional signal on the time-frequency and then to obtain a separation vector by applying an independent element analysis technique and Provide a method.

 An object of the present invention is to provide a fetal ECG separation apparatus and method for detecting a fetal heartbeat with a single channel signal obtained from the mother's abdomen.

Another technical problem of the present invention is to reconstruct a multidimensional signal over time-frequency by applying the smooth pseudo Wigner-Ville distribution (SPWVD) function, and then apply an independent element analysis technique to obtain a separation vector. An apparatus and method for separating fetal ECG from

Hereinafter, a configuration and an operation of an apparatus and a method for separating fetal ECG from a single channel abdominal electrocardiogram according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

2 is a block diagram illustrating the configuration of a fetal ECG separation apparatus from a single channel abdominal electrocardiogram according to an exemplary embodiment of the present invention, wherein the electrode portion 100, the signal preprocessor 200, and the A / D converter ( 300), the fetal ECG separation module 400, and the display unit 500. The fetal ECG separation module 400 includes a separated vector extracting unit 1000, a first storage unit 1900, a fetal heart rate extracting unit 2000, a second storage unit 2900, and a fetal ECG complex calculating unit 3000. .

The electrode unit 100 detects a biological signal from the mother's abdomen. The electrode unit 100 may include two electrodes.

The signal preprocessor 200 includes an amplifier and a filter, and amplifies the biosignal detected from the electrode 100 and removes noise.

The A / D converter 300 converts the output signal of the signal detector 200 into a digital signal. The A / D converter 300 may use an A / D converter having a sampling rate of 500 Hz and a resolution of 12 bits.

The fetal ECG separation module 400 separates the maternal electrocardiogram and the fetal ECG from the signal for the first predetermined time received from the A / D converter 300 and extracts the fetal heartbeat therefrom. The fetal ECG separation module 400 includes a separated vector extracting unit 1000, a first storage unit 1900, a fetal heart rate extracting unit 2000, a second storage unit 2900, and a fetal ECG complex calculating unit 3000. . The fetal ECG separation module unit 400 may be configured as a microprocessor.

The separation vector extractor 1000 obtains a demixing vector from a signal received from the A / D converter 300 for an initial predetermined time. The separation vector extractor 1000 includes a first time-frequency domain converter 1100 and a blind signal separator 1200.

The first time-frequency domain converter 1100 reconstructs a signal received from the A / D converter 300 into a multidimensional signal on the time-frequency domain. That is, the first time-frequency domain converter 1100 reconstructs the single channel signal into a multidimensional signal on the time-frequency.

The blind signal separator 1200 may extract a signal corresponding to the fetal ECG from the signal reconstructed by the first time-frequency domain converter 1100 into a multi-dimensional signal on the time-frequency domain through a blind signal separation method. Find the separation vector.

The first storage unit 1900 stores the separation vector obtained from the blind signal separation unit 1200.

The fetal heartbeat extraction unit 2000 uses the separation vector stored in the first storage unit 1900, that is, the fetus from the signal received from the A / D converter 300 using the separation vector obtained from the separation vector extraction unit 1000. Find your heart rate. That is, the fetal heartbeat extraction unit 2000 detects the position of the fetal heartbeat by multiplying the separation vector after constructing the multi-dimensional signal on the time-frequency by the signal received from the A / D converter 300. In addition, the present invention can obtain the fetal ECG complex through the template technique from the detected position information of the heart rate. The fetal heartbeat extractor 2000 includes a second time-frequency domain converter 2100 and a fetal heartbeat detector 2200.

The second time-frequency domain converter 1100 reconstructs a signal received from the A / D converter 300 into a multidimensional signal on the time-frequency domain. That is, the second time-frequency domain converter 1100 reconstructs a single channel signal into a multidimensional signal on the time-frequency domain. At this time, the signal input from the A / D converter 300 to the second time-frequency domain converter 1100 is simultaneously stored in the second storage unit 2900.

The fetal heartbeat detection unit 2200 reads a separation vector stored in the first storage unit, multiplies the separation vector by a signal reconstructed into a multi-dimensional signal on the time-frequency domain from the second time-frequency domain converter 1100, and stores the fetus. Detect the location of the heart rate.

The second storage unit 2900 simultaneously stores signals input from the A / D converter 300 to the second time-frequency domain converter 1100. That is, the original signal is stored in the second storage unit 2900, and the second time-frequency domain converter 1100 reconstructs the signal stored in the second storage unit 2900 into a multidimensional signal on the time-frequency. same.

The fetal ECG complex calculator 3000 uses fetal ECG complexes through a template technique by using the original signal (the maternal abdominal electrocardiogram) from the second storage unit 2900 and the positional information of the fetal heartbeat from the fetal heart rate detection unit 2200. Obtain

The display unit 500 displays an output signal from the fetal ECG separation module 400, that is, fetal ECG complex, fetal health information, fetal heart rate, and the like.

3 is a flowchart illustrating an example of a first time-frequency domain converter of FIG. 2.

The first time-frequency domain converter 1100 receives a signal from the A / D converter 300 for an initial predetermined time (S100). In the case of FIG. 3, the first time-frequency domain converter 1100 receives a signal of 512 points from the A / D converter 300 during an initial predetermined time.

Multi-dimensional in the time-frequency domain by applying the Smooth Pseudo Wigner-Ville Distribution (hereinafter referred to as SPWVD) function from the signal received from the A / D converter 300 during the predetermined time. Reconstruct the signal (S200). In the present invention, the frequency section is set to 256, and thus, in FIG. 3, a 256 × 512 multidimensional virtual sensor on the time-frequency domain can be obtained.

A time axis signal of 512 points for 256 frequencies is extracted from the multi-dimensional signals reconstructed on the time-frequency domain (S300).

The present invention reconstructs a single channel signal received from the A / D converter 300 through the first time-frequency domain converter 1100 into a multidimensional signal on time-frequency, which is independent to extract fetal heartbeats. This is because it is necessary to apply the element analysis technique.

In general, in order to apply the independent element analysis technique to a single channel signal, it is necessary to make it into a multi-channel signal. There are two methods for configuring a multi-channel, which is configured in the time domain and in the time-frequency domain. . The construction method in the time domain is to apply a transform to the original signal to create a new signal that is statistically independent of the original signal while preserving the information of the mixed signal of the original signal. There is a method of generating a new signal by creating and adding a simulation signal having information. However, there is almost no research on transforms or simulation signals that generate appropriate new signals, and there is a limit to generating as many transforms or simulation signals in order to create a plurality of channels. On the other hand, in the time-frequency domain, there is an advantage that a desired number of signals of a frequency band can be selected.

In the present invention, a smooth pseudo Wigner-Ville distribution (SPWVD) is used to project a single channel signal to a multichannel signal in the time-frequency domain.

The following describes the smooth pseudo-Wigner-Ville distribution (SPWVD) function.

According to the Wiener-Khinchin definition, the power spectrum is the Fourier transform of the correlation function (Equation 1).

The correlation function R (τ) is defined as Equation 2 as an average value of x (t) x ^* (t-τ), and information on time cannot be known as an integral value for the entire time. To find the change in frequency of the frequency component, add the time variable t to the R (τ) term in Equation 2 and replace it with R (t, τ) to obtain the frequency component of the signal for a specific time zone. It is defined as Wigner-Ville Distribution (hereinafter referred to as WVD) in Equation 3.

Whereas Short Time Fourier Transform (STFT) determines the time-frequency resolution by the length of the window function, WVD does not use the window function, so it is not affected by window length and provides higher resolution. However, in WVD, cross interference occurs when there are signals with two or more different frequency components. In order to remove this, a second low pass filter φ (τ, t) is applied to the WVD on the time and frequency axis. The low pass filter is called a kernel, and the present invention uses a kernel such as Equation 4. Implement SPWVD.

In this case, h (·) is a window function for removing interference on the frequency axis, and G (·) is a window function for removing interference on the time axis. SPWVD eliminates interference by taking independent windows on the time and frequency axes, and it is possible to implement filtering to match the characteristics of the signal according to the choice of window length or type, and to achieve fast calculation speed.

In the present invention, in order to improve the implementation speed, x is divided by 512 points in the time-frequency domain, and the time interval and the frequency interval are set to 256. As a result, a 256 × 256 multidimensional virtual sensor in the time-frequency domain can be obtained from a single channel 512 point signal in the time domain. In this case, the acquired virtual signal may be used as an input for independent element analysis (ICA), and the dimension may be reduced to 16th order, and then 16 estimated signal sources may be output.

4 is a connection diagram illustrating an application of multidimensional signals and independent element analysis from a single channel signal. That is, the multi-channel signal is generated from the single channel signal detected by the electrode unit 100 through the first time-frequency domain converter 1100 and the independent multi-dimensional analysis is applied to the generated multi-dimensional signal.

Entropy maximization, mutual information minimization, and maximum likelihood methods are used to make signal components independently in ICA. Although the Kullback-Liebler divergence minimization method is used, the maximum and minimum kurtosis is widely used in the present invention, which is very simple, fast, and effective in analyzing independent components. We use a law-based fast fixed-point algorithm.

FIG. 5 is a flowchart for describing the blind signal separation unit of FIG. 2.

 The first time-frequency domain converter 1100 receives a virtual sensor signal which is a signal reconstructed into a multi-dimensional signal on the time-frequency (S400). That is, as shown in FIG. 3, when the first time-frequency domain converter 1100 receives a signal of 512 points from the A / D converter 300 during an initial predetermined time, the frequency section is 256 in the present invention. Since the first time-frequency domain converter 1100 obtains a 256 × 512 multidimensional virtual sensor signal X on the time-frequency domain, the first multi-time domain converter 1100 obtains the multidimensional virtual signal. 1200 receives.

The number of estimated signal sources is set (S550). That is, it determines how many signals are mixed in the virtual sensor signal. In this case, the dimension may be reduced to 16 orders and then 16 estimated signal sources may be output.

)을 결정한다(S600). 즉, 독립요소해석 기법을 통해 서로 독립적인 신호원을 추정한다.Estimated signal source

Determine (S600). In other words, independent signal sources are estimated through independent element analysis.

로 부터 구한다.The separation matrix W is determined (S700). That is, a vector that separates the estimated signal source from the virtual sensor signal.

Obtain from.

A target signal source is selected (S800). Of the 16 estimated signal sources, an estimated signal source that best represents the fetal ECG information is selected.

).The separation vector D is determined (S900). From the separation matrix, select the row from which the target signal source is calculated to determine the separation vector (

).

Next, the blind signal separation will be described in more detail.

Blind signal separation is a technique for recovering signal sources from a mixed measurement signal when the signal sources are linearly mixed. If information about the source and the mixed structure is unknown, only mutual independence between the signal sources is obtained. Solve the problem by assuming

Assuming that m independent signal sources are s and n linearly mixed signals are x, it can be expressed as follows.

However, A is an unknown mixed matrix, and in general, it is assumed that m≥n since the number of sensors must be equal to or greater than the number of signal sources. The goal of blind signal separation is to estimate the signal source s and the mixing matrix A when only the measured sensor x is known, so that the elements of the output signal vector x are statistically independent of each other for source separation. Find it.

Signals can be estimated through this process, but there are ambiguity problems with scaling and permutation.

As a method for this blind signal separation, principal component analysis (hereinafter referred to as PCA) has been proposed, but the performance of PCA is highly dependent on the orthogonality of the columns of the mixing matrix. In the case of an electrocardiogram, the source signals generated from the heart do not satisfy the spatially orthogonal conditions of each other. Therefore, the PCA is not suitable for separating ECG elements because it does not yield as good a result as separating each source from ECG through ICA, but also assumes that the source is a Gaussian distribution. not. In addition, in order to measure the independence of non-Gaussian signals more accurately, HOS (High Order) used by Independent Component Analysis (hereinafter referred to as ICA) rather than SOS (Second Order Statistic) used by PCA. Statistic is more effective. The ICA not only has no limitation on the geometrical structure of the mixed matrix, but also considers the non-Gaussian characteristics of the source signal. For this reason, the present invention in the blind signal separation unit uses the independent element analysis (ICA).

6 is a flowchart illustrating the fetal heartbeat extraction unit of FIG. 2.

The second time-frequency domain converter 2100 receives a signal from the A / D converter 300 for a predetermined time (S1000). In the case of FIG. 6, the second time-frequency domain converter 2100 receives a signal of 512 points from the A / D converter 300 for a predetermined time.

The second time-frequency domain conversion unit 2100 performs a Smooth Pseudo Wigner-Ville Distribution (hereinafter referred to as SPWVD) from a signal received from the A / D converter 300 during the predetermined time. Reconstruct the multi-dimensional signal on the time-frequency domain by applying a function (S1100). In the present invention, the frequency range is set to 256, and thus, in FIG. 6, a 256 × 512 multidimensional virtual sensor on the time-frequency domain can be obtained.

The fetal heartbeat detection unit 2200 is a separation vector received from the first storage unit in a 256 × 512 multidimensional virtual sensor on the time-frequency domain obtained from the second time-frequency domain converter 2100. Multiplying to generate a detection signal containing the fetal ECG information (S1200).

The detection signal is rectified to eliminate the downward peak (S1300).

The threshold is set to detect a peak that is greater than or equal to the threshold (S1400).

If it is determined that the detected peak is actually within the normal rhythm range, it is not detected if it appears earlier than the refractory, and if it does not appear for a long time, the threshold is adjusted downward (S1500 and S1600).

The detected peak is regarded as the fetal heartbeat position (S1700), and the process is repeated until there is no more input signal (S1800).

In more detail about the separation vector, the separation vector includes information on a signal to be detected when a signal source is estimated using an ICA input as a virtual signal generated by projecting an initial single-channel maternal abdominal electrocardiogram into a time-frequency domain. Is defined as a vector that can extract the estimated signal source including the most, can be expressed as shown in Equation 6.

However, S _t is an estimated target signal source, S _v is a virtual signal, D is a separation vector, and the separation vector belongs to the separation matrix W.

FIG. 7 is a flowchart illustrating the fetal ECG complex calculation unit of FIG. 2.

The fetal ECG complex calculating unit 3000 sets a peak region through a threshold in the detection signal received from the fetal heartbeat detection unit 2200 (S3000).

The same area as the detection signal is set in the original signal received from the second storage unit 2900 (S3100).

The peak corresponding to the R point is detected in the region of the original signal (S3200).

A predetermined template region is set before and after the R point which is the detected peak (S3300). At this time, the template region may vary depending on the sampling rate or the maturity of the fetus. For example, if the abdominal electrocardiogram at 37 weeks of pregnancy is sampled at 600 Hz, the template areas before and after the R point can be 65 and 70 points, respectively.

The slope of the template start value and the end value is calculated to determine whether the baseline exists (S3400), and if the baseline exists, it is discarded without being used for the representative template (S3450).

A representative template is configured by averaging the last 8 bits where no baseline exists (S3500).

Information necessary for detecting and diagnosing feature points is calculated from the representative template (S3200).

8 is a flowchart schematically illustrating a method of separating fetal ECG from a maternal abdominal electrocardiogram according to a preferred embodiment of the present invention.

Maternal abdominal electrocardiogram is detected on the maternal abdomen as a single channel and abdominal electrocardiogram data input is started through the A / D converter 300 (S150).

In operation S620, it is determined whether a separation vector exists in the first memory storage unit 1900.

If the separation vector does not exist in the first memory storage unit 1900, the smooth pseudo-wigg from the signal received from the A / D converter 300 for a predetermined time in the first time-frequency domain converter 1100. The non-ville distribution (SPWVD) function is applied to reconstruct the multidimensional signal on the time-frequency domain (S650).

The implicit signal separation unit 1200 applies an independent element analysis (ICA) to the multi-dimensional signal on the time-frequency domain reconstructed by the first time-frequency domain converter 1100 (S850) and sets a separate vector (S900). .

If there is a separation vector in the first memory storage unit 1900, the smooth pseudo-weigner from the signal received from the A / D converter 300 for a predetermined time in the second time-frequency domain converter 2100. Reconstruct the multi-dimensional signal on the time-frequency domain by applying the -bill distribution (SPWVD) function (S1050).

The fetal heartbeat detection unit 2200 multiplies the multi-dimensional signal obtained from the second time-frequency domain transforming unit 2100 by the separation vector received from the first storage unit (S1050), so that the bits of the fetal ECG, that is, the R point ( Heart rate) (S1750).

The fetal ECG complex calculation unit 3000 calculates a fetal heart rate change rate (S3350) and configures a fetal ECG template (S3550). This provides the fetal health state information through the display unit 500.

FIG. 9A is an example of maternal abdominal electrocardiogram and smoothed pseudo Wigner-Ville distribution (SPWVD) function application results using the fetal ECG separation device from the single-channel abdominal electrocardiogram of FIG. 2, and FIG. 9B is the smoothed pseudo diagram of FIG. 9A. It is an example of estimated signal sources obtained by applying the Independent Element Analysis (ICA) in the Wigner-Ville distribution (SPWVD) function.

9A and 9B show a result of obtaining an estimated signal source by projecting an actual maternal data into the time-frequency domain and applying an independent element analysis (ICA). In this case, the 14th estimated signal source (indicated in red in FIG. 9B) best represents the fetal ECG, and a vector for generating the 14th signal from the virtual signal is set as the separation vector. For data after the initial 512 points, fetal ECG was detected by multiplying the separation vector by the signal projected in the time-frequency domain via SPWVD.

FIG. 10A is an example of maternal abdominal electrocardiogram and fetal ECG bit detection results using a separation vector when using the fetal ECG separation apparatus from the single-channel abdominal electrocardiogram of FIG. 2. FIG. 10B is a maternal electrocardiogram and a fetus's ECG diagram of FIG. 10A. This is an example of overlapping cases.

10A and 10B, the apparatus and method for fetal ECG separation from a single channel abdominal ECG of the present invention can detect 100% fetal ECG bits in actual maternal data. In particular, the present invention was able to find the fetal ECG bit (R point) where the maternal ECG and the fetal ECG overlap or are difficult to visually identify, as shown by the arrow in FIG. 10B. Therefore, fetal heart rate change rate can be calculated by calculating the RR interval from fetal ECG bits (R points) detected from Kendo 10A and FIG. 10B, and a representative template can also be obtained.

FIG. 11A is an example of fetal heart rate change rate calculated from detected fetal ECG bits when using the fetal ECG separation device from the single channel abdominal electrocardiogram of FIG. 2, and FIG. 11B is an example of a representative template obtained from FIG. 11A.

11A and 11B show fetal heart rate change rates and representative templates calculated from detection results of fetal ECG bits (R point, heart rate) of FIG. 10A. In the case of Figs. 11A and 11B, the fetus is almost mature and has not only the QRS complex but also P- and T-waves. The length of the template is 65 points and 70 points, respectively, based on the R point to 136 points (≒ 227 ms). It was.

Thus, fetal ECG separation device and method from the maternal abdominal electrocardiogram of the present invention detects fetal ECG bits from the single-channel maternal abdominal electrocardiogram, through which fetal heart rate change and morphology information can be obtained and the fetal health status Basic data on evaluating the data are provided.

The present invention is not limited to the above described and illustrated in the drawings, and of course, more modifications and variations are possible to those skilled in the art within the scope of the following claims.

As described above, the fetal ECG separation apparatus and method from the maternal abdominal electrocardiogram according to the present invention can detect the fetal heartbeat by using a single channel signal obtained from the maternal abdomen, and the present invention is a smooth pseudo-Wigner Apparatus and method for reconstructing fetal ECG from maternal abdominal electrocardiogram by reconstructing multidimensional signals on time-frequency by applying the -Bill distribution (SPWVD) function and then applying independent element analysis techniques.

The method and apparatus for obtaining fetal ECG bits from the single channel abdominal ECG of the present invention is applicable to fetal health monitoring systems.

In the present invention, when the application of the smooth pseudo Wigner-Ville distribution (SPWVD) function is implemented in an ARM processor or a DSP chip, fetal ECG can be detected in real time from a single channel abdominal electrocardiogram and the fetal health state can be provided. The system can be implemented.                     

The present invention is also useful as a new algorithm for detecting fetal ECG from a single-channel abdominal electrocardiogram mixed with maternal ECG and fetal ECG measured in the mother's abdomen.

Claims (17)

An electrode unit for detecting a biological signal from a mother's abdomen; A signal preprocessor for amplifying a biosignal detected from the electrode and removing noise; An A / D converter for converting the output signal of the signal detector into a digital signal; A fetal ECG separation module unit for extracting a fetal heartbeat from a signal for a predetermined time received from the A / D converter; In the fetal ECG separation apparatus from a single channel maternal abdominal electrocardiogram provided with; Display for displaying the output signal from the fetal ECG separation module unit, The fetal ECG separation module unit  A separation vector extracting unit that obtains a demixing vector from a signal received from the A / D converter during an initial predetermined time period; A fetal heartbeat extracting unit for finding the position of the fetal heartbeat from the signal received from the A / D converter using the separation vector obtained from the separation vector extracting unit; Fetal ECG separation apparatus from a single channel maternal abdominal electrocardiogram characterized in that it comprises at least. The method of claim 1, The fetal ECG separation module unit The fetal ECG complex calculation unit is further provided to obtain a fetal ECG complex using a template technique using the maternal abdominal electrocardiogram, which is the original signal received from the A / D converter, and the fetal heart rate from the fetal heart rate extractor. Fetal ECG separation apparatus from a single channel maternal abdominal electrocardiogram. The method of claim 1, The separation vector extraction unit A first time-frequency domain converter reconstructing a signal received from the A / D converter into a multidimensional signal on a time-frequency domain; An implicit signal separation unit for obtaining a separation vector of the signal reconstructed into a multidimensional signal on the time-frequency domain by the first time-frequency domain conversion unit through a blind signal separation method; A first storage unit for storing the separation vector obtained from the blind signal separation unit; Fetal ECG separation apparatus from a single channel maternal abdominal electrocardiogram comprising a. The method of claim 1, The fetal heart rate extraction unit A second time-frequency domain converter reconstructing a signal received from the A / D converter into a multidimensional signal on a time-frequency domain; A fetal heartbeat receiving the separation vector extracted from the separation vector extracting unit and multiplying the separation vector by a signal reconstructed into a multi-dimensional signal on the time-frequency domain from the second time-frequency domain converter to detect the position of the fetal heartbeat. Detection unit; Fetal ECG separation apparatus from a single channel maternal abdominal electrocardiogram comprising a. The method according to claim 3 or 4 The first time-frequency domain converter and the second time-frequency domain converter Apparatus for separating fetal ECG from a single channel maternal abdominal electrocardiogram characterized by reconstruction with multidimensional signals on the time-frequency domain by applying the smooth pseudo-Wigner-Ville distribution (SPWVD) function. An abdominal electrocardiogram detection step of detecting an electrocardiogram signal from the mother's abdomen; A signal preprocessing step of amplifying a biosignal detected in the abdominal electrocardiogram detecting step and removing noise; An A / D conversion step of converting the output signal of the signal preprocessing step into a digital signal; A separation vector extracting step of obtaining a demixing vector from a signal received from the A / D conversion step for a predetermined time; A fetal heartbeat extraction step of finding the position of the fetal heartbeat from the signal received from the A / D conversion step using the separation vector obtained from the separation vector extraction step; A fetal ECG complex calculating step of obtaining a fetal ECG complex using maternal abdominal ECG, which is the original signal received from the A / D conversion step, and fetal heartbeat position information from the fetal heart rate extraction step; Fetal ECG separation apparatus from a single channel maternal abdominal electrocardiogram characterized in that it comprises at least. The method of claim 6, The separation vector extraction step A first time-frequency domain conversion step of reconstructing a signal received from the A / D converter into a multidimensional signal on a time-frequency domain; An implicit signal separation step of obtaining a separation vector of a signal reconstructed into a multidimensional signal on the time-frequency domain in the first time-frequency domain conversion step by a blind signal separation method; Fetal ECG separation method from a single channel maternal abdominal electrocardiogram characterized in that it comprises a. The method of claim 6, The fetal heart rate extraction unit A second time-frequency domain conversion step of reconstructing the signal received from the A / D conversion step into a multidimensional signal on the time-frequency domain; The separation vector extracted from the separation vector extraction step is received, and the signal reconstructed into the multi-dimensional signal on the time-frequency domain from the second time-frequency domain conversion step is multiplied by the separation vector to detect the position of the fetal heartbeat. Detecting fetal heart rate; Fetal ECG separation method from a single channel maternal abdominal electrocardiogram characterized in that it comprises a. The method according to claim 7 or 8 The first time-frequency domain conversion step and the second time-frequency domain conversion step A method of separating fetal ECG from a single channel maternal abdominal electrocardiogram characterized by reconstructing a multidimensional signal on the time-frequency domain by applying a smooth pseudo-Wigner-Ville distribution (SPWVD) function. The method of claim 7, wherein The blind signal separation step A fetal ECG separation method from a single channel maternal abdominal electrocardiogram characterized by using independent element analysis (ICA). The method of claim 7, wherein The blind signal separation step A fetal ECG separation method from a single channel maternal abdominal electrocardiogram characterized by using a fast fixed-point algorithm in independent element analysis (ICA). The method of claim 7, wherein The blind signal separation step, A virtual signal receiving step of receiving a virtual sensor signal which is a signal reconstructed into a multi-dimensional signal on a time-frequency domain from a first time-frequency domain converter; An estimated signal number setting step of setting the number of estimated signal sources; Estimated signal source through independent factor analysis

Determining an estimated signal source; A separation matrix determining step of determining a separation matrix W; A target signal source determining step of selecting a target signal source; A separation vector determination step of determining a separation vector D; Fetal ECG separation method from a single channel maternal abdominal electrocardiogram characterized in that it comprises a. The method of claim 8, The second time-frequency domain conversion step A signal input step of receiving a signal from the A / D conversion step for a predetermined time; A time-frequency phase reconstruction step of reconstructing a multi-dimensional signal in the time-frequency domain by applying a smooth pseudo Wigner-Ville distribution (SPWVD) function from the signal received from the A / D conversion step for the predetermined time; Fetal ECG separation method from a single channel maternal abdominal electrocardiogram characterized in that it comprises a. The method of claim 13, The fetal heart rate detection step A detection signal generation step of generating a detection signal including fetal ECG information by multiplying the separation vector by a multi-dimensional virtual signal obtained from the second time-frequency domain conversion step; A downward peak elimination step of rectifying the detection signal to eliminate downward peaks; Detecting a peak above a threshold for detecting a peak that is greater than or equal to a predetermined threshold; A threshold adjustment step of determining whether the detected peak is actually within a normal rhythm range and not detecting if it appears earlier than the refractory, and lowering the threshold if it does not appear for a long time; Fetal heartbeat position estimation step of estimating the detected peak as fetal heartbeat position; Fetal ECG separation method from a single channel maternal abdominal electrocardiogram comprising a. The method of claim 14, The fetal ECG complex calculation step A peak region setting step of setting a peak region through a threshold in the detection signal received from the fetal heartbeat detection step; An original signal region setting step of setting the same region as the detection signal in the maternal abdominal electrocardiogram which is the original signal received in the A / D conversion step; An R point detecting step of detecting a peak corresponding to the R point in the region of the original signal; A template region setting step of setting a predetermined template region before and after the detected R point; A representative template constructing step of determining whether a baseline exists by calculating slopes of the template start value and the end value, and constituting a representative template by averaging the last 8 bits where no baseline exists; A feature point detection step of calculating information required for feature point detection and diagnosis from the representative template; Fetal ECG separation method from a single channel maternal abdominal electrocardiogram comprising a. A data input step of detecting maternal abdominal electrocardiogram from a maternal abdomen in a single channel and starting abdominal electrocardiogram data input through an A / D converter; A determination step of determining whether a separation vector exists; If the separation vector does not exist, the independent vector is reconstructed into a multi-dimensional signal in the time-frequency domain by applying the smooth pseudo Wigner-Ville distribution (SPWVD) function from the signal received from the A / D converter for a predetermined time. A separation vector setting step of applying a component analysis (ICA) to set a separation vector; If the separation vector exists, the multi-dimensional signal reconstruction step of reconstructing the multi-dimensional signal in the time-frequency domain by applying the smooth pseudo Wigner-Ville distribution (SPWVD) function from the signal received from the A / D converter for a predetermined time ; A bit detection step of detecting a bit (R point, heart rate) of the fetal ECG by multiplying the separation vector by the multidimensional signal reconstructed in the multidimensional signal reconstruction step; Calculating a fetal heart rate rate and configuring the fetal ECG template; Fetal ECG separation method from a single channel maternal abdominal electrocardiogram characterized in that it comprises a. An abdominal electrocardiogram detection step of detecting an electrocardiogram signal from the mother's abdomen; An A / D conversion step of converting the output signal of the abdominal electrocardiogram detection step into a digital signal; A time-frequency phase reconstruction step of reconstructing a multi-dimensional signal in the time-frequency domain by applying a smooth pseudo Wigner-Ville distribution (SPWVD) function from the signal received from the A / D conversion step for the predetermined time; A detection signal generation step of generating a detection signal including fetal ECG information by multiplying a separation vector by a multi-dimensional signal on the time-frequency domain obtained from the second time-frequency phase reconstruction step; A downward peak elimination step of rectifying the detection signal to eliminate downward peaks; Detecting a peak above a threshold for detecting a peak that is greater than or equal to a predetermined threshold; A threshold adjustment step of determining whether the detected peak is actually within a normal rhythm range and not detecting if it appears earlier than the refractory, and lowering the threshold if it does not appear for a long time; Fetal heartbeat position estimation step of estimating the detected peak as fetal heartbeat position; Fetal ECG separation apparatus from a single channel maternal abdominal electrocardiogram comprising a.

Images