JP4644268B2 - Shunt condition detector - Google Patents

Description

  The present invention relates to a shunt state detection device, and more particularly to a shunt state detection device for detecting a change in the state of a shunt that is a part bypassing a vein and an artery of a patient.

  The method of purifying the blood by drawing the blood of a patient with renal insufficiency whose renal function has declined or stopped, and exchanging the substance between the blood and the dialysate via a dialysis membrane is hemodialysis or artificial dialysis. Well known as.

  In this hemodialysis, blood access is usually performed by introducing a needle or a catheter into a blood vessel. In this case, the blood access is composed of one or more needles or catheters, and blood is passed through an arterial needle. After being taken out into the blood circuit and purified, it is returned to the body again via the venous needle.

  However, for the blood access for chronic maintenance blood purification, a standard internal shunt method is used instead of temporary catheter placement and direct arterial puncture as in the acute phase.

  In this method, depending on normal needle puncture, blood flow is insufficient with venous blood alone to extract blood flow of 150 to 250 ml / min. Therefore, the radial artery is one of the arteries located deep in the body. And the cephalic vein located on the body surface side are anastomosed or end-to-side anastomosed to form a so-called inner shunt and puncture this part to take out a necessary amount of blood from the body.

  In such a method using an inner shunt, repeated puncture into the shunt over a long period of time gradually narrows the arteries and veins, or causes a blood access trouble.

  To detect this access trouble, for example, a stethoscope generates a shunt sound that is generated when a large amount of blood rapidly passes through the anastomosis between the radial artery of the inner shunt and the cephalic vein and vibrates the blood vessel wall. It is done with tracking.

  That is, the magnitude of the shunt sound is related to the blood volume, and the frequency of the shunt sound is related to the inner diameter of the blood vessel. When the inner diameter is large, the sound is low, and when the inner diameter is small, the sound is high. It can be seen that narrowing occurred.

  Also, if the sound accompanying the pulsation is heard as a relatively continuous sound, the elasticity of the vein wall and the diameter of the arteriovenous anastomosis are maintained normally, and conversely if the sound is heard as a short discontinuous sound Venous sclerosis is strong or the diameter of the anastomosis part is narrow.

  Furthermore, when a wide variety of sounds can be heard, it is considered that turbulence has occurred due to irregularities in the inner wall of the blood vessel or bending of the access blood vessel itself.

  Thus, diagnosis with a stethoscope can be easily performed and a lot of important information can be obtained, so if it is performed carefully, it can be up to 70 to 80% of the blood access function evaluation, and the state of the shunt portion is to some extent. There is an advantage that it can be grasped.

  However, this diagnosis by auscultation requires medical staff to continuously obtain information on the same patient over a long period of time, so if the person in charge is replaced by another staff due to an accident, etc., it is necessary to arrange the information from the beginning. In addition, since the evaluation content based on the diagnosis results depends on the ability and experience of each staff member, even if the same shunt sound of the same patient is heard, it is inevitable that a difference occurs in the blood access evaluation.

Therefore, a method has been proposed in which shunt sound signals are wavelet transformed to image temporal changes in frequency components of shunt sounds (see, for example, Patent Document 1).
JP 2005-40518 A

  However, even if the shunt sound signal is imaged by wavelet transform, when judging the change over time of the shunt sound, humans will judge the difference in the images, and it cannot be said that quantitative judgment is always possible. There was a problem.

  In addition, although there are certain rules such as heartbeat, there is a problem that an error occurs in the determination of the difference when comparing data with variations in period.

  The present invention has been made in view of these points, and an object of the present invention is to provide a shunt state change detection device that can quantitatively determine a change over time in a state in the shunt.

  In the present invention, in order to solve the above problem, in a shunt state detection device for detecting a change in the state of a shunt that is a site bypassing a patient's vein and artery, the sound of blood flow flowing through the shunt is collected from the sound collector. An input unit that receives the signal, an image generation unit that generates a wavelet image by performing wavelet transform on the blood flow sound signal received by the input unit, and a heartbeat from the blood flow signal received by the input unit. A period detection unit that detects one cycle corresponding to the above, a cutout unit that cuts out the one cycle from the wavelet image, an image storage unit that stores the image of the cut out one cycle in association with the patient, and When an image is received from the clipping means, the patient image stored in the image storage means is read out and one of the images is stored. And processing means for processing the image to match the period of the other image in the period, the possible and a calculating means for calculating a correlation between the two images shunt state detection apparatus according to claim is provided.

  Thereby, the input unit receives the sound of the blood flow flowing through the shunt from the sound collecting device, and the image generation unit generates a wavelet image by performing wavelet transform on the signal of the blood flow sound received by the input unit. Further, the period detecting means detects one period corresponding to one heartbeat from the blood flow sound signal received by the input unit. The cutout unit cuts out the one period from the wavelet image, and the image storage unit stores the cut out image for the one period in association with the patient. Then, when the processing means receives the image from the cutout means, the image of the patient stored in the image storage means is read, and the image is adjusted so that the period of one of the images matches the period of the other image. And the calculating means calculates the correlation between the two images.

  Further, in the present invention, in the shunt state detection device for detecting a change in the state of the shunt that is a part bypassing the vein and artery of the patient, an input unit that receives sound of blood flow flowing through the shunt from the sound collection device; Image generation means for generating a wavelet image by wavelet transforming the blood flow sound signal received by the input unit, and one cycle corresponding to one heartbeat from the blood flow sound signal received by the input unit A period detection means for detecting the image, a cutout means for cutting out the one period from the wavelet image, an image storage means for storing the extracted image for the one period in association with the patient, and an image from the cutout means. When received, the image of the patient stored in the image storage means is read out, and the image of the other image is read in one image period of both images. And processing means for processing the image to match the period, the calculated a histogram for each image of the two images, the shunt state detection device, characterized in that it comprises a comparing means for comparing is provided.

  Thereby, the input unit receives the sound of the blood flow flowing through the shunt from the sound collecting device, and the image generation unit generates a wavelet image by performing wavelet transform on the signal of the blood flow sound received by the input unit. Further, the period detecting means detects one period corresponding to one heartbeat from the blood flow sound signal received by the input unit. The cutout unit cuts out the one period from the wavelet image, and the image storage unit stores the cut out image for the one period in association with the patient. Then, when the processing means receives the image from the cutout means, the image of the patient stored in the image storage means is read, and the image is adjusted so that the period of one of the images matches the period of the other image. The comparison means calculates and compares histograms for each of the two images.

  According to the shunt state detection device of the present invention, one period is detected from a blood flow sound signal, one period is cut out from the wavelet image, and the cut out image is stored in association with the patient. Then, after processing the image so that the cycle of one image matches the cycle of the other image of both the stored image and the clipped image, the correlation between the two images is calculated. It is possible to appropriately calculate the correlation of images depending on the pulse whose period is not always constant even if it is of the same patient.

That is, since it is possible to quantitatively determine how much both images have changed, it is possible to appropriately determine the state of the shunt.
This eliminates the need for the same person to observe the patient's shunt sound over a long period of time.
Further, since it is not necessary to judge the difference by human eyes, it is possible to eliminate arbitraryness and to make an appropriate judgment.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[First Embodiment]
FIG. 1 is a block diagram of the shunt state detection device of the present embodiment.
As shown in FIG. 1, the shunt state detection apparatus 100 includes an input unit 110 that receives a shunt sound signal flowing in a shunt input from the outside, and an image generation unit that generates a wavelet image by wavelet transforming the shunt sound signal. 120, period detection means 130 for detecting one period of the shunt sound from the shunt sound signal, extraction means 140 for extracting one period from the wavelet image, and image storage means 150 for storing the extracted one period image in association with the patient. A processing unit 160 for processing the image so as to match the cycle of the cut-out image with the cycle of the image stored in the image storage unit 150, and a calculation unit 170 for calculating the correlation between the two images having the same cycle. I have.

  The input unit 110 receives a shunt sound signal flowing in a shunt collected by a sensor (not shown) connected to the shunt state detection device 100. The input unit 110 includes, for example, a filter, an amplifier, and an A / D (Analog / Digital) converter.

  The filter uses a bandpass filter that does not pass sound in a frequency band unnecessary for measurement of shunt sound and passes only sound in a frequency band required for measurement of shunt sound. Here, a band-pass filter that passes only sound having a frequency of 25 Hz to 1600 Hz is used.

  Moreover, it is possible to adjust a gain suitable for measurement by an amplifier, and an analog signal input by an A / D converter can be converted into a digital signal.

  When receiving a shunt sound signal from the input unit 110, the image generation unit 120 performs wavelet transform on the signal to generate a wavelet image.

  When the period detector 130 receives a shunt sound signal from the input unit 110, the period detector 130 removes and averages the noise of the signal, and detects the time between the peaks of the signal as a period. A specific description will be given later.

  When the clipping unit 140 receives the wavelet image from the image generation unit 120 and receives the cycle detected from the cycle detection unit 130, the clipping unit 140 performs a process of cutting out only one cycle in the wavelet image. This clipping process will also be described later.

  When the clipping unit 140 performs the clipping process, the clipping unit 140 stores the clipped image in the image storage unit 150. At this time, the image stored in the image storage means 150 is stored in association with the collected patient identifier or the like. Further, the stored (measured) date and time are stored in association with each other.

  When the processing unit 160 receives the cut-out image from the cut-out unit 140, the processing unit 160 reads the patient image stored in the image storage unit 150. As the read image, an image cut out before the cut out image is read out. Mainly, a sound that has been collected, wavelet transformed, and cut out immediately after forming a shunt that is a reference image is read out.

  The shunt condition is considered to be the best condition immediately after the shunt plastic surgery, so that the intima of the blood vessel wall due to repeated insertion and removal of the needle for artificial dialysis and blood flow to the blood vessel wall vigorously This is because, in order to know the degree of stenosis of the blood vessel due to thickening, it is considered desirable to compare with the best state.

  The processing unit 160 compares the period of the image received from the cropping unit 140 with the period of the image read from the image storage unit 150, and performs reduction or enlargement processing on one or both images so that the periods of both images match. Do. Image processing for both images will also be described later.

The calculating means 170 calculates the correlation between both images that have undergone image processing.
At this time, the image is considered as a vector having the luminance of each coordinate as an element, and the vectors of both images to be compared
,
Then, the normal correlation coefficient R indicating the correlation is calculated by the following equation (1).
……… (1)
here,
,
Represents the average brightness of both images.

  By calculating such a normal correlation coefficient R, anyone using the coefficient R can quantitatively determine the degree of change of the shunt with time.

This eliminates the need for the same person to observe the patient's shunt sound over a long period of time.
Further, since it is not necessary to judge the difference by human eyes, it is possible to eliminate arbitraryness and to make an appropriate judgment.

Next, each process described above will be described in detail with reference to the drawings.
FIG. 2 is a diagram showing a waveform of a shunt sound and its cubic spline curve.
As shown in FIG. 2, it can be expressed as a waveform 200 of a shunt sound collected by a sensor. At this time, the cubic spline curve 210 is as shown in FIG.

  When the period detection unit 130 receives the signal of the shunt sound waveform 200 from the input unit 110, the period detection unit 130 calculates an average value per certain time in the waveform data, and smoothly connects the average value to derive the cubic spline curve 210. .

  The average value of the waveform data per fixed time is the average value of the waveform data in the data unit 220 (indicated by a dotted line in FIG. 2) obtained by dividing the time axis of the waveform 200 (horizontal axis in FIG. 2) per fixed interval. is there. In FIG. 2, it is indicated by a hollow circle.

  When the period detection unit 130 derives the cubic spline curve 210, the period detection unit 130 detects the time T between the peaks 222 and 223 of the cubic spline curve 210 as one period of the shunt sound waveform 200.

  Next, a process in which the cutout unit 140 cuts out the wavelet image generated by the image generation unit 120 will be described in detail with reference to the drawings.

Figure 3 is a wavelet image, a diagram showing one period of the wavelet image, FIG. 3 (a) is a wavelet image of the waveform of the shunt sound period T _1, FIG. 3 (b), the period T It is a wavelet image of the waveform of ₂ shunt sound.

  As shown in FIG. 3, in the wavelet image, the horizontal axis is the time axis and the vertical axis is the frequency axis, and the intensity of the audio signal at the time and frequency is indicated by color shading. In the wavelet image shown in FIG. 3, the whiter the color, the stronger the intensity of the audio signal. Hereinafter, it is assumed that the wavelet images are the same unless otherwise specified.

As shown in FIG. 3, when T ₁ is greater than T ₂ , the width of the image 311 in a portion divided into one period T _{1 in} the wavelet image 310 is equal to one period T in the wavelet image 320. _It becomes larger than the width of the image 321 of the part corresponding to ₂ .

  In order to know the change over time of the shunt sound, when one period of the wavelet image is cut out and compared, the images 311 and 321 cannot be cut out and compared as they are. That is, if the image is left as it is, images of different sizes are compared, and the difference in one cycle cannot be clarified.

Therefore, after the images 311 and 321 are cut out, image processing is performed so that the width of one image matches the width of the other image. In the present embodiment, the reference is the image 321 and the width of the T ₃ image 311 is reduced to T ₂ .

  Next, how the image 312 is extracted from the wavelet image 310 will be described in detail with reference to the drawings.

4, the period is extracted images of one period from the wavelet image T _1, is a diagram showing an image after processed.

As shown in FIG. 4, the period since the period of the waveform of the original and became shunt sound wavelet image 310 is T _1, the period of the waveform of the original and became shunt sound image as a reference is T ₂ It is necessary to match. Therefore, by dividing by T ₁ in the time axis direction, to generate an image 312 of one period by performing the processing of multiplication by T _2.
Since the frequency axis directions are the same, there is no need to perform processing.

Figure 5 is a diagram showing how the cycle extract the image of one period from the wavelet image T _2.
As shown in FIG. 5, since the period of the waveform of the original and became shunt sound wavelet image 320 is T _2, is an image in which one cycle of the image 322 of the wavelet image 320 is a reference, the image No processing is performed on 322.

  As a result, the image sizes for both one cycle of the images 312 and 322 to be compared coincide with each other, and it has become possible to compare how the frequency component has changed in one cycle. .

  Next, how to perform processing for adjusting the period of the image for one period will be described in detail with reference to the drawings.

FIG. 6 is a diagram showing an image for one cycle extracted from the wavelet image, and FIG. 6A is a diagram showing an image for one cycle extracted from the wavelet image with the cycle T ₁ . (b) is a diagram showing an image of one cycle which cycle is extracted from the wavelet image T _2.

As shown in FIG. 6, the width of the image 311 whose period is T ₁ is larger than the width of the image 321 whose period is T ₂ .

In this case, the period T ₁ is assumed to consist of 10 dots in the data, the period T ₂ are assumed to consist 8 dot data. The data at the position with the lowest frequency in the image 311 are 10, 8, 6, 4, 2, 2, 4, 6, 8, and 10, and the data at the position with the lowest frequency in the image 321 are 9, 7, 5, 2, 2, 5, 7, 9.

  Comparing the images 311 and 321 is to compare the data amount of each dot, but it cannot be compared if the number of dots is different as described above. Therefore, the image 311 can be compared by reducing it in the time axis direction.

Next, the reduction processing procedure will be described in detail with reference to the drawings.
FIG. 7 is a diagram showing a procedure of reduction processing.
As shown in FIG. 7, the procedure for reducing the data at the lowest frequency position in the image 311 in the time axis direction will be described.

T ₁ is a 10-dot, since _{T 2} is 8 dots, it is sufficient to reduce the 4/5. Therefore, each dot of the image 311 is first expanded four times (first stage).

  Next, five expanded dots are selected from the ends, and an average value of the five selected dots is calculated. For example, at the left end in FIG. 7, if five dots are selected from the left end, 10, 10, 10, 10, and 8 are selected, and the average value is 9 (second stage).

Then, the calculated average value is reconstructed in order from the left end (third stage).
By performing this operation sequentially in the frequency axis direction, the image 311 can be reduced in the time axis direction.

FIG. 8 is a diagram illustrating both images after the processing.
As shown in FIG. 8, both the images 312 and 322 have the same length in both the time axis direction and the frequency axis direction by performing a reduction process in accordance with the width in one time axis direction.
In this way, the calculation means 170 calculates the correlation coefficient between the images having the same size.

Next, a specific example of the correlation coefficient calculated by the calculation unit will be described in detail with reference to the drawings.
FIG. 9 is a graph in which the correlation coefficients of subjects are arranged in time series.
The shunt sound collection date is circled 1 from the left (hereinafter shown in parentheses for convenience of explanation), but the 8th day (1), 15th day (2), 22nd day (3 ), Day 36 (4), Day 43 (5), Day 50 (6), Day 83 (7), Day 99 (8), Day 104 (9), Day 120 (10) , And 132 days (11).

  In the example shown in this graph, PTA (percutaneous transluminal angioplasty) is performed again on the 78th day from the reference data collection date.

  The reference data collection date is after the first PTA, and the normal correlation coefficient R is calculated from the image at that time and the image of the data collection date.

Regular correlation coefficients R calculated at each sampling date, the value of 8 days from the reference data collection date R _1, the value of the day 15 R _2, · · ·, a value of 132 days and R ₁₁ To do.

  At this time, the normal correlation coefficient R gradually decreases after collecting the reference data. This test subject repeatedly repeats stenosis of about 5 cm toward the central side (heart side) immediately after the anastomosis, and has been improved by PTA each time.

  However, in response to the progress of stenosis again, the shunt sound also changed from a continuous low pitch with mainly low frequency components to a high intermittent frequency with high frequency components mainly corresponding to wavelets. It is considered that the normal correlation coefficient R between the images after the conversion also decreased.

Actually, the blood removal amount during dialysis, which was 250 ml / min at the time of collecting the reference data, gradually decreased, and the blood removal was poor at around 100 to 150 ml / min around day 50 (6). Therefore, when the PTA was performed again on the 78th day from the reference data collection (the 99th day since the previous PTA), R ₇ recovered to the same level as R _{1 on the} 8th day on the 83rd day, Although a slight decrease is observed after the 99th day, the normal correlation coefficient R maintains a high level close to 0.7. In the meantime, the blood removal amount was maintained at 200 ml / min or more, and dialysis was successfully performed.

  From this result, it can be seen that the change over time in the normal correlation coefficient R between the wavelet transform images of the shunt sound signal well represents the change over time in the VA (basic access) function as the stenosis progresses. .

  By the way, from the date of reference data collection to the 50th day (6), even if the auscultation of a shunt using a stethoscope, which is a typical physical finding, has been performed, a shunt sound is clear in auscultation. Unable to identify changes.

  That is, according to the present embodiment, changes in the VA function that cannot be identified by auscultation can be grasped over time and quantitatively, and an angiographic examination is performed after it is determined that there is an abnormality. This contributes to the establishment of a monitoring program for VA function evaluation that can be implemented in an efficient manner.

  In addition, if you set a threshold to determine that the situation is normal, such as a decrease in the normal correlation coefficient over time, the dialysis staff only collects the shunt sound every few minutes before the start of dialysis, It is possible to automatically determine the VA function such as an alarm when an abnormality occurs.

Next, the effect of sensor mounting reproducibility on the normal correlation coefficient R will be described in detail with reference to the drawings.
FIG. 10 is a graph showing the relationship with the normal correlation coefficient when the sensor is repeatedly attached and detached.
As shown in FIG. 10, the maximum fluctuation range of the correlation coefficient was 0.016 for three subjects. On the other hand, the difference between R ₆ and R ₈ in FIG. 9 described above is 0.17, and from this result, the influence of reproducibility of sensor mounting is about 10% at the maximum, which is given to the monitoring of the VA function by this method. It can be confirmed that the impact is small.

[Second Embodiment]
Next, a second embodiment of the present invention will be described. The shunt state detection apparatus of this embodiment is the same as that of the first embodiment except that the comparison unit calculates a histogram and compares the calculated histogram. For this reason, about the component similar to the said 1st Embodiment, the same code | symbol is attached | subjected etc., and the description is abbreviate | omitted suitably.

FIG. 11 is a block diagram of the shunt state detection device of the present embodiment.
As shown in FIG. 11, the shunt state detection device 2100 includes an input unit 110 that receives a shunt sound signal flowing in a shunt input from the outside, and an image generation unit that generates a wavelet image by wavelet transforming the shunt sound signal. 120, period detection means 130 for detecting one period of the shunt sound from the shunt sound signal, extraction means 140 for extracting one period from the wavelet image, and image storage means 150 for storing the extracted one period image in association with the patient. The cut-out image of one period and the processing means 160 for processing the image so as to match the period of the image stored in the image storage means 150, and the histograms of both images having the same period are calculated. Comparing means 2170 for comparing the values of the histogram is provided.

The comparison unit 2170 calculates a histogram of both images that have undergone image processing.
Hereinafter, the process in which the comparison unit 2170 calculates the histogram will be specifically described with reference to the drawings.

FIG. 12 is a diagram illustrating a calculation procedure of a histogram calculated by the comparison unit.
As shown in FIG. 12, the image 312 reduced by the processing means 160 is divided into three equal parts in the time axis direction, and divided into four sections of 25 to 200 Hz, 200 to 400 Hz, 400 to 600 Hz, and 600 to 800 Hz in the frequency axis direction. Divide and set 12 areas as a whole.

  For each area, a circled number from 1 to 12 is set for convenience. Then, the calculation unit 170 calculates a histogram for each area.

  For the image 322 (not shown) to be compared, 12 areas are set in the same way, and a histogram for each area is calculated in the same manner as for the image 312.

  Then, the histograms of the areas with the same numbers are compared. As a result of this comparison, when the waveform level is higher in an area of a higher frequency band, it can be determined that stenosis is progressing.

It is a block diagram of the shunt state detection apparatus of this Embodiment. It is a figure which shows the waveform of a shunt sound, and its cubic spline curve. It is a figure which shows a wavelet image and one period in the wavelet image. Period extracting an image of one period from the wavelet image T _1, is a view showing a state of working. Period is a diagram showing a state of extracting the image of one period from the wavelet image T _2. It is a figure which shows the image for one period extracted from the wavelet image. It is a figure which shows the procedure of a reduction process. It is a figure which shows both the images after performing a process. It is the graph which arranged the test subject's correlation coefficient in time series. It is a graph which shows the relationship between the frequency | count of sampling of shunt sound, and a normal correlation coefficient. It is a block diagram of the shunt state detection apparatus of this Embodiment. It is a figure which shows the calculation procedure of the histogram which a comparison means calculates.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Shunt state detection apparatus 110 Input part 120 Image generation means 130 Period detection means 140 Clipping means 150 Image storage means 160 Processing means 170 Calculation means

Claims (3)

In a shunt state detection device for detecting a change in the state of a shunt that is a site bypassing a vein and an artery of a patient,
An input unit for receiving a sound of blood flow flowing through the shunt from a sound collecting device;
Image generating means for generating a wavelet image by wavelet transforming a blood flow sound signal received by the input unit;
A cycle detecting means for detecting one cycle corresponding to one beat of the heart from a blood flow sound signal received by the input unit;
Clipping means for cutting out the one period from the wavelet image;
Image storage means for storing the extracted image for one period in association with the patient;
Processing means for reading the image of the patient stored in the image storage means upon receiving an image from the clipping means, and processing the image so that the period of one of the images matches the period of the other image; ,
Calculating means for calculating a correlation between the two images;
A shunt state detection device comprising: The image storage means stores the cut-out image for one period in time series,
The calculating means calculates a correlation between the image of the extracted one cycle stored most recently after shunt formation and the image received from the cutting means among the patient images. The shunt state detection device according to claim 1. In a shunt state detection device for detecting a change in the state of a shunt that is a site bypassing a vein and an artery of a patient,
An input unit for receiving a sound of blood flow flowing through the shunt from a sound collecting device;
Image generating means for generating a wavelet image by wavelet transforming a blood flow sound signal received by the input unit;
A cycle detecting means for detecting one cycle corresponding to one beat of the heart from a blood flow sound signal received by the input unit;
Clipping means for cutting out the one period from the wavelet image;
Image storage means for storing the extracted image for one period in association with the patient;
Processing means for reading the image of the patient stored in the image storage means upon receiving an image from the clipping means, and processing the image so that the period of one of the images matches the period of the other image; ,
A comparison means for calculating and comparing a histogram for each image of the two images;
A shunt state detection device comprising: