Myocardial infarction detection system based on diffusion speckles

Abstract

The invention provides a myocardial infarction detection system based on diffusion speckles, and relates to the field of wearable health detection. The system comprises a laser used for exciting a laser signal; the optical fiber probe is connected with the laser through a multimode optical fiber and is used for guiding the laser signal to an object to be detected and collecting a reflected light signal of the object to be detected; the camera is connected with the optical fiber probe through a multimode optical fiber and is used for obtaining a diffusion speckle image based on the reflected light signal; the processing equipment is electrically connected with the camera and used for obtaining an initial blood perfusion index signal and an initial pulse wave signal based on the diffusion speckle image, the processing equipment conducts signal reconstruction on the initial blood perfusion index signal and the initial pulse wave signal to obtain a target blood perfusion index signal and a target pulse wave signal, and the target blood perfusion index signal and the target pulse wave signal are sent to the camera. And myocardial infarction detection is carried out based on the target blood perfusion index signal and the target pulse wave signal.

Description

Technical Field

[0001] This application relates to the field of wearable health monitoring, specifically to a myocardial infarction detection system based on diffuse speckle. Background Technology

[0002] Myocardial infarction (heart attack) is a serious cardiovascular disease caused by acute blockage of the coronary arteries, leading to myocardial ischemia, hypoxia, and even necrosis. It is extremely dangerous, causing not only acute symptoms such as chest pain, shortness of breath, and arrhythmias, but in severe cases, it can even lead to cardiogenic shock and cardiac arrest. Therefore, early detection and warning of myocardial infarction are of paramount importance.

[0003] Currently, wearable devices' real-time detection capabilities can promptly identify potential risk signals such as abnormal electrocardiograms (ECGs) and sudden heart rate changes, thereby improving the prevention and control of cardiovascular diseases. However, their early detection and warning of myocardial infarction still face many problems and limitations. First, the detection accuracy of these devices is limited, especially in complex environments or during user activity, where ECG signals are easily interfered with by noise, potentially leading to false alarms or missed alarms. Second, data processing algorithms are still immature; most devices rely on AI (Artificial Intelligence) models that still exhibit a high false positive rate when identifying ECG features specific to myocardial infarction. Summary of the Invention

[0004] To address the aforementioned problems, this application provides a myocardial infarction detection system based on diffuse speckle, comprising:

[0005] A laser is used to excite a laser signal; a fiber optic probe, connected to the laser via a multimode fiber, is used to guide the laser signal to the object under test and collect the reflected light signal from the object; a camera, connected to the fiber optic probe via a multimode fiber, is used to obtain a speckle image based on the reflected light signal; and a processing device, electrically connected to the camera, is used to obtain an initial blood perfusion index signal and an initial pulse wave signal based on the speckle image. The processing device reconstructs the initial blood perfusion index signal and the initial pulse wave signal to obtain the target blood perfusion index signal and the target pulse wave signal, and performs myocardial infarction detection based on the target blood perfusion index signal and the target pulse wave signal.

[0006] According to an embodiment of this application, the processing device is used to perform empirical mode decomposition on any initial signal, either an initial blood perfusion index signal or an initial pulse wave signal, to obtain multiple first mode components; based on frequency domain kurtosis and energy proportion, the multiple first mode components are filtered to obtain multiple second mode components; variational mode decomposition is performed on the multiple second mode components to obtain multiple third mode components; and signal reconstruction is performed based on the multiple third mode components to obtain a target blood perfusion index signal and a target pulse wave signal.

[0007] According to an embodiment of this application, the processing device is further configured to inject Gaussian noise into an initial signal based on multiple noise figures to obtain multiple signals to be decomposed; perform empirical mode decomposition on the signals to be decomposed to obtain multiple initial mode components of the target order; and obtain a first mode component of the signal to be decomposed based on the average value of the multiple initial mode components to obtain multiple first mode components.

[0008] According to an embodiment of this application, the processing device is further configured to calculate the frequency domain kurtosis and energy percentage of a plurality of first modal components respectively; and to select modal components from the plurality of first modal components whose frequency domain kurtosis is greater than a first preset value and whose energy percentage is greater than a second preset value, so as to obtain a plurality of second modal components.

[0009] According to an embodiment of this application, the processing device is further configured to perform spectral peak search based on multiple second mode components to obtain multiple peak frequencies; perform clustering based on multiple peak frequencies to obtain a center frequency; and perform variational mode decomposition on multiple second mode components based on the center frequency to obtain multiple third mode components.

[0010] According to an embodiment of this application, the processing device is further configured to accumulate multiple third-mode components to obtain a target blood perfusion index signal and the target pulse wave signal.

[0011] According to an embodiment of this application, the processing device is further used to perform joint feature extraction on the target blood perfusion index signal and the target pulse wave signal to obtain correlation features; and to perform myocardial infarction detection based on the correlation features.

[0012] According to an embodiment of this application, the processing device is further configured to divide the target blood perfusion index signal into multiple first sub-signals and the target pulse wave signal into multiple second sub-signals based on a preset window and a preset step size; perform correlation calculation based on the multiple first sub-signals and multiple second sub-signals to obtain a first-order correlation coefficient matrix; normalize the first-order correlation coefficient matrix to obtain a second-order correlation coefficient matrix; and perform matrix multiplication of the first-order correlation coefficient matrix and the second-order correlation coefficient matrix to obtain correlation features.

[0013] According to an embodiment of this application, the processing device is further configured to perform correlation calculations with the first sub-signal and the second sub-signal corresponding to the same window to obtain low-order frequency components; and to obtain a first-order correlation coefficient matrix based on multiple low-order frequency components.

[0014] According to an embodiment of this application, the processing device is further configured to segment the relevant features to obtain multiple feature blocks; encode the multiple feature blocks respectively to obtain multiple bottleneck features; and input the multiple bottleneck features into a classifier to obtain myocardial infarction detection results.

[0015] Based on the principle of speckle diffusion, this application achieves simultaneous detection of blood perfusion index and pulse wave signals, and performs myocardial infarction detection based on the reconstructed signals. On one hand, multimodal detection provides this application with more dimensional information; on the other hand, the signal reconstruction process significantly improves the signal-to-noise ratio and enhances the signal feature recognition, thereby reducing signal interference caused by human activity and external environmental factors. Therefore, the accuracy of the myocardial infarction detection system provided by this application is improved. Attached Figure Description

[0016] The above and other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0017] Figure 1 A schematic diagram of a myocardial infarction detection system based on diffuse speckle according to an embodiment of this application is shown.

[0018] Figure 2 An initial signal processing flowchart according to an embodiment of this application is shown.

[0019] Figure 3 A flowchart of a relevance feature extraction method according to an embodiment of this application is shown.

[0020] Figure 4 A schematic diagram illustrating the principle of obtaining myocardial infarction detection results from correlation features according to an embodiment of this application is shown.

[0021] Figure 5 A flowchart of myocardial infarction detection according to an embodiment of this application is shown.

[0022] Figure 6 The initial blood perfusion signal and initial pulse wave signal of the rabbit measured in an experiment according to an embodiment of this application are shown.

[0023] Figure 7 The experiment according to an embodiment of this application shows multiple first mode components obtained based on the rabbit's initial blood perfusion signal and initial pulse wave signal.

[0024] Figure 8 The experiment according to an embodiment of this application shows multiple second mode components obtained by screening based on frequency domain kurtosis and energy proportion method.

[0025] Figure 9 Multiple third-mode components obtained in experiments according to embodiments of this application are shown.

[0026] Figure 10 The reconstructed signal obtained in an experiment according to an embodiment of this application is shown. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.

[0028] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0029] All terms used herein, including technical and scientific terms, have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0030] Traditional myocardial infarction detection devices only analyze pulse wave signals, such as detecting the amplitude and symmetry of the pulse wave. However, these characteristics often cannot fully reflect the complex morphology of the cardiovascular system, especially changes in microcirculation and hemodynamics.

[0031] Therefore, this application, based on the principle of speckle diffusion, simultaneously measures blood perfusion signals and pulse wave signals, combining the two signals to more comprehensively characterize the dynamic response of the cardiovascular system. Blood perfusion signals reflect key physiological parameters such as microcirculation status, vascular elasticity, and blood flow resistance, while pulse wave signals reflect changes in blood volume and cardiac pumping function. The combined analysis of both improves the accuracy of myocardial infarction detection.

[0032] Figure 1 A schematic diagram of a myocardial infarction detection system based on diffuse speckle according to an embodiment of this application is shown.

[0033] like Figure 1 As shown, the myocardial infarction detection system based on diffuse speckle may include: a laser 10, a fiber optic probe 20, a camera 30, and a processing device 40.

[0034] Laser 10 is used to excite a laser signal; fiber optic probe 20 is connected to the laser via a multimode fiber to guide the laser signal to the object under test and to collect the reflected light signal from the object under test; camera 30 is connected to the fiber optic probe via a multimode fiber to obtain a diffuse speckle image based on the reflected light signal; and processing device 40 is electrically connected to the camera to obtain an initial blood perfusion index signal and an initial pulse wave signal based on the diffuse speckle image. The processing device 40 reconstructs the initial blood perfusion index signal and the initial pulse wave signal to obtain a target blood perfusion index signal and a target pulse wave signal, and performs myocardial infarction detection based on the target blood perfusion index signal and the target pulse wave signal.

[0035] According to embodiments of this application, laser 10 can be a near-infrared semiconductor laser. The emission of light from a near-infrared semiconductor laser is based on stimulated emission of a semiconductor PN junction (Positive-Negative Junction). When a forward bias voltage is applied to the PN junction of the laser, an external current causes holes in the P region and electrons in the N region to migrate towards the active layer. Within the active layer, the number of electrons at high energy levels far exceeds the number of holes at low energy levels. At this time, high-energy electrons, excited by external photons or their own spontaneously emitted photons, transition to lower energy levels and release energy, forming stimulated emission photons. These photons are continuously reflected and amplified within the resonant cavity of the laser, ultimately forming a near-infrared laser with good monochromaticity and strong directionality, which is emitted from the output end of the resonant cavity.

[0036] The laser emitted from laser 10 is further precisely guided by fiber optic probe 20 to the object under test 50, which can be human tissue. Fiber optic probe 20 typically consists of a fiber core and cladding. When the near-infrared laser emitted from laser 10 is incident at the interface between the fiber core and cladding at an angle less than the critical angle, total internal reflection occurs at the interface. This total internal reflection continues along the length of the fiber, allowing the light signal to be transmitted from the incident end to the human tissue under test with almost no loss, thus preventing light diffusion or absorption by the external environment during transmission. Similarly, the reflected light signal after interacting with the object under test 50 is also collected by the fiber optic probe using the same principle of total internal reflection and transmitted to camera 30 through multimode fiber.

[0037] According to an embodiment of this application, camera 30 can be a CCD (Charge-coupled Device) camera. When the reflected light signal transmitted through multimode fiber illuminates the photosensitive unit array of the CCD, the photosensitive unit converts photons into photogenerated charges proportional to the light intensity through the photoelectric effect. The bright and dark speckle grain features correspond to the charge distribution differences at different pixel positions. Subsequently, under the control of the driving circuit, these photogenerated charges are transferred in an orderly manner to the register along the vertical and horizontal directions. The charges transferred to the signal amplifier are amplified and then converted into digital signals by the analog-to-digital converter. These digital signals are then arranged in rows and columns of pixels to generate a digital image, i.e., a diffuse speckle image.

[0038] The diffuse speckle image is further transmitted to the processing device 40. Based on the diffuse speckle image, the processing device 40 performs image processing and calculation to obtain the initial blood perfusion index signal and the initial pulse wave signal.

[0039] According to the principle of speckle diffusion, the speckle contrast value reflects the blood flow perfusion index (BFI) signal, while the photoplethysmography (PPG) signal can be represented by the average speckle intensity. The expressions for the initial blood flow perfusion index signal B(t) and the initial photoplethysmography signal P(t) are shown in Equation (1) and Equation (2), respectively:

[0040] (1)

[0041] (2)

[0042] in The standard deviation of pixel intensity within the diffuse speckle region. This represents the average of the standard deviations of pixel intensity within the diffuse speckle region.

[0043] Since diffuse speckle images are easily affected by uneven intensity distribution of multimode fiber, tissue scattering noise, and human motion artifacts during actual imaging, the processing device 40 will further reconstruct the initial blood perfusion index signal and the initial pulse wave signal. Finally, myocardial infarction detection is performed based on the reconstructed target blood perfusion index signal and target pulse wave signal.

[0044] Based on the principle of speckle diffusion, this application achieves simultaneous detection of blood perfusion index and pulse wave signals, and performs myocardial infarction detection based on the reconstructed signals. On one hand, multimodal detection provides this application with more dimensional information; on the other hand, the signal reconstruction process significantly improves the signal-to-noise ratio and enhances the signal feature recognition, thereby reducing signal interference caused by human activity and external environmental factors. Therefore, the accuracy of the myocardial infarction detection system provided by this application is improved.

[0045] Figure 2 An initial signal processing flowchart according to an embodiment of this application is shown.

[0046] like Figure 2 As shown, the initial signal processing flow includes operations S210~S240.

[0047] In operation S210, for any initial signal among the initial blood perfusion index signal and the initial pulse wave signal, empirical mode decomposition is performed on the initial signal to obtain multiple first mode components.

[0048] In operation S220, multiple first mode components are filtered based on frequency domain kurtosis and energy ratio to obtain multiple second mode components.

[0049] In operation S230, variational mode decomposition is performed on multiple second mode components to obtain multiple third mode components.

[0050] In operation S240, signal reconstruction is performed based on multiple third-mode components to obtain the target blood perfusion index signal and the target pulse wave signal.

[0051] The process of obtaining the first mode component includes the following steps: Gaussian noise is injected into the initial signal based on multiple noise figures to obtain multiple signals to be decomposed; empirical mode decomposition is performed on the signals to be decomposed to obtain multiple initial mode components of the target order; the first mode component of the signal to be decomposed is obtained based on the average value of the multiple initial mode components, so as to obtain multiple first mode components.

[0052] The following will take the processing of the initial blood perfusion index signal B(t) as an example to describe in detail the process of obtaining the first modal component corresponding to B(t).

[0053] First, Q pre-generated Gaussian noises are injected into the initial blood perfusion index signal B(t), resulting in Q signals to be decomposed. For the i-th signal B(t) to be decomposed... i (t), whose expression is shown in formula (3):

[0054] (3)

[0055] In the formula, It is zero-mean Gaussian noise. , This represents the noise figure.

[0056] Subsequently, Empirical Mode Decomposition (EMD) is performed on each noisy signal to obtain multiple initial mode components of the target order, where the j-th initial mode component is denoted as... ,extract The average value of the first initial modal components is taken as the first modal component. Its expression is shown in formula (4):

[0057] (4)

[0058] In the formula, Q is the number of signal groups to be decomposed. This indicates that the residual signal is subjected to the first... Extraction of the initial mode. The initial pulse wave signal P(t) is processed in the same way as described above to obtain the first mode component corresponding to P(t).

[0059] To accurately identify effective signal components and efficiently eliminate noise interference, this application uses a frequency domain kurtosis and energy ratio method to screen the first mode components. The specific steps are as follows: Calculate the frequency domain kurtosis and energy ratio of multiple first mode components; screen mode components with a frequency domain kurtosis greater than a first preset value and an energy ratio greater than a second preset value from the multiple first mode components to obtain multiple second mode components.

[0060] First, based on the frequency domain kurtosis method, a sliding window function is used to filter the first mode component. A sliding time window of appropriate length and step size is set for the first mode component, and the window slides successively along the signal time axis, dividing the long-time signal into several short-time signal segments. Then, the original time-domain first mode component signal is transformed to the frequency domain through a short-time Fourier transform to obtain the power spectral density corresponding to each frequency point. Subsequently, the kurtosis value of each frequency point or frequency interval is calculated. Kurtosis is a statistical measure of the degree to which a signal deviates from a normal distribution. If the signal component conforms to the distribution of stationary Gaussian noise, its kurtosis value is close to 3, while the signal component containing effective information exhibits a kurtosis value much greater than 3. Therefore, by setting a reasonable first preset value, frequency intervals with kurtosis values ​​higher than the first preset value are selected. These intervals are the key frequency bands containing effective signal characteristics. Frequency domain kurtosis... The expression is shown in formula (5):

[0061] (5)

[0062] In the formula, Initial blood perfusion index signal The short-time Fourier transform, This represents the expected value within a time window.

[0063] The energy proportion method involves time-frequency conversion or decomposition of the original first-mode component signal to obtain several signal components representing different frequencies. Then, the energy of each component is calculated, and the ratio of that component's energy to the total energy of the original signal is determined—the energy proportion. Finally, based on a set second preset value, components with energy proportions higher than this value are selected. These components contain the main information from the original signal, while components with low energy proportions are mostly noise or redundant information, thus achieving accurate extraction of effective signal components. Specifically, in the characteristic frequency band... Energy percentage The expression is shown in formula (6):

[0064] (6)

[0065] In the formula, Sampling frequency, power spectral density The estimation is performed using Welch's method, and the calculation formula is as follows:

[0066] (7)

[0067] In the formula, For window functions, H represents the number of segments, and H refers to the window function. The length.

[0068] The first modal components corresponding to the initial blood perfusion index signal and the initial pulse wave signal are filtered according to the above screening method to obtain the corresponding second modal components.

[0069] Spectral peak search is performed based on multiple second-mode components to obtain multiple peak frequencies; clustering is performed based on multiple peak frequencies to obtain center frequencies; variational mode decomposition is performed on multiple second-mode components based on center frequencies to obtain multiple third-mode components.

[0070] According to an embodiment of this application, in order to further extract effective signals, a spectral peak search is performed on the second mode component selected by the above-mentioned dual preset values ​​to obtain a series of peak frequencies. M). Next, the K-means method was used to... Clustering can yield center frequency information. K), The expression is shown in formula (8):

[0071] (8)

[0072] In the formula, This represents the K-means clustering operation. This indicates the number of cluster centers.

[0073] Finally, variational mode decomposition is performed on the second mode component based on the center frequency to extract the third mode component containing the effective signal. .

[0074] The processing device 40 accumulates multiple third-mode components to obtain the target blood perfusion index signal and the target pulse wave signal. Among them, the target blood perfusion signal... The expression is shown in formula (9):

[0075] (9)

[0076] In the formula, D is the number of third-mode components containing valid signals.

[0077] Joint feature extraction is performed on the target blood perfusion index signal and the target pulse wave signal to obtain correlation features; myocardial infarction is detected based on the correlation features.

[0078] The correlation feature extraction steps are as follows: Based on a preset window and a preset step size, the target blood perfusion index signal is divided into multiple first sub-signals, and the target pulse wave signal is divided into multiple second sub-signals; correlation is calculated based on the multiple first sub-signals and multiple second sub-signals to obtain a first-order correlation coefficient matrix; the first-order correlation coefficient matrix is ​​normalized to obtain a second-order correlation coefficient matrix; the first-order correlation coefficient matrix and the second-order correlation coefficient matrix are multiplied together to obtain the correlation features.

[0079] The steps for obtaining the first-order correlation coefficient matrix are as follows: Correlation calculation is performed on the first and second sub-signals corresponding to the same window to obtain low-order frequency components; based on multiple low-order frequency components, the first-order correlation coefficient matrix is ​​obtained.

[0080] Figure 3 A flowchart of a relevance feature extraction method according to an embodiment of this application is shown. Figure 3 As shown, firstly, a sliding window correlation method is applied to the target blood perfusion index signal and the target pulse wave signal respectively, dividing them into several time windows to obtain multiple sub-signals. The preset window length is... To represent a complete pulse cycle, the signal is extracted within one minute. A segmented signal, of which The expression is shown in formula (10):

[0081] (10)

[0082] In the formula, T is the total duration of the original signal. The preset step size for each sliding window. To round down, the sub-signal after dividing the target perfusion index signal is the first sub-signal, and the sub-signal after dividing the target pulse wave signal is the second sub-signal. The first sub-signal of the nth segment is represented as follows: The second sub-signal of the nth segment is represented as .

[0083] Subsequently, the correlation between the target perfusion index signal and the target pulse wave signal was quantified based on Pearson correlation calculation. First, the lower-order frequency component (LoFC) of the first and second sub-signals corresponding to the same window was calculated. The calculation formula is shown in formula (11):

[0084] (11)

[0085] in and They are respectively and The mean.

[0086] Calculate according to the formula respectively The LoFC of the segment signal yields the first-order correlation coefficient matrix as shown in formula (12):

[0087] (12)

[0088] matrix The Middle Column vectors are represented as To each The Middle Columns and After performing Z-score normalization on the signal, the correlation is calculated to obtain... The expression is shown in formula (13):

[0089] (13)

[0090] in and They are respectively and The mean of the correlation coefficients is obtained. Correspondingly, the second-order correlation coefficient matrix (14) is obtained:

[0091] (14)

[0092] Finally, the first-order correlation coefficient matrix is ​​then... and second-order correlation coefficient matrix Perform matrix multiplication to obtain the correlation features (15):

[0093] (15)

[0094] According to an embodiment of this application, the processing device 40 is further configured to segment the relevant features to obtain multiple feature blocks; encode the multiple feature blocks to obtain multiple bottleneck features; and input the multiple bottleneck features into a classifier to obtain the myocardial infarction detection result.

[0095] Figure 4 A schematic diagram illustrating the principle of obtaining myocardial infarction detection results from correlation features according to an embodiment of this application is shown.

[0096] like Figure 4As shown, to reduce data dimensionality, the relevance features are first segmented and projected into the feature space, resulting in multiple feature blocks. After passing through the encoder, these blocks are encoded as bottleneck features and further input into the classifier. The bottleneck features are low-dimensional, highly condensed core feature vectors obtained by the encoder through progressive dimensionality reduction and information compression of the input feature blocks. During the encoding of bottleneck features, the encoder automatically removes redundant information from the original features, retaining only the most critical and representative information for subsequent tasks.

[0097] The classifier mainly consists of a Dropout layer, a fully connected layer, and an output layer. The Dropout layer helps prevent overfitting. After the bottleneck features are input into the classifier, they are first passed to the fully connected layer, where linear transformations and non-linear activation functions are used to initially integrate the features and extract more discriminative intermediate features. Then, they are passed to the Dropout layer, where some neurons are randomly disabled according to a preset probability during the training phase to prevent the network from over-relying on specific neurons and causing overfitting. Subsequently, the features processed by Dropout are passed to subsequent fully connected layers for further in-depth integration and dimensionality adjustment. Finally, the output layer outputs the myocardial infarction detection results, completing the entire process from feature input to detection result output.

[0098] Figure 5 A flowchart of myocardial infarction detection according to an embodiment of this application is shown.

[0099] like Figure 5 As shown, after the processing device acquires the initial blood perfusion index signal and the initial pulse wave signal, it injects Gaussian noise into the initial signals (i.e., the initial blood perfusion index signal and the initial pulse wave signal) and performs empirical mode decomposition to obtain multiple first mode components IMF1(t), IMF2(t)...IMF n (t). Subsequently, the first modal signal is filtered using a dual-preset value method based on frequency domain kurtosis and energy proportion to obtain the second modal component. Next, variational mode decomposition is performed on the second modal component to obtain the third modal component. The third modal component is then reconstructed, and correlation features are extracted. Finally, the myocardial infarction detection result is obtained based on the correlation features.

[0100] To verify the effectiveness of the myocardial infarction detection system based on diffusion speckle in this application, the following rabbit myocardial infarction experimental model was conducted:

[0101] The experiment employed coronary artery ligation to simulate the pathological process of myocardial ischemia and infarction. New Zealand white rabbits weighing 2.5-3.5 kg were selected as experimental animals, as their cardiac anatomy is similar to that of humans, especially the distribution of the left anterior descending coronary artery (LAD), facilitating surgical manipulation and making them sensitive to myocardial ischemia. Before the experiment, the rabbits were fasted for 12 hours, and anesthesia was administered via the marginal ear vein. After complete anesthesia, they were fixed to the operating table. An incision was made in the 3rd-4th intercostal space on the left side of the sternum, the pericardium was carefully cut, and the heart was removed and flushed with heparinized saline to prevent thrombosis. Subsequently, the middle segment of the LAD was ligated with silk sutures to block blood flow, inducing myocardial ischemia and infarction. Postoperatively, the heart was returned to the thoracic cavity, the incision was sutured, and vital signs such as heart rate, blood pressure, and respiration were closely monitored. After the experiment, the rabbits were euthanized, and cardiac tissue was harvested for pathological analysis to determine the extent of infarction and to assess the degree of myocardial damage by detecting serum myocardial enzyme levels.

[0102] Figure 6 The initial blood perfusion signal and initial pulse wave signal of the rabbit measured in an experiment according to an embodiment of this application are shown. Figure 7 The illustration shows multiple first mode components obtained based on the initial blood perfusion signal and initial pulse wave signal of a rabbit in an experiment according to an embodiment of this application. Figure 6 Empirical mode decomposition is performed on the two signals to obtain, as follows: Figure 7 The first modal component 1, the first modal component 2, and the first modal component 3 are shown.

[0103] Figure 8 The illustration shows multiple second-mode components obtained by screening based on frequency domain kurtosis and energy proportion in an experiment according to an embodiment of this application. The multiple first-mode components were then screened according to the principle of frequency domain kurtosis > 1 and energy proportion > 0.8, resulting in the following... Figure 8 The second mode component 1, second mode component 2, and second mode component 3 are shown.

[0104] Figure 9 The diagram illustrates multiple third mode components obtained in an experiment according to an embodiment of this application. Variational mode decomposition is performed on the aforementioned multiple second mode components to obtain, as shown below... Figure 9 The third mode component 1 and the third mode component 2 are shown.

[0105] Figure 10 The reconstructed signal obtained in an experiment according to an embodiment of this application is shown. Signal reconstruction based on the aforementioned third mode component yields the following result: Figure 10 The reconstructed signal is shown.

[0106] It can be seen that the signal reconstruction method of this application can significantly improve the signal-to-noise ratio and obtain blood flow pulse wave characteristics more clearly.

[0107] Finally, initial blood perfusion index and initial pulse wave signals were obtained from normal animals (17 animals) and animals with myocardial infarction (13 animals) using animal models. The signals before and after reconstruction were used as input signals for the discrimination model to determine myocardial infarction. The results are shown in Table 1.

[0108] Table 1

[0109]

[0110] The experimental results show that using the reconstructed signal can improve the accuracy of discrimination in animals with myocardial infarction.

[0111] The myocardial infarction detection system based on diffuse speckle provided in this application, by simultaneously acquiring blood perfusion index signals and pulse wave signals, more comprehensively characterizes the dynamic response of the cardiovascular system, significantly improving the accuracy of myocardial infarction detection. Furthermore, this application reconstructs the initial blood perfusion index signals and initial pulse wave signals, reducing signal interference caused by displacement of skin, muscles, and bones due to the subject's physical activity, as well as interference caused by fluctuations in ambient light, further improving the accuracy of myocardial infarction detection.

[0112] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this application. It should be understood that the above descriptions are merely specific embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A myocardial infarction detection system based on diffuse speckle, comprising: A laser is used to generate a laser signal. An optical fiber probe, connected to the laser via a multimode optical fiber, is used to direct the laser signal to the object under test and to collect the reflected light signal from the object under test. A camera, connected to the fiber optic probe via a multimode fiber, is used to obtain a diffuse speckle image based on the reflected light signal; as well as The processing device, electrically connected to the camera, is used to obtain an initial blood perfusion index signal and an initial pulse wave signal based on the diffuse speckle image. The processing device reconstructs the initial blood perfusion index signal and the initial pulse wave signal to obtain a target blood perfusion index signal and a target pulse wave signal, and performs myocardial infarction detection based on the target blood perfusion index signal and the target pulse wave signal.

2. The system according to claim 1, characterized in that, The processing equipment is used for: For any one of the initial signals, namely the initial blood perfusion index signal and the initial pulse wave signal, empirical mode decomposition is performed on the initial signal to obtain multiple first mode components; Based on frequency domain kurtosis and energy proportion, multiple first mode components are filtered to obtain multiple second mode components; Variational mode decomposition is performed on multiple second mode components to obtain multiple third mode components; Signal reconstruction is performed based on multiple third mode components to obtain the target blood perfusion index signal and the target pulse wave signal.

3. The system according to claim 2, characterized in that, The processing equipment is used for: Based on multiple noise figures, Gaussian noise is injected into the initial signal to obtain multiple signals to be decomposed; Empirical mode decomposition is performed on the signal to be decomposed to obtain multiple initial mode components of the target order; Based on the average value of the multiple initial modal components, the first modal component of the signal to be decomposed is obtained, thereby obtaining multiple first modal components.

4. The system according to claim 2, characterized in that, The processing equipment is used for: Calculate the frequency domain kurtosis and energy percentage of each of the first modal components; From a plurality of first modal components, modal components with frequency domain kurtosis greater than a first preset value and energy percentage greater than a second preset value are selected to obtain a plurality of second modal components.

5. The system according to claim 2, characterized in that, The processing equipment is used for: Multiple peak frequencies are obtained by searching the spectrum based on multiple second mode components; Clustering is performed based on multiple peak frequencies to obtain the center frequency; Based on the center frequency, variational mode decomposition is performed on multiple second mode components to obtain multiple third mode components.

6. The system according to claim 2, characterized in that, The processing device is used to: accumulate multiple third modal components to obtain the target blood perfusion index signal and the target pulse wave signal.

7. The system according to claim 1, characterized in that, The processing equipment is used for: Joint feature extraction is performed on the target blood perfusion index signal and the target pulse wave signal to obtain correlation features; Myocardial infarction detection is performed based on the aforementioned correlation features.

8. The system according to claim 7, characterized in that, The processing equipment is used for: Based on a preset window and a preset step size, the target blood perfusion index signal is divided into multiple first sub-signals, and the target pulse wave signal is divided into multiple second sub-signals. A first-order correlation coefficient matrix is ​​obtained by performing correlation calculations based on multiple first sub-signals and multiple second sub-signals. The first-order correlation coefficient matrix is ​​normalized to obtain the second-order correlation coefficient matrix; The correlation features are obtained by multiplying the first-order correlation coefficient matrix and the second-order correlation coefficient matrix.

9. The system according to claim 8, characterized in that, The processing equipment is used for: Correlation calculations are performed on the first and second sub-signals corresponding to the same window to obtain low-order frequency components; The first-order correlation coefficient matrix is ​​obtained based on multiple low-order frequency components.

10. The system according to claim 7, characterized in that, The processing equipment is used for: The correlation features are segmented to obtain multiple feature blocks; Encode each of the multiple feature blocks to obtain multiple bottleneck features; By inputting multiple bottleneck features into a classifier, the myocardial infarction detection results are obtained.

Images