Physiological signal detection method and system for magnetic resonance scanning and storage medium

Abstract

The invention relates to the technical field of physiological signal detection. The invention discloses a physiological signal detection method and system oriented to magnetic resonance scanning and a storage medium. The method and system can achieve accurate detection of multi-dimensional physiological parameters in a high-intensity magnetic field environment. The physiological signal detection method for magnetic resonance scanning comprises the following steps: in the process of scanning a patient by magnetic resonance equipment of the physiological signal detection system for magnetic resonance scanning, acquiring a multispectral pulse wave signal of the face of the patient by adopting a time sequence light splitting method; when the multispectral pulse wave signals are obtained, heart vibration signals of the chest of the patient are obtained through a vibration measurement method; preprocessing the multispectral pulse wave signal to obtain a first target signal; preprocessing the heart vibration signal to obtain a second target signal; extracting a physiological signal with a plurality of physiological parameters based on the first target signal and the second target signal; the physiological signals are fed back to medical staff.

Description

Technical Field

[0001] The present application relates to the technical field of physiological signal detection, and more specifically, to a physiological signal detection method, system and storage medium for magnetic resonance scanning. Background Art

[0002] Cardiovascular Magnetic Resonance (CMR), as a high-resolution medical imaging technology, plays a vital role in evaluating the structure and function of the heart and blood vessels. However, the periodic motion caused by physiological activities such as heart and breathing will produce motion artifacts in high-field magnetic resonance environments (such as 3T and above), which will affect the accurate identification of lesions. Although existing gating technologies, such as electrocardiogram (ECG) and photoplethysmogram (PPG), can alleviate the artifact problem to a certain extent, ECG and pulse waves can only detect a single physiological signal and cannot achieve accurate detection of multi-dimensional physiological parameters. They cannot meet the current needs of detecting patients' multi-dimensional physiological parameters in a magnetic resonance environment. Summary of the invention

[0003] The purpose of the embodiments of the present application is to provide a physiological signal detection method, system and storage medium for magnetic resonance scanning, which can realize accurate detection of multi-dimensional physiological parameters in a strong magnetic field environment. The embodiments of the present application are mainly implemented through the following technical solutions: According to a first aspect of an embodiment of the present application, a physiological signal detection method for magnetic resonance scanning is provided, comprising: In the process of scanning a patient by a magnetic resonance device of a physiological signal detection system for magnetic resonance scanning, a multi-spectral pulse wave signal of the patient's face is obtained by using a time-series spectroscopic method; While acquiring the multi-spectral pulse wave signal, using a vibrometric method to acquire a cardiac vibration signal from the patient's chest; Preprocessing the multi-spectral pulse wave signal to obtain a first target signal; Preprocessing the heart vibration signal to obtain a second target signal; extracting a physiological signal having a plurality of physiological parameters based on the first target signal and the second target signal; The physiological signal is fed back to medical personnel.

[0004] According to one embodiment of the present application, the step of acquiring a multi-spectral pulse wave signal of the patient's face using a time-series spectroscopic method includes: receiving a first image sequence transmitted by a first camera of the physiological signal detection system for magnetic resonance scanning, wherein the first image sequence comprises an image sequence of multiple frame periods, and the image sequence of each frame period is an image sequence obtained by the first camera using multiple spectra of different wavelengths to sequentially photograph the face in a spectral order; In the first image sequence, first images corresponding to spectrums of the same wavelength are combined into a single spectrum video to obtain a plurality of spectrum videos; Performing a time series analysis on all pixels in a target spectral video, and using a skin reflectance model to model the change in the reflectance value of each pixel in the RGB channel of the target spectral video over time, to obtain a single spectrum pulse wave signal corresponding to the target spectral video, wherein the target spectral video is any one of the multiple spectral videos; All single-spectrum pulse wave signals constitute the multi-spectrum pulse wave signal.

[0005] According to one embodiment of the present application, the step of obtaining the cardiac vibration signal of the patient's chest by using the vibrometric method includes: receiving a second image sequence transmitted by a second camera of the physiological signal detection system for magnetic resonance scanning; Calculating a first motion amplitude in the X direction and a second motion amplitude in the Y direction of each second image in the second image sequence by using an optical flow method; combining the first motion amplitudes of all the second images in the X direction into a first motion amplitude sequence; combining the second motion amplitudes of all the second images in the Y direction into a second motion amplitude sequence; extracting motion angles from the first motion amplitude sequence and the second motion amplitude sequence using a sliding window method; The heart vibration signal is obtained based on the movement angle.

[0006] According to one embodiment of the present application, the step of preprocessing the multi-spectral pulse wave signal to obtain a first target signal includes: Performing high-pass filtering on the multi-spectral pulse wave signal to obtain a first signal to be processed; Performing motion compensation processing on the first signal to be processed to obtain a second signal to be processed; Perform noise reduction processing on the second signal to be processed to obtain the first target signal.

[0007] According to one embodiment of the present application, the step of preprocessing the heart vibration signal to obtain the second target signal includes: Performing high-pass filtering on the heart vibration signal to obtain a third signal to be processed; Performing motion compensation processing on the third signal to be processed to obtain a fourth signal to be processed; Perform noise reduction processing on the fourth signal to be processed to obtain the second target signal.

[0008] According to one embodiment of the present application, the step of extracting a physiological signal having multiple physiological parameters based on the first target signal and the second target signal includes: Extracting the blood oxygen saturation of the physiological signal from the first target signal; Calculating the heart rate and heart rate variability of the physiological signal based on the second target signal; The respiratory rate of the physiological signal is calculated from the second target signal using an optical flow method.

[0009] According to one embodiment of the present application, after the step of extracting a physiological signal having multiple physiological parameters based on the first target signal and the second target signal, the physiological signal detection method for magnetic resonance scanning further includes: When at least one physiological parameter of the physiological signal reaches a first preset threshold, an alarm mechanism is activated; When the trend change state of at least one physiological parameter of the physiological signal reaches a dangerous state, the early warning mechanism is activated.

[0010] A second aspect of an embodiment of the present application provides a physiological signal detection system for magnetic resonance scanning, comprising: A multi-spectral pulse wave signal acquisition module, used for acquiring a multi-spectral pulse wave signal of the face of the patient by using a time-series spectroscopic method during the process of scanning the patient by a magnetic resonance device of a physiological signal detection system for magnetic resonance scanning; A cardiac vibration signal acquisition module, used to acquire the cardiac vibration signal of the patient's chest by using a vibration measurement method while acquiring the multi-spectral pulse wave signal; A first target signal acquisition module, used for preprocessing the multi-spectral pulse wave signal to obtain a first target signal; A second target signal acquisition module, used for preprocessing the heart vibration signal to obtain a second target signal; A physiological signal extraction module, configured to extract a physiological signal having a plurality of physiological parameters based on the first target signal and the second target signal; The feedback module is used to feed back the physiological signal to medical staff.

[0011] According to a third aspect of an embodiment of the present application, a physiological signal detection system for magnetic resonance scanning is provided, comprising: a processor, a memory, a multimodal non-contact multi-parameter detection optical device, a laser, and a magnetic resonance device, wherein the multimodal non-contact multi-parameter detection optical device comprises a first camera and a second camera, the memory is used to store a computer program, the processor is used to call and run the computer program stored in the memory, and execute the steps of the physiological signal detection method for magnetic resonance scanning described in the first aspect of the embodiment of the present application, the magnetic resonance device, the first camera, and the second camera are all communicatively connected to the processor, and the laser is arranged on one side of the second camera; The first camera is used to perform an image acquisition operation on the skin on the patient's face to obtain a first image sequence, and transmit the first image sequence to the processor; The second camera is used to perform an image acquisition operation on the patient's chest to obtain a second image sequence, and transmit the second image sequence to the processor; When the second camera performs an image acquisition operation on the chest, the laser emits laser light to irradiate the chest.

[0012] According to a fourth aspect of an embodiment of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium is used to store a computer program, wherein the computer program enables a computer to execute the steps of the physiological signal detection method for magnetic resonance scanning provided in the first aspect of the embodiment of the present application.

[0013] The beneficial effects of the embodiments of the present application include: The embodiment of the present application uses a multi-spectral pulse wave signal and a heart vibration signal to detect the patient's physiological signals. Specifically, the embodiment of the present application uses a time-series spectroscopic method to obtain a multi-spectral pulse wave signal on the patient's face during the process of scanning the patient with a magnetic resonance device of a physiological signal detection system for magnetic resonance scanning; while obtaining the multi-spectral pulse wave signal, a vibration measurement method is used to obtain a heart vibration signal on the patient's chest; the multi-spectral pulse wave signal is pre-processed to obtain a first target signal; the heart vibration signal is pre-processed to obtain a second target signal; a physiological signal with multiple physiological parameters is extracted based on the first target signal and the second target signal; and the physiological signal is fed back to medical staff. Therefore, the embodiment of the present application can realize the precise detection of multi-dimensional physiological parameters in a strong magnetic field environment, so that medical staff can monitor multiple physiological indicators in real time to ensure the life safety of patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the conventional technology, the drawings required for use in the embodiments or the conventional technology descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0015] Figure 1 A flowchart of a physiological signal detection method for magnetic resonance scanning of the present application in some embodiments; Figure 2 A schematic diagram of the principle of the physiological signal detection method for magnetic resonance scanning in some embodiments of the present application; Figure 3 A flowchart for building an early warning model in this application; Figure 4 A flowchart of another embodiment of the physiological signal detection method for magnetic resonance scanning of the present application; Figure 5 A principle block diagram of a physiological signal detection system for magnetic resonance scanning of the present application in some embodiments; Figure 6 It is a principle block diagram of the physiological signal detection system for magnetic resonance scanning of the present application in other embodiments; Figure 7 It is a principle block diagram of a multi-modal non-contact multi-parameter detection optical device in some embodiments of the present application; Figure 8 This is a principle block diagram of the physiological signal detection system for magnetic resonance scanning of the present application in some further embodiments. DETAILED DESCRIPTION

[0016] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present application, so the present application is not limited by the specific embodiments disclosed below.

[0017] It should be noted that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0018] The terms "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0019] The terms "comprises," "includes," or any other variations thereof are intended to cover a non-exclusive inclusion, for example, a process, method, system, product, or apparatus that comprises a series of steps or elements is not necessarily limited to those steps or elements explicitly listed but may include other steps or elements not explicitly listed or inherent to such process, method, product, or apparatus.

[0020] Unless otherwise defined, all technical and scientific terms used in the specification of this application have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application. The term "and / or" used in the specification of this application includes any and all combinations of one or more related listed items.

[0021] The specific implementation of the present application is further described below in conjunction with the accompanying drawings.

[0022] refer to Figure 1 FIG. 1 is a flow chart of a physiological signal detection method for magnetic resonance scanning provided in the first aspect of the embodiment of the present application. Figure 1 In the present invention, the physiological signal detection method for magnetic resonance scanning comprises: S1. In the process of scanning a patient by a magnetic resonance device of a physiological signal detection system for magnetic resonance scanning, a time-series spectroscopic method is used to obtain a multi-spectral pulse wave signal of the patient's face.

[0023] The multi-spectral pulse wave signal can be expressed as a rPPG (Remote Photoplethysmography) signal. It should be understood that rPPG technology captures the changes in subcutaneous vascular volume caused by heart beats, affects the reflected light of the skin, forms a pulsation signal in the video, and thus realizes the detection of heart rate and heart rate variability.

[0024] Furthermore, the step of acquiring the multi-spectral pulse wave signal of the patient's face by using the time-series spectroscopic method includes: S11. Receive a first image sequence transmitted by a first camera of the physiological signal detection system for magnetic resonance scanning, wherein the first image sequence includes an image sequence of multiple frame periods, and the image sequence of each frame period is an image sequence obtained by the first camera using a plurality of different wavelengths of spectra to sequentially photograph the face in spectral order.

[0025] The first camera needs to be aimed at the skin area of ​​the patient's face to shoot, while ensuring that the facial skin is included in the picture and reducing the possibility of frame loss by the first camera.

[0026] The first camera is a multispectral camera.

[0027] The frame period refers to the time required for the first camera to sequentially use a plurality of spectra of different wavelengths to photograph the face. Figure 2 shown.

[0028] In this embodiment, the multiple spectra of different wavelengths are spectra of four different wavelengths, specifically blue light, green light, red light and infrared light. Exemplarily, the blue light is 465nm, the green light is 530nm, the red light is 680nm, and the infrared light is 850nm. In the process of acquiring the first image sequence, the first camera alternately uses blue light, green light, red light and infrared light for fill light, thereby being able to capture image data of different spectra (that is, the first image sequence), providing support for the measurement of heart rate, heart rate variability and blood oxygen saturation.

[0029] The spectral order can be set by those skilled in the art according to actual needs, and no further limitation is made herein.

[0030] S12. In the first image sequence, first images corresponding to spectra of the same wavelength are combined into a single spectral video to obtain multiple spectral videos.

[0031] It should be understood that each image in the first image sequence is a first image.

[0032] In the embodiment of the present application, four spectral videos can be obtained, that is, one spectral video corresponding to blue light, one spectral video corresponding to green light, one spectral video corresponding to red light, and one spectral video corresponding to infrared light.

[0033] S13. Perform time series analysis on all pixels in the target spectral video, and use a skin reflectance model to model the change in the reflectance value of each pixel in the target spectral video in the RGB channel (i.e., the red, green, and blue channels) over time to obtain a single spectrum pulse wave signal corresponding to the target spectral video, wherein the target spectral video is any one of the multiple spectral videos.

[0034] The skin reflection model is a mathematical model used to describe how light interacts with the skin. As a multi-layered structure, the reflection characteristics of the skin are determined by many factors, including the oil on the skin surface, the melanin in the epidermis, the hemoglobin in the dermis, etc. When light shines on the skin, part of the light will be reflected by the oil on the skin surface, forming a specular reflection (or highlight), while the other part of the light will penetrate the skin, be scattered and then reflected, forming a diffuse reflection.

[0035] In the embodiment of the present application, the reflection signal of each pixel in the RGB channel of the skin reflection model is mainly composed of light intensity, specular reflection, diffuse reflection and quantization noise. Among them, specular reflection is mainly composed of light reflected from the skin surface. Although it does not contain physiological information of heartbeat, the changes in light angle and intensity caused by the movement of the subject (ie, the patient) may cause noise interference. Diffuse reflection is closely related to the absorption and scattering of light by skin tissue, and is the key source of pulse signal. It is affected by the periodic changes in hemoglobin concentration, thereby forming a periodic reflection signal consistent with the heartbeat. Although melanin and carotene contribute to light absorption and scattering, their reflection intensity does not change with heart activity.

[0036] The single spectrum pulse wave signal can refer to Figure 2 The signals corresponding to labels 1, 2, 3 and 4, among which, the signal pointed to by label 1 is the single-spectrum pulse wave signal corresponding to blue light, the signal pointed to by label 2 is the single-spectrum pulse wave signal corresponding to green light, the signal pointed to by label 3 is the single-spectrum pulse wave signal corresponding to red light, and the signal pointed to by label 4 is the single-spectrum pulse wave signal corresponding to infrared light.

[0037] S14. All single-spectrum pulse wave signals are combined to form the multi-spectrum pulse wave signal.

[0038] The embodiment of the present application accurately describes the effect of changes in hemoglobin concentration on the intensity of reflected light, and combines the pigment distribution of skin tissue and the depth of skin color, so that the first camera can effectively extract a multispectral rPPG signal (that is, the multispectral pulse wave signal) containing multidimensional vital signs information such as heart rate, heart rate variability and blood oxygen.

[0039] S2. While acquiring the multi-spectral pulse wave signal, a vibrometric method is used to acquire a cardiac vibration signal of the patient's chest.

[0040] The heart vibration signal may be expressed as a SCG (Seismocardiography) signal.

[0041] The SCG signal is based on the mechanical vibration of the heart and can effectively extract parameters such as heart rate, heart rate variability and respiratory rate, thereby supplementing and enriching the detection dimensions of the overall physiological parameters.

[0042] Furthermore, the step of obtaining the cardiac vibration signal of the patient's chest by using a vibrometric method comprises: S21, receiving a second image sequence transmitted by a second camera of the physiological signal detection system for magnetic resonance scanning.

[0043] The second camera is a through-focus camera.

[0044] During the patient's heart beat, the tiny movement of the chest (i.e., chest) surface causes the three-dimensional interference field formed by the reflected laser to change, resulting in the interference speckle pattern to move periodically on the imaging plane of the second camera, and synchronized with the contraction and relaxation process of the heart. In order to achieve accurate capture, the physiological signal detection system for magnetic resonance scanning adjusts the angle of the second camera so that the laser is irradiated to the fourth rib on the lower left of the chest. In the fourth rib area, the laser and the surface reflected light interfere with each other to form a periodically changing speckle pattern, which is then recorded by the second camera to obtain the heart vibration signal. As the heart beats, the speckle pattern shows periodic changes. The physiological signal detection system for magnetic resonance scanning can accurately reconstruct the heart movement process and extract the SCG signal by analyzing the movement pattern in the image.

[0045] In order to further enhance the visibility of tiny chest movements, the physiological signal detection system for magnetic resonance scanning adjusts the focal length of the lens of the second camera to move the focal plane of the second camera from the heart plane to a specific position, thereby significantly amplifying the interference speckle image reflected by the laser. When the focal plane is adjusted to an appropriate position, the speckle image will change periodically with the heart beat, and the defocus degree of the lens of the second camera is proportional to the motion amplification gain of the heart shock signal (i.e., the heart vibration signal).

[0046] The physiological signal detection system for magnetic resonance scanning uses a second camera with a high frame rate of at least 200 fps and a resolution of 400×300 to capture laser speckle images (ie, the second image sequence).

[0047] S22: Calculate a first motion amplitude in the X direction and a second motion amplitude in the Y direction for each second image in the second image sequence by using an optical flow method.

[0048] S23. Combine the first motion amplitudes of all second images in the X direction into a first motion amplitude sequence.

[0049] S24. Combine the second motion amplitudes of all the second images in the Y direction into a second motion amplitude sequence.

[0050] S25. Extract motion angles from the first motion amplitude sequence and the second motion amplitude sequence using a sliding window method.

[0051] Furthermore, the step S25 includes: S251. Set a preset window and a preset step.

[0052] The size of the preset window can be set by those skilled in the art according to actual needs. The specific value of the preset stride can be set by those skilled in the art according to actual needs.

[0053] S252: Extract a first subsequence set from the first motion amplitude sequence according to the preset window and the preset stride.

[0054] S253: Extract a second subsequence set from the second motion amplitude sequence according to the preset window and the preset step.

[0055] S254: Calculate a to-be-processed angle based on the first subsequence set and the second subsequence set.

[0056] Furthermore, the calculation formula of step S254 is: ; in, represents the main motion quadrant, which needs to be obtained from the four quadrants using the calculation formula in step S254. The purpose is that the actual motion of the speckle may be noisy, but the motion direction of the speckle motion caused by the heart is highly concentrated and located in one of the four quadrants; It means that the quadrant with the largest 90-point value is selected from the four quadrants as the main motion quadrant. The 90-point value refers to the statistical index used when analyzing the motion data in the four quadrants. Represents the ninetieth digit amplitude value in the corresponding quadrant; represents the first subsequence set elements; represents the first elements; Represents the current Axis coordinate position; Represents the current Coordinate location; represents the four quadrants in the Cartesian coordinate system, ; represents the angle to be processed; represents the median; represents the tangent function; Representative Points in The integral value in the direction, that is, the first The first element The integral value of each point; Representative Points in The integral value in the direction, that is, the first subsequence set The first element The integral value of each point; Representative Points in The coordinate position of the axis; Representative Points in The coordinate position of the axis.

[0057] When processing cardiac vibration signals, the physiological signal detection system for magnetic resonance scanning calculates the 90th percentile value of the motion data in each quadrant, that is, it calculates that 90% of the data points of all motion data in each quadrant are lower than or equal to a specific value in the quadrant. By comparing the 90th percentile values ​​of the four quadrants, the largest quadrant is selected as the "main motion quadrant". The "main motion quadrant" contains the most significant cardiac motion information, because its distribution concentration of motion amplitude is high, which can effectively eliminate noise and other non-cardiac motion interference.

[0058] In the process of calculating the motion angle using the optical flow method, the physiological signal detection system for magnetic resonance scanning divides the image plane into four quadrants. The motion data in each quadrant represents the direction and amplitude of the object's motion in that quadrant. By analyzing the motion data in each quadrant, the system can determine the object's motion trend on the image plane, thereby helping to extract the heart vibration signal.

[0059] S255: Calculate the average value of all the angles to be processed to obtain the movement angle.

[0060] S26. Obtain the heart vibration signal based on the movement angle.

[0061] Furthermore, the calculation formula of step S26 is: ; in, represents the heart vibration signal; represent A signal sequence, namely, the first motion amplitude sequence; represents the angle of movement; represent Signal sequence, that is, the second motion amplitude sequence.

[0062] The S2 step provides a contactless, non-invasive and high-precision means of detecting heart health, which can support cardiac magnetic resonance (CMR) image acquisition or accurately record the position of heart movement.

[0063] S3. Preprocess the multi-spectral pulse wave signal to obtain a first target signal.

[0064] Furthermore, the S3 step includes: S31, performing high-pass filtering on the multi-spectrum pulse wave signal to obtain a first signal to be processed.

[0065] Since low-frequency signals usually come from physiological activities such as breathing or body movement, which are unrelated to the pulse wave signal, these low-frequency noises will affect the accuracy and stability of the multi-spectral pulse wave signal. In practical applications, low-frequency noise often manifests itself as fluctuations with a longer period, usually below 0.1 Hz. These fluctuations mainly come from respiratory rhythm, head movement or slight shaking of the first camera. In order to effectively remove these irrelevant low-frequency interferences, the physiological signal detection system for magnetic resonance scanning uses a high-pass filter to perform high-pass filtering on the multi-spectral pulse wave signal to obtain the first signal to be processed. The high-pass filter can allow signals above the preset cutoff frequency to pass through, while filtering out low-frequency components below the preset cutoff frequency.

[0066] Through the above method, the physiological signal detection system for magnetic resonance scanning can effectively remove low-frequency noise caused by breathing or body movement, and retain high-frequency components related to the pulse wave signal, thereby improving the accuracy of the multi-spectral pulse wave signal (i.e., rPPG signal), and providing a clearer basis for the subsequent extraction of heart rate and other physiological parameters.

[0067] S32: Perform motion compensation processing on the first signal to be processed to obtain a second signal to be processed.

[0068] Further, the step S32 includes: S321. Divide the first signal to be processed into multiple first signal segments.

[0069] S322. Detect a first feature point in each first signal segment.

[0070] The first characteristic point is a peak, a trough or an inflection point in the corresponding first signal segment.

[0071] S323. Using an optical flow method to combine all the first feature points and all the first signal segments, calculate a first optical flow field between adjacent first signal segments in all the first signal segments, where the first optical flow field describes the movement direction and speed of each first feature point in the adjacent first signal segments.

[0072] The optical flow method is a gradient-based Lucas-Kanade method or a global energy optimization-based Horn-Schunck method (this method is a classic dense optical flow algorithm).

[0073] S324: Calculate a first compensation amount for each first feature point based on all first optical flow fields.

[0074] The first compensation amount includes motion parameters such as translation and rotation, which are used to correct signal distortion caused by motion.

[0075] S325 . Perform motion compensation processing on the first signal to be processed based on the compensation amount of each first feature point to obtain the second signal to be processed.

[0076] The embodiment of the present application uses an optical flow method to eliminate noise generated by motion in the first signal to be processed, thereby obtaining the second signal to be processed.

[0077] The motion compensation processing in step S32 is intended to correct the interference of the patient's movement on the first signal to be processed, and the result is to obtain a corrected signal, that is, the second signal to be processed. The second signal to be processed can more accurately reflect the patient's physiological state. After compensation, the waveform of the second signal to be processed is more stable and the fluctuation is more regular, which not only facilitates the subsequent extraction of physiological parameters, but also effectively reduces the noise and interference components in the signal, thereby significantly improving the signal-to-noise ratio of the signal. More importantly, the signal processed by motion compensation can more truly reflect the patient's physiological changes, provide medical staff with a reliable basis for diagnosis, and thus ensure the patient's life safety.

[0078] S33: Perform noise reduction processing on the second signal to be processed to obtain the first target signal.

[0079] Specifically, the embodiment of the present application uses independent component analysis (ICA) to perform noise reduction processing on the second signal to be processed.

[0080] Independent component analysis is a blind source separation technology that can extract independent components from multiple mixed signals. Independent component analysis is widely used in the denoising of complex signals. In the physiological signal detection system for magnetic resonance scanning, the signal is often disturbed by factors such as facial movement, environmental noise, illumination changes and equipment errors, resulting in signal confusion, thereby affecting the accurate analysis of physiological parameters. Independent component analysis separates independent components from mixed signals by assuming that the signal sources are independent of each other and non-Gaussian. During the processing, the ICA optimization algorithm restores the mixed signal to multiple independent components by maximizing the independence of the signal, thereby removing noise components that are not related to blood flow, such as facial movement and environmental interference. This technology can significantly improve the purity and reliability of the signal, ensure that the physiological signals extracted from the physiological signal detection system for magnetic resonance scanning are more accurate, and provide more accurate data support for the analysis of parameters such as heart rate and blood oxygen saturation, thereby improving the performance of the non-contact health detection system (that is, the physiological signal detection system for magnetic resonance scanning).

[0081] Before performing high-pass filtering on the multi-spectrum pulse wave signal, the multi-spectrum pulse wave signal may be normalized.

[0082] The S3 step can ensure the accuracy and stability of the first target signal.

[0083] S4. Preprocess the heart vibration signal to obtain a second target signal.

[0084] Furthermore, the S4 step includes: S41. Perform high-pass filtering on the heart vibration signal to obtain a third signal to be processed.

[0085] The embodiment of the present application uses a high-pass filter to perform high-pass filtering on the heart vibration signal.

[0086] S42: Perform motion compensation processing on the third signal to be processed to obtain a fourth signal to be processed.

[0087] Furthermore, the step S42 includes: S421. Divide the third signal to be processed into multiple second signal segments.

[0088] S422: Detect a second feature point in each second signal segment.

[0089] The second characteristic point is a peak, a trough or an inflection point in the corresponding second signal segment.

[0090] S423. Using an optical flow method to combine all the second feature points and all the second signal segments, calculate a second optical flow field between adjacent second signal segments in all the second signal segments, where the second optical flow field describes the movement direction and speed of each second feature point in the adjacent second signal segments.

[0091] The optical flow method is a gradient-based Lucas-Kanade method or a global energy optimization-based Horn-Schunck method (this method is a classic dense optical flow algorithm).

[0092] S424: Calculate a second compensation amount for each second feature point based on all second optical flow fields.

[0093] The second compensation amount includes motion parameters such as translation and rotation, and is used to correct signal distortion caused by motion.

[0094] S425 . Perform motion compensation processing on the third signal to be processed based on the compensation amount of each second feature point to obtain the fourth signal to be processed.

[0095] The embodiment of the present application adopts the optical flow method to eliminate the noise generated by motion in the third signal to be processed, thereby obtaining the fourth signal to be processed. The motion compensation processing in step S42 is intended to correct the interference of the patient's movement on the third signal to be processed, and the result is to obtain a corrected signal, that is, the fourth signal to be processed. The fourth signal to be processed can more accurately reflect the physiological state of the patient. After compensation, the waveform of the fourth signal to be processed is more stable and the fluctuation is more regular, which not only facilitates the subsequent extraction of physiological parameters, but also effectively reduces the noise and interference components in the fourth signal to be processed, thereby significantly improving the signal-to-noise ratio of the fourth signal to be processed.

[0096] S43: Perform noise reduction processing on the fourth signal to be processed to obtain the second target signal.

[0097] The embodiment of the present application uses independent component analysis to perform noise reduction on the fourth signal to be processed.

[0098] S5. Extracting a physiological signal having multiple physiological parameters based on the first target signal and the second target signal.

[0099] Furthermore, the step S5 includes: S51, extracting the blood oxygen saturation of the physiological signal from the first target signal. The step S51 can be understood as using a blood oxygen saturation detection model to detect the blood oxygen saturation.

[0100] Specifically, the step S51 includes (that is, the implementation steps of the blood oxygen saturation detection model): S511. Perform detrending processing on the first target signal to obtain a third target signal.

[0101] The step S511 can remove the long-term trend component in the first target signal.

[0102] S512. Perform bandpass filtering on the third target signal to obtain a fourth target signal.

[0103] The embodiment of the present application uses a bandpass filter to implement step S512, and the bandpass filter can selectively pass signals within a specific frequency range while attenuating or blocking signals of other frequencies. For adults, the specific frequency range is 0.7 Hz to 3 Hz; for infants, the specific frequency range is 1.5 Hz to 5 Hz.

[0104] The step S512 can retain the pulse wave signal within the heart rate frequency band.

[0105] Furthermore, the step S512 can be expressed as a formula: PPG filtered (t) = BandPass(PPG raw (t)); where PPG filtered (t) is the fourth target signal, BandPass() is the bandpass filter, PPG raw (t) is the third target signal, and t is the current time.

[0106] S513 , detecting a peak corresponding to a red light channel (ie, R channel), a trough corresponding to the red light channel, a peak corresponding to a green light channel (ie, G channel), and a trough corresponding to the green light channel in the fourth target signal.

[0107] S514: Calculate the amplitude difference corresponding to the red light channel by using the difference between the peak corresponding to the red light channel and the trough corresponding to the red light channel.

[0108] The calculation formula of step S514 is: ; in, is the amplitude difference corresponding to the red light channel, is the peak corresponding to the red light channel, is the trough corresponding to the red light channel.

[0109] S515: Calculate the amplitude difference corresponding to the green light channel by using the difference between the peak corresponding to the green light channel and the trough corresponding to the green light channel.

[0110] The calculation formula of step S515 is: ; in, is the amplitude difference corresponding to the green light channel; is the peak corresponding to the green light channel; is the trough corresponding to the green light channel.

[0111] S516: The relative amplitude feature is obtained by dividing the amplitude difference corresponding to the green light channel by the amplitude difference corresponding to the red light channel.

[0112] The calculation formula of step S516 is: ; in, is the relative amplitude characteristic, is the amplitude difference corresponding to the green light channel, is the amplitude difference corresponding to the red light channel.

[0113] In other implementations, the relative amplitude feature may also be calculated by calculating the amplitude of the heart rate frequency corresponding to the red light channel and the amplitude of the heart rate frequency corresponding to the green light channel through Fourier transform. The specific calculation formula is as follows: ; in, is the relative amplitude characteristic, is the amplitude of the heart rate frequency corresponding to the green light channel, is the amplitude of the heart rate frequency corresponding to the red light channel, That is the heart rate.

[0114] S517. Obtain the blood oxygen saturation by calculation based on the relative amplitude feature.

[0115] The embodiment of the present application uses a regression model established by a polynomial regression method to implement the step S517.

[0116] Specifically, the calculation formula of step S517 is: ; Wherein, SpO2 is the blood oxygen saturation, k is the cross-band absorption difference correction factor, RRs is the relative amplitude characteristic, and b is the red band reference value.

[0117] The two parameters k and b reflect the correction relationship between the red light channel and the green light channel for the difference in hemoglobin absorption, and both are obtained through pre-training with a large amount of experimental data.

[0118] S52: Calculate the heart rate and heart rate variability of the physiological signal based on the second target signal. The step S52 can be understood as detecting the heart rate and the heart rate variability through a heart rate detection model.

[0119] Exemplarily, the step S52 includes (that is, the step of implementing the heart rate detection model detection): S521, identifying the R wave position of the second target signal, and calculating the RR interval (ie, the time interval between two adjacent R wave positions) according to the R wave position.

[0120] S522. Calculate the heart rate and heart rate variability based on the RR interval.

[0121] In other embodiments, the heart rate detection model combines multispectral rPPG signals and SCG signals, CNN is used to extract time domain features and frequency domain features, and a long short-term memory network (LSTM) or a Transformer model is used to capture long-term dependencies in the time series and calculate the heart rate value in real time.

[0122] S53, using the optical flow method to calculate the respiratory rate of the physiological signal from the second target signal. The step S53 can be understood as detecting the respiratory rate through a respiratory detection model.

[0123] The embodiment of the present application uses the optical flow method to convert chest movement into optical flow information and encode it, and then extracts the pixel brightness sequence to obtain respiratory waveform information. This method can effectively compensate for image deviation and blur caused by facial or body movement, thereby improving the accuracy and stability of physiological signal extraction.

[0124] In other embodiments, for breathing detection, a convolutional neural network (CNN) can be used to extract chest or abdominal motion features from video images, and these features are converted into frequency domain features through fast Fourier transform (FFT) to calculate the breathing frequency.

[0125] It should be understood that the respiratory rate can also be estimated through the movement change signal of the patient's abdomen. The respiratory rate is the respiratory frequency.

[0126] S6. Feedback the physiological signal to medical staff.

[0127] In the embodiment of the present application, the physiological signal may be displayed in real time on a display screen outside the MRI room so that the medical staff can monitor the physiological signal data in real time.

[0128] The embodiment of the present application uses multi-spectral pulse wave signals and heart vibration signals to detect the patient's physiological signals. Therefore, the embodiment of the present application can realize the accurate detection of multi-dimensional physiological parameters in a strong magnetic field environment, so that medical staff can monitor multiple physiological indicators in real time to ensure the patient's life safety.

[0129] The embodiment of the present application integrates multi-spectral optical signals (i.e., the multi-spectral pulse wave signals) and cardiac mechanical vibration signals (i.e., the cardiac vibration signals), and the method realizes high-precision detection of multiple physiological signals, and improves the reliability and accuracy of data through signal fusion. This design provides technical support for the application of multi-parameter detection systems (i.e., physiological signal detection systems for magnetic resonance scanning), and also lays a foundation for health management, clinical diagnosis, and early screening of cardiovascular diseases.

[0130] In addition, the embodiment of the present application also achieves deeper physiological information mining by calculating the pulse wave transmission time difference between the multi-spectral rPPG signal and the SCG signal, as well as the time difference between rPPG signals of different spectra. These time difference parameters combined with the multi-dimensional physiological features extracted from the signal can be used to construct a blood pressure estimation model based on machine learning, providing a new method for non-contact blood pressure measurement.

[0131] In some embodiments, the embodiments of the present application use multidimensional pulse wave transmission time (Multidimensional Pulse Transit Time, MD-PTT) blood pressure detection technology combined with signal trajectories at different points and different wavelengths at the same point to calculate multi-point pulse wave transmission time and multi-spectral pulse wave transmission time, and use traditional machine learning algorithms to establish a mapping relationship with blood pressure.

[0132] Specifically, the multi-point pulse wave transmission time (i.e., multi-point PTT) is obtained by calculating the peak time interval between the SCG signal (the cardiac vibration signal) and the single-spectrum pulse wave signal under any spectrum in the same cardiac cycle; the multi-spectral pulse wave transmission time is obtained by calculating the peak time interval of the single-spectrum pulse wave signal corresponding to the spectra of all wavelengths in the same cardiac cycle (i.e., calculating the peak time interval of the rPPG signal extracted under different spectra in the same cardiac cycle).

[0133] The embodiment of the present application can establish a mapping relationship with blood pressure based on the heart rate and heart rate variability extracted by MD-PTT and using a traditional machine learning algorithm.

[0134] Furthermore, the step of establishing a mapping relationship with blood pressure using a traditional machine learning algorithm includes: calculating the first systolic peak time difference (PPG-PPW-PTT) between the forehead PPG (G channel) signal and the neck pressure pulse wave (PPW) signal, calculating the second systolic peak time difference (PPG-PPG-PTT) between the forehead PPG and the neck PPG signal, and extracting key features related to blood pressure changes based on the first systolic peak time difference and the second systolic peak time difference. In an embodiment of the present application, the forehead PPG signal can be obtained by the multi-spectral pulse wave signal. The neck pressure pulse wave can be obtained by pressure sensor detection or Doppler ultrasound detection.

[0135] In order to ensure the accuracy of peak detection, the findpeaks function in MATLAB (MatrixLaboratory, an advanced numerical computing and programming environment) is used, and the main peaks in the signal are accurately located by setting appropriate parameters to ensure the time alignment of each signal segment; the synchronously collected blood pressure value timestamps are used to accurately calculate the average PTT value of each segment (that is, the first systolic peak time difference and the second systolic peak time difference). In addition, the embodiment of the present application also extracts the dicrotic index (AIx) feature. By detecting the peaks, valleys and dicrotic notches in the signal, and dividing the signal into multiple segments according to the blood pressure value timestamps, the dicrotic index of each signal segment is calculated. Subsequently, the polynomial regression model (that is, ) establish a mapping relationship between PTT (that is, the first systolic peak time difference and the second systolic peak time difference) and blood pressure, and use the polyfit (polynomial fitting) function and polyval (polynomial evaluation) function in MATLAB to optimize the regression coefficient through the least squares method to obtain a blood pressure detection model (that is, a blood pressure estimation model).

[0136] In other implementations, the blood pressure detection model may also be trained using machine learning algorithms such as support vector machine (SVM) and random forest (RF).

[0137] In the embodiment of the present application, during the real-time detection process, the pre-processed multi-spectral pulse wave signal and the heart vibration signal (that is, the first target signal and the second target signal) are input into the trained blood pressure estimation model, and the blood pressure estimation model calculates and outputs the systolic and diastolic blood pressure values ​​of the patient in real time according to the mapping relationship obtained through training. This ensures the accuracy and real-time performance of blood pressure detection.

[0138] The blood pressure estimation model can accurately extract and calculate key physiological parameters from physiological signals collected non-contactly, providing comprehensive, real-time and accurate physiological data support for the detection of physiological signals of patients during magnetic resonance scanning.

[0139] In some embodiments, features such as heartbeat interval (also known as the RR interval), heart rate variability (HRV), and heart rate change trend are extracted from the heart rate signal (also known as the heart vibration signal). These features can effectively reflect the stability of the heart rhythm and the activity status of the autonomic nervous system.

[0140] The embodiment of the present application extracts the respiratory cycle changes and the numerical values ​​of the respiratory frequency from the respiratory frequency signal, detects too fast or too slow breathing (for example, less than 10 times / minute or more than 25 times / minute), and uses Fast Fourier Transform (FFT) to extract frequency domain features (such as low-frequency and high-frequency components of the power spectral density).

[0141] The embodiment of the present application realizes real-time monitoring of changes in blood oxygen levels by extracting waveform features and actual values ​​from blood oxygen saturation (SpO2) signals. As one of the most sensitive physiological parameters, the blood oxygen saturation can capture small changes through trend analysis every 10 or 20 seconds. Once it is detected that the blood oxygen saturation is lower than the set threshold (usually 95%) or the downward trend exceeds 3%, the physiological signal detection system for magnetic resonance scanning will immediately trigger an alarm.

[0142] The embodiment of the present application extracts the real-time values ​​of systolic and diastolic pressure from the blood pressure signal, and combines the fluctuation of blood pressure to promptly detect abnormalities (such as systolic pressure exceeding 180 mmHg or diastolic pressure below 60 mmHg). The peak detection algorithm is used in the blood pressure waveform analysis to capture the key change points of blood pressure fluctuations.

[0143] In terms of personalized modeling, the physiological signal detection system for magnetic resonance scanning combines the normal range standards of physiological parameters published by authoritative medical institutions to provide a basis for threshold setting, and integrates machine learning algorithms with clinician experience to refine the threshold setting to ensure the personalization and accuracy of the test results.

[0144] Due to the particularity of the magnetic resonance scanning process, patients may be in different physiological states. The physiological signal detection system for magnetic resonance scanning can dynamically adjust the threshold to meet the needs of different populations and special cases. Figure 3 As shown in FIG, the physiological signal detection system for magnetic resonance imaging introduces deep learning methods such as convolutional neural network (CNN), long short-term memory network (LSTM) and Transformer (i.e. Figure 3”Transformer model” in the MRI scan to further optimize personalized modeling. Specifically, CNN can effectively extract spatiotemporal features (that is, time-domain features, such as heartbeat intervals or breathing cycles) in physiological signals and dynamically adjust them according to individual differences; LSTM can capture long-term dependencies in time series and provide accurate predictions and anomaly detection; through the Transformer model, the system can process and learn multi-dimensional physiological signals from different patient groups, and perform more accurate modeling and predictions. For the elderly, high-risk patients or special disease groups, the physiological signal detection system for magnetic resonance scanning can set precise thresholds based on individual differences, and combine clustering analysis (such as K-means algorithm, also known as K-means clustering algorithm) to establish a personalized normal physiological range to ensure the accuracy of real-time detection. In other embodiments, the physiological signal detection system for magnetic resonance scanning also introduces a decision tree to further optimize personalized modeling.

[0145] It should be understood that CNN can effectively extract the spatiotemporal features when processing physiological signals. CNN slides the convolution kernel in the time dimension through the convolution layer to extract local temporal features, and convolutions in the spatial dimension of multi-channel signals to extract spatial features. The pooling layer further reduces the dimensions of temporal and spatial features and enhances feature robustness. The fully connected layer integrates these features and outputs the final spatiotemporal features.

[0146] CNN is able to dynamically adjust its internal parameters, such as weights and biases, according to individual differences to adapt to the physiological characteristics of different individuals. This adjustment not only includes the optimization of model parameters, but also involves the automatic selection and emphasis of feature extraction, as well as the dynamic setting of the threshold of the normal range of physiological parameters. For example, for individuals with generally high heart rates, the model will automatically adjust the normal range of heart rate to avoid false alarms. In this way, CNN is able to not only extract spatiotemporal features, but also dynamically adjust according to individual differences, providing more personalized and accurate physiological detection and early warning.

[0147] LSTM captures long-term dependencies in time series through its unique gating mechanism (input gate, forget gate, and output gate). The input gate determines how much of the current input information is written into the cell state, the forget gate determines how much information in the cell state is retained or forgotten, and the output gate determines how much information in the cell state is output. These gating mechanisms enable LSTM to dynamically update the cell state at each time step, thereby retaining long-term contextual information. For example, when detecting heart rate, LSTM can remember the heart rate changes at previous time points, which can affect the current output even if they are far away from the current time point. This ability enables LSTM to capture long-term dependencies in time series and provide support for accurate physiological parameter prediction and anomaly detection.

[0148] Regarding long-term dependencies in time series, LSTM can more accurately predict future physiological parameters such as heart rate, respiratory rate, blood oxygen saturation, and blood pressure, which is crucial for early warning and intervention of potential health problems. Second, long-term dependencies help understand the context of the signal. For example, understanding the changing trend of heart rate can help the model better identify anomalies even if these changes occur over a long time span. Finally, by capturing normal patterns, LSTM can more easily identify anomalies such as sudden changes in heart rate, abnormal fluctuations in respiratory rate, etc., which may indicate potential health risks.

[0149] In addition, LSTM's anomaly detection mainly focuses on abnormal heart rate, abnormal breathing, abnormal blood oxygen saturation, and abnormal blood pressure. Specifically, LSTM can detect sudden changes in heart rate, such as arrhythmia, tachycardia, or bradycardia; detect abnormal fluctuations in respiratory rate, such as apnea or hyperventilation; detect a decrease in blood oxygen saturation to detect hypoxemia in a timely manner; detect a sudden increase or decrease in blood pressure to warn of hypertension or hypotension. By comprehensively analyzing multiple physiological parameters, LSTM can also detect potential comprehensive abnormalities, such as multi-parameter changes before cardiac arrest. These anomaly detection functions enable LSTM to detect the patient's physiological state in real time and immediately issue an alarm when an abnormality is detected, reminding doctors to take timely intervention measures to ensure the safety and health of patients.

[0150] The steps of establishing a personalized normal physiological range by cluster analysis include: first, removing missing values ​​and outliers of physiological signals to ensure the integrity and accuracy of the signals, and standardizing or normalizing the physiological signals to eliminate the dimensional differences between different physiological parameters (such as heart rate, respiratory rate, blood oxygen saturation, etc.), and ensuring that the influence of each parameter on the cluster analysis is balanced. Next, the K-means clustering algorithm is selected for data analysis, and the number of clusters K is pre-set. The selection of this K value can be determined by methods such as the elbow rule or the silhouette coefficient to find the optimal number of clusters. The process of the K-means algorithm includes randomly selecting K cluster centers and assigning data points to the nearest cluster by calculating the distance from each data point to each cluster center. Then, the center of each cluster (i.e., the mean of all data points in the cluster) is recalculated, and the process is repeated until the cluster center no longer changes significantly, and finally all data points are divided into K clusters, each cluster representing a physiological state. After clustering is completed, the clustering results need to be analyzed in detail to identify the cluster representing the normal physiological range, which usually contains data of healthy individuals. According to the statistical characteristics of the cluster (such as mean and standard deviation), the range of normal physiological parameters can be set. For example, the normal range of heart rate can be set as the mean of the heart rate data in the cluster plus or minus one standard deviation. Finally, the rationality of the normal range is verified in combination with historical data to ensure that it can accurately reflect the physiological state of healthy individuals, and dynamically adjusted through real-time detection data to adapt to the differences between different individuals.

[0151] In some embodiments, reference Figure 4 As shown, after the step of extracting the physiological signal having multiple physiological parameters based on the first target signal and the second target signal, the physiological signal detection method for magnetic resonance scanning further includes: S7. When at least one physiological parameter of the physiological signal reaches a first preset threshold, an alarm mechanism is activated.

[0152] The heart rate, the heart rate variability, the blood oxygen saturation and the respiratory rate are all physiological parameters.

[0153] The first preset threshold may be a pre-set normal fluctuation range, which may be specifically set by those skilled in the art according to actual needs.

[0154] The alarm mechanism reminds medical personnel or patients themselves through visual (such as flashing red light) signals (i.e. light signals) and auditory (such as alarm sound) signals (i.e. sound signals), ensuring that timely response measures can be taken in different environments. The alarm mechanism is instantaneous and can quickly notify relevant personnel to deal with abnormal situations that have occurred.

[0155] S8. When the trend change state of at least one physiological parameter of the physiological signal reaches a dangerous state, start an early warning mechanism.

[0156] The trend change state of physiological parameters refers to the regular or continuous change pattern of physiological parameters over time. This change pattern can be rising, falling, fluctuating or remaining stable.

[0157] The early warning mechanism is a trend forecast.

[0158] The early warning mechanism predicts potential risks in advance by dynamically detecting the changing trends of physiological parameters (such as combining the dynamic time warping (DTW) algorithm). When the physiological signal detection system for magnetic resonance scanning detects that the trend changes of key physiological parameters may reach a dangerous state (that is, reach the dangerous threshold), the early warning mechanism will issue a prompt to buy more reaction time for medical staff and patients and intervene in potential safety issues of patients in advance.

[0159] The embodiment of the present application uses accurate health detection and risk prediction, combined with the dual mechanism of alarm and early warning, so that the physiological signal detection system for magnetic resonance scanning can significantly reduce health risks. Especially in the CMR scanning process of patients with cardiovascular diseases, this dual mechanism of combining alarm and early warning significantly improves the safety of patients and the accuracy of medical staff's diagnosis, provides more sufficient response time for clinical intervention, and further optimizes the intelligence and precision level of health management.

[0160] In some embodiments, the physiological signal detection method for magnetic resonance scanning further includes saving the physiological signal. In some embodiments, the first preset threshold is generated by a health detection model.

[0161] The training method of the health detection model is: obtaining the historical record data and the real label set of the patient, each data in the historical record data has a one-to-one correspondence with one of the real labels in the real label set; inputting the target training data into the Transformer model to obtain a prediction threshold, and the target training data is any data in the historical record data; calculating the loss function based on the prediction threshold and the real label corresponding to the target training data; adjusting the network parameters of the Transformer model based on the loss function to obtain the health detection model.

[0162] The historical record data refers to the physiological signals obtained by real-time detection in the past. The historical record data may also include annotated initial data. The physiological signals obtained by real-time detection in the past may include heart rate (HR), heart rate variability (HRV), respiratory rate (RR) and blood oxygen saturation (SpO2), heartbeat interval (RR interval), heart rate change trend, respiratory cycle change, respiratory frequency value, frequency domain characteristics, waveform trend change and peak point, etc. The specific mechanism of the Transformer model includes multi-head self-attention, feed forward neural network, residual connection and layer normalization, wherein multi-head self-attention simultaneously focuses on the features of different positions through multi-head self-attention to capture long-distance dependencies in the sequence; the feed forward neural network performs nonlinear transformation on the features of each position to extract higher-level features; the residual connection avoids the problems of gradient disappearance and gradient explosion, and improves the training stability of the model; and layer normalization further improves the training stability and performance of the model. These mechanisms enable the Transformer model to effectively process multi-dimensional physiological signals, provide accurate prediction results, and help doctors detect potential health problems in a timely manner and take intervention measures.

[0163] Furthermore, the calculation formula of the loss function is: ;in, is the loss function, is the total number of historical record data, is the target training data (the first The corresponding true label of the prediction threshold of the data, is the prediction threshold.

[0164] Furthermore, the embodiment of the present application can also optimize the performance of the health detection model through cross-validation and hyperparameter tuning.

[0165] It should be understood that the embodiment of the present application regularly updates the health detection model according to real-time data to ensure that it can adapt to changes in the patient's physiological state. If it is detected that the value of the physiological parameter exceeds the normal range of the personalized setting (that is, the first preset threshold) or the change trend of the physiological parameter reaches a dangerous state, the health detection model will automatically adjust its internal parameters and immediately start the alarm mechanism or the early warning mechanism. This adjustment of internal parameters may involve changing feature weights, updating thresholds, or adjusting the decision boundaries of the model to better adapt to the patient's current physiological state. Exemplarily, if the patient's blood pressure continues to be high, the health detection model may increase the weight of blood pressure-related features and adjust the blood pressure threshold accordingly to more sensitively detect blood pressure changes.

[0166] In other embodiments, the Transformer model can be replaced by a convolutional neural network (CNN) or a long short-term memory network (LSTM), linear regression, a support vector machine, or a random forest.

[0167] It should also be mentioned that, considering the need for non-contact detection, wireless sensor technology or the combination of PPG, SCG and other physiological information detection technologies can be explored to replace existing optical sensing technologies. These technologies can further reduce interference to patients while providing real-time physiological parameter detection.

[0168] In addition, it is also possible to consider using bioelectric signals (such as electrocardiogram ECG) to replace existing optical sensing technology. Although the acquisition of bioelectric signals requires the use of contact sensors, the patient's discomfort can be minimized by improving the design of the sensor.

[0169] refer to Figure 5 FIG. 1 is a schematic block diagram of a physiological signal detection system for magnetic resonance scanning provided in accordance with the second aspect of an embodiment of the present application. Figure 5 In the embodiment, the physiological signal detection system 100 for magnetic resonance scanning comprises: The multi-spectral pulse wave signal acquisition module 101 is used to acquire the multi-spectral pulse wave signal of the face of the patient by using a time-series spectroscopic method during the process of scanning the patient by the magnetic resonance device of the physiological signal detection system for magnetic resonance scanning; A heart vibration signal acquisition module 102, configured to acquire a heart vibration signal of the patient's chest by using a vibration measurement method while acquiring the multi-spectral pulse wave signal; A first target signal acquisition module 103 is used to pre-process the multi-spectral pulse wave signal to obtain a first target signal; A second target signal obtaining module 104, configured to pre-process the heart vibration signal to obtain a second target signal; A physiological signal extraction module 105, configured to extract a physiological signal having multiple physiological parameters based on the first target signal and the second target signal; The feedback module 106 is used to feed back the physiological signal to medical personnel.

[0170] refer to Figure 6 FIG. 1 is a block diagram of a physiological signal detection system for magnetic resonance scanning provided in accordance with the third aspect of the present application. Figure 6 In the present invention, the physiological signal detection system 200 for magnetic resonance scanning includes: a processor 201, a memory 202, a multi-modal non-contact multi-parameter detection optical device 203, and a laser 204 (see Figure 2 The laser light in the magnetic resonance device 205, wherein the reference Figure 7 As shown, the multimodal non-contact multi-parameter detection optical device 203 includes a first camera 2031 and a second camera 2032, the memory 202 is used to store computer programs, the processor 201 is used to call and run the computer programs stored in the memory 202, and perform the steps of the physiological signal detection method for magnetic resonance scanning described in the first aspect of the embodiment of the present application, the magnetic resonance device 205, the first camera 2031 and the second camera 2032 are all connected to the processor 201 for communication, and the laser 204 is arranged on one side of the second camera 2032; The first camera 2031 is used to perform an image acquisition operation on the skin on the patient's face to obtain a first image sequence, and transmit the first image sequence to the processor 201; The second camera 2032 is used to perform an image acquisition operation on the patient's chest to obtain a second image sequence, and transmit the second image sequence to the processor 201; When the second camera 2032 performs an image acquisition operation on the chest, the laser 204 emits laser light to irradiate the chest.

[0171] In some embodiments, reference Figure 8 As shown, the physiological signal detection system 200 for magnetic resonance scanning also includes an early warning module 206. The early warning module 206 combines the dual mechanisms of alarm and early warning, which can not only quickly notify relevant personnel to deal with abnormal situations that have occurred, but also predict potential risks in advance based on trend changes in physiological characteristics, thereby gaining more response time for medical staff and patients, intervening in potential problems in advance, significantly reducing health risks, providing personalized decision-making support, and further ensuring the life safety of patients.

[0172] According to a fourth aspect of an embodiment of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium is used to store a computer program, wherein the computer program enables a computer to execute the steps of the physiological signal detection method for magnetic resonance scanning provided in the first aspect of the embodiment of the present application.

[0173] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0174] The technical features of the above embodiments can be combined without changing the basic principles of the present application. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0175] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the scope of patent protection of the present application shall be subject to the attached claims.

Claims

1. A physiological signal detection method for magnetic resonance scanning, characterized in that: include: In the process of scanning a patient by a magnetic resonance device of a physiological signal detection system for magnetic resonance scanning, a multi-spectral pulse wave signal of the patient's face is obtained by using a time-series spectroscopic method; While acquiring the multi-spectral pulse wave signal, using a vibration measurement method to acquire a heart vibration signal of the patient's chest; Preprocessing the multi-spectral pulse wave signal to obtain a first target signal; Preprocessing the heart vibration signal to obtain a second target signal; extracting a physiological signal having a plurality of physiological parameters based on the first target signal and the second target signal; The physiological signal is fed back to medical personnel.

2. The physiological signal detection method for magnetic resonance scanning according to claim 1, characterized in that: The steps of acquiring the multi-spectral pulse wave signal of the patient's face by using the time-series spectroscopic method include: receiving a first image sequence transmitted by a first camera of the physiological signal detection system for magnetic resonance scanning, wherein the first image sequence comprises an image sequence of multiple frame periods, and the image sequence of each frame period is an image sequence obtained by the first camera using multiple spectra of different wavelengths to sequentially photograph the face in a spectral order; In the first image sequence, first images corresponding to spectrums of the same wavelength are combined into a single spectrum video to obtain a plurality of spectrum videos; Performing a time series analysis on all pixels in a target spectral video, and using a skin reflectance model to model the change in the reflectance value of each pixel in the RGB channel of the target spectral video over time, to obtain a single spectrum pulse wave signal corresponding to the target spectral video, wherein the target spectral video is any one of the multiple spectral videos; All single-spectrum pulse wave signals constitute the multi-spectrum pulse wave signal.

3. The physiological signal detection method for magnetic resonance scanning according to claim 1, characterized in that: The steps of obtaining the cardiac vibration signal of the patient's chest by using the vibration measurement method include: receiving a second image sequence transmitted by a second camera of the physiological signal detection system for magnetic resonance scanning; Calculating a first motion amplitude in the X direction and a second motion amplitude in the Y direction of each second image in the second image sequence by using an optical flow method; combining the first motion amplitudes of all the second images in the X direction into a first motion amplitude sequence; combining the second motion amplitudes of all the second images in the Y direction into a second motion amplitude sequence; extracting motion angles from the first motion amplitude sequence and the second motion amplitude sequence using a sliding window method; The heart vibration signal is obtained based on the movement angle.

4. The physiological signal detection method for magnetic resonance scanning according to claim 1, characterized in that: The step of preprocessing the multi-spectral pulse wave signal to obtain a first target signal includes: Performing high-pass filtering on the multi-spectral pulse wave signal to obtain a first signal to be processed; Performing motion compensation processing on the first signal to be processed to obtain a second signal to be processed; Perform noise reduction processing on the second signal to be processed to obtain the first target signal.

5. The physiological signal detection method for magnetic resonance scanning according to claim 1, characterized in that: The step of preprocessing the heart vibration signal to obtain a second target signal comprises: Performing high-pass filtering on the heart vibration signal to obtain a third signal to be processed; Performing motion compensation processing on the third signal to be processed to obtain a fourth signal to be processed; Perform noise reduction processing on the fourth signal to be processed to obtain the second target signal.

6. The physiological signal detection method for magnetic resonance scanning according to claim 1, characterized in that: The step of extracting a physiological signal having a plurality of physiological parameters based on the first target signal and the second target signal comprises: Extracting the blood oxygen saturation of the physiological signal from the first target signal; Calculating the heart rate and heart rate variability of the physiological signal based on the second target signal; The respiratory rate of the physiological signal is calculated from the second target signal using an optical flow method.

7. The physiological signal detection method for magnetic resonance scanning according to claim 1, characterized in that: After the step of extracting a physiological signal having a plurality of physiological parameters based on the first target signal and the second target signal, the physiological signal detection method for magnetic resonance scanning further includes: When at least one physiological parameter of the physiological signal reaches a first preset threshold, an alarm mechanism is activated; When the trend change state of at least one physiological parameter of the physiological signal reaches a dangerous state, the early warning mechanism is activated.

8. A physiological signal detection system for magnetic resonance scanning, characterized in that: include: A multi-spectral pulse wave signal acquisition module, used for acquiring a multi-spectral pulse wave signal of the face of the patient by using a time-series spectroscopic method during the process of scanning the patient by a magnetic resonance device of a physiological signal detection system for magnetic resonance scanning; A cardiac vibration signal acquisition module, used to acquire the cardiac vibration signal of the patient's chest by using a vibration measurement method while acquiring the multi-spectral pulse wave signal; A first target signal acquisition module, used for preprocessing the multi-spectral pulse wave signal to obtain a first target signal; A second target signal acquisition module, used for preprocessing the heart vibration signal to obtain a second target signal; A physiological signal extraction module, configured to extract a physiological signal having a plurality of physiological parameters based on the first target signal and the second target signal; The feedback module is used to feed back the physiological signal to medical staff.

9. A physiological signal detection system for magnetic resonance scanning, characterized in that: include: A processor, a memory, a multimodal non-contact multi-parameter detection optical device, a laser and a magnetic resonance device, wherein the multimodal non-contact multi-parameter detection optical device includes a first camera and a second camera, the memory is used to store a computer program, the processor is used to call and run the computer program stored in the memory, and execute the steps of the physiological signal detection method for magnetic resonance scanning according to any one of claims 1 to 7, the magnetic resonance device, the first camera and the second camera are all connected to the processor in communication, and the laser is arranged on one side of the second camera; The first camera is used to perform an image acquisition operation on the skin on the patient's face to obtain a first image sequence, and transmit the first image sequence to the processor; The second camera is used to perform an image acquisition operation on the patient's chest to obtain a second image sequence, and transmit the second image sequence to the processor; When the second camera performs an image acquisition operation on the chest, the laser emits laser light to irradiate the chest.

10. A computer-readable storage medium, characterized in that: Used to store a computer program, wherein the computer program enables a computer to execute the steps of the physiological signal detection method for magnetic resonance scanning as described in any one of claims 1 to 7.

Images