Portable cardiac multi-modal intelligent imaging system and method

Abstract

This invention relates to a portable cardiac multimodal intelligent imaging system and method, belonging to the field of medical image processing technology. The system performs an initial ultrasound-photoacoustic fusion scan to collect sample information data and send it to the cloud. The collected data is preprocessed using imaging features and then compared to determine the most similar sample type. A twin neural network is used for feature recognition and registration. A minimum power consumption strategy for this type is sent to the beamforming network, and a photoacoustic emission strategy is sent to the pulsed laser. A precise multimodal scan of the sample is performed, and the collected sample information data is uploaded and sent to the cloud. The collected data is preprocessed using imaging features and then compared. This data is input into the network for training. After detail interpolation, correction, and multimodal imaging fusion using similar empirical images, the fused imaging result of the sample is obtained.

Description

Technical Field

[0001] This invention belongs to the field of medical image processing technology, and relates to a portable cardiac multimodal intelligent imaging system and method. Background Technology

[0002] Traditional ultrasound imaging equipment typically improves image quality by increasing the number of ultrasound signal processing channels, the number of transmitter / receiver elements, and the operating frequency. These measures, increasing the number of channels, lead to complex circuitry, high power consumption, large size, and increased manufacturing and quality control difficulties, resulting in high production costs. This severely hinders the widespread adoption of rapid and accurate medical diagnosis in field and out-of-hospital settings. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide a portable cardiac multimodal intelligent imaging system and method. By employing a series of phase consistency enhancement techniques for the receiving path of the receiving front-end, a high-phase-consistency receiving front-end based on ultrasound-photoacoustic fusion is constructed, achieving high-precision phased array scanning functionality. The significant increase in the number of control processing channels and the number of transmitting / receiving array elements solves the practical problems of complex circuitry, high power consumption, large size, and high production costs due to complex manufacturing processes and quality control. Through rapid transceiver beamforming controlled by multi-dimensional fusion using artificial intelligence, the scanning speed and channel utilization of array elements are improved, thereby increasing the overall refresh rate of the device and shortening the scanning and imaging time for multimodal imaging.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A portable cardiac multimodal intelligent imaging system, comprising an ultrasound transceiver phased array and pulsed laser module connected in sequence, an optical-acoustic fusion receiving front end, a phased array beamforming network module, and an artificial intelligence optical-acoustic multidimensional fusion cloud module;

[0006] The components in the ultrasonic transceiver phased array of the system undergo phase consistency screening:

[0007] Place the batch of tested components in an environment with a temperature of 22-27℃ and a relative humidity of 40-55%, and apply rated frequency, power signal and load to keep them in operation for 24 hours to achieve a stable working state.

[0008] The test environment should be maintained at 22-27℃ and relative humidity at 40-55%, with no detectable vibration or air turbulence. The vector network analyzer should be powered on and warmed up for ≥10 minutes. After the working display readings stabilize, use calibration parts with appropriate frequency and interface to perform calibration tests for open circuit, short circuit and load at both ports in sequence.

[0009] Select array elements or RC test fixtures according to the function of the components and test them at a predetermined frequency. Number each component and record its measured impedance and phase parameters. After the test is completed, the components are clustered and sorted according to the set numerical accuracy. At the same time, components whose phase difference from the average value of the same type and batch of components is ≥90° are removed. Components with N different phase numerical clusters are obtained.

[0010] According to the required number of array elements and components, take them from the component group of each phase value cluster. If the number of components in the cluster is sufficient, obtain the required number of array elements or the capacitors and resistors in the ultrasonic transceiver phased array and pulsed laser module, and the photoacoustic fusion receiver front end. If insufficient, take them from other value cluster component groups, or repeat the above steps to expand the number of available components.

[0011] The ultrasonic transceiver phased array and pulsed laser module in the system, and the capacitor and resistor phase consistency screening in the optoacoustic fusion receiving front end.

[0012] By selecting the RC phase test fixture and following the test sorting method in 1.1, the capacitors and resistors in the phase-consistent ultrasonic transceiver phased array and pulsed laser module and the optoacoustic fusion receiver front-end required for constructing the transceiver front-end can be obtained.

[0013] After the ultrasonic transceiver phased array completes a full-channel scan signal transmission from channel 1 to n, the reflected signal from the detected tissue returns to the full-channel transceiver phased array to generate image data. The upper-level processing system identifies and distinguishes the specific type of tissue based on AI imaging pattern comparison, and issues optimized ultrasonic transmission AI beamforming for the detected tissue to control the phased array to work in the corresponding ab channel for detecting the tissue, reducing the single phased array scan time, and adjusting the pulse laser emission second pulse parameter.

[0014] A portable cardiac multimodal intelligent imaging method based on the aforementioned system includes the following steps:

[0015] S1: Improve the overall phase consistency of the receiving path of the ultrasonic-optical fusion precision receiving front-end;

[0016] S2: Perform an initial ultrasound-photoacoustic fusion scan to collect sample information data and send it to the device cloud;

[0017] S3: The collected data is preprocessed through imaging features and then identified and compared to obtain the most similar type of sample. The Siamese neural network is used for feature identification and registration.

[0018] S4: The background sends the minimum power consumption strategy of this type to the beamforming network and the photoacoustic emission strategy to the pulsed laser to achieve efficient ultrasonic and photoacoustic transceiver beamforming with multi-dimensional fusion of artificial intelligence.

[0019] S5: Perform precise multimodal scanning on the sample, upload the sample information data collected by the scan, and send it to the cloud;

[0020] S6: The collected data is preprocessed for imaging and then identified and compared. The data is then input into the network for training. After detail interpolation, correction and multimodal imaging fusion of similar experience images, the fused imaging result of the sample is obtained.

[0021] Optionally, Gaussian noise is added to the detail interpolation of the image, and the Gaussian distribution is reduced to the original distribution during the denoising process using a reversible Markov chain. The diffusion model consists of latent variable models for two processes: a forward process and a reverse process. The forward process is defined by the following Markov chain:

[0022]

[0023] Where, x t It follows a Gaussian distribution over time t, with a mean of 1 / 2. Variance σ t (x t )=(1-α t )I, where α t It is a learnable variable that changes over time;

[0024] Learning diffusion by predicting the original data x0 can be viewed as a process of predicting noise; the optimization problems are expressed as follows:

[0025]

[0026] Where ε is the input noise, ε θ (x t (t) represents the predicted noise at time t, and finally, image data interpolation is performed; a cross-modal attention mechanism is used in the image registration process.

[0027]

[0028] Among them, c i and p j The features of c and p at positions i and j, θ(.), Both g(.) and f(.) are linear embeddings, and f(.) = exp(.). In the equation, f(.) computes a scalar representing the position, c i and p j The correlation between features; result y i It is a normalized summary of the features at all positions of P, weighted by their correlation with the cross-modal features at position i; therefore, by y i The resulting matrix Y integrates nonlocal information from each position from P to C;

[0029] Optionally, in step S2, the series of data is sent to the device cloud, and a Siamese neural network is used for feature recognition and comparison to obtain the type discrimination loss of the most similar samples:

[0030]

[0031] in, This is used to calculate the Euclidean distance between two sample features X1 and X2. P represents the feature dimension of the sample, Y is the label indicating whether the two samples match, Y=1 means the two samples are similar or match, Y=0 means they do not match, m is the set threshold, and N is the number of samples.

[0032] The most suitable ultrasonic transceiver beamforming strategy and photoacoustic transceiver strategy for this type of sample are delivered from the cloud to the beamforming network and pulsed laser, realizing rapid transceiver beamforming with multi-dimensional integration of artificial intelligence.

[0033] The beneficial effects of this invention are: by improving the overall phase consistency of the receiving path of the precision receiving front end through ultrasonic-photoacoustic fusion, the number of array elements and channels of the device is reduced.

[0034] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0035] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein:

[0036] Figure 1 This is a schematic diagram of the overall concept of the present invention;

[0037] Figure 2 This is a system composition diagram of the present invention.

[0038] Figure 3 A process for screening the phase consistency of capacitors and resistors in array elements / ultrasonic transceiver phased arrays and pulsed laser modules, and optoacoustic fusion receiving front-ends.

[0039] Figure 4 This is a schematic diagram of waveguide-like embedded wiring.

[0040] Figure 5 This is a schematic diagram of rapid transmission beamforming based on multi-mode fusion of artificial intelligence. Detailed Implementation

[0041] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0042] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0043] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0044] Please refer to the examples. Figures 1-5 .

[0045] 1. Improved overall phase consistency of the receiving path in the ultrasonic-optical acoustic fusion precision receiving front-end

[0046] Improving the overall phase consistency of the front-end receiving path is one of the most effective ways to enhance the imaging performance of ultrasonic phased arrays. Compared to the current industry practice of increasing the number of array elements and channels, it has the advantages of using fewer components, simpler circuit structure, and thus more PCB layout space. However, it is rarely used because the industry currently lacks effective solutions to ensure high phase consistency of components / arrays, PCB traces, and capacitors and resistors in the ultrasonic transceiver phased array, pulsed laser module, and photoacoustic fusion receiving front-end. This invention achieves high phase consistency of the receiving path of the receiving front-end through the following method.

[0047] 1.1 Screening for Phase Consistency of Components in Ultrasonic Transceiver Phased Arrays

[0048] Elements in single-unit and array-based ultrasonic transceiver phased arrays are typically made of PZT ceramic sintering. Composites of ceramic and polymer materials, silicon-based materials, and other crystals with significant piezoelectric effects are fabricated using MEMS methods. Due to the structural dimensional losses caused by uncertain processes, it is difficult to ensure phase consistency between individual array elements without screening, making it challenging to achieve impedance and phase accuracy errors better than 0.1% for 32 or more individual elements. Various new material MEMS processes have been used to fabricate array transducers due to their high processing precision; however, their high processing cost limits their application to small arrays in the millimeter range. For centimeter-scale or larger arrays with more than 256 elements, the overall yield rate for accuracy and phase consistency remains low. Furthermore, the high cost of large-size MEMS fabrication and the resulting yield issues severely hinder the mass production of transducer arrays.

[0049] This invention employs a pre-defined, streamlined testing and screening method to screen components in an ultrasonic transceiver phased array for phase consistency.

[0050] (1) Preparation for component aging and stabilization

[0051] Place the batch of tested components in an environment with a temperature of 22-27°C and a relative humidity of 40-55%, and apply rated frequency, power signal and load to keep them in operation for 24 hours to achieve a stable working state.

[0052] (2) Test Environment Preparation

[0053] The test environment should be maintained at 22-27℃ and relative humidity at 40-55%, with no detectable vibration or air turbulence. The vector network analyzer should be powered on and allowed to warm up for ≥10 minutes. After the working display readings stabilize, calibration tests for open circuit, short circuit, and load at both ports should be performed sequentially using calibration components with appropriate frequencies and interfaces.

[0054] (3) Phase testing and sorting

[0055] Select array elements or RC test fixtures according to the component's function and test it at a predetermined frequency. Number each component and record its measured impedance and phase parameters. After testing, perform phase value clustering and sorting on the components according to the set numerical accuracy (e.g., 0.1%, 0.5%). At the same time, discard components whose phase difference from the average value of the same type and batch of components is ≥90°. Obtain N component groups with different phase value clusters.

[0056] (4) Statistical analysis of phase numerical clustering elements

[0057] According to the required number of array elements and components (e.g., 32, 64, 96, etc.), take from the component group of each phase numerical cluster. If the number of components in the cluster is sufficient, the required number of array elements or ultrasonic transceiver phased array and pulsed laser module, and capacitors and resistors in the photoacoustic fusion receiver front end are obtained. If insufficient, take from other numerical cluster component groups, or repeat steps 1-3 to expand the number of available components.

[0058] 1.2 Phase Consistency Screening of Capacitors and Resistors in Ultrasonic Transceiver Phased Array and Pulsed Laser Modules, and Opto-Acoustic Fusion Receiving Front End

[0059] By selecting the RC phase test fixture and following the test sorting method in 1.1, the capacitors and resistors in the phase-consistent ultrasonic transceiver phased array and pulsed laser module and the optoacoustic fusion receiver front-end required for constructing the transceiver front-end can be obtained.

[0060] 1.3 Refined Design of PCB Routing Phase Consistency

[0061] In the receiving path of the ultrasonic-optical fusion receiver front-end, the PCB routing of each channel is also a crucial component. Compared to current ultrasonic equipment designs that typically use outer layer traces on the top or bottom layer of the PCB and control the trace length of each channel to achieve impedance matching, this invention employs a waveguide-like embedded routing design that isolates the traces from the external environment and between channels, thereby achieving a high degree of phase consistency among the traces in each channel.

[0062] In the waveguide-like embedded routing design schematic, A is a top view of the three-channel partial routing, and B is a routing cross-section. a, b, and c represent the routing of three adjacent channels at the receiving front end. Feature 1: The routing is contained within the center of the PCB, unaffected by external signal interference. Testing shows a shielding efficiency of ≥70dB relative to outer layer routing. Feature 2: Vias are symmetrically distributed on both sides of routing a, and similar via arrangements are also present in the unshown portions of b and c. The vias connect to the upper and lower ground planes, forming a phase-stable coaxial waveguide-like structure with the routing. Therefore, the isolation between channels is significantly improved compared to routing without vias. Testing shows that, while maintaining the same spacing, adding vias can achieve an isolation improvement of ≥40dB. Simultaneously, the waveguide-like structure also provides higher impedance stability than typical surface routing, thus achieving highly phase-consistent channel routing.

[0063] 2. Rapid transmit / receive beamforming with multi-dimensional integration of artificial intelligence

[0064] Based on the phased array scanning imaging principle, an artificial intelligence feedback model is used to determine a reasonable number of scanning array elements and array shape. After optimizing beamforming, the phased array can achieve a shorter scanning time, thereby improving the speed of the entire scanning imaging process.

[0065] 2.1 Rapid Transmission Beamforming Based on Artificial Intelligence Multimodal and Multidimensional Fusion

[0066] After transmitting signals across all channels (1 to n) in a single ultrasound transceiver phased array scan, the reflected signals from the detected tissue return to the phased array, generating image data. The upper-level processing system can identify the specific tissue type based on AI-generated image comparison and then issue optimized ultrasound transmission AI beamforming for the detected tissue. This controls the phased array to operate at the most suitable ab channel for detecting that tissue, reducing the single phased array scan time and adjusting the pulsed laser emission pulse parameters to improve photoacoustic imaging quality. This same principle and operating method is not limited to the one-dimensional linear phased array described in the figure; it also applies to improving the imaging scanning speed of 1.5D and 2D phased arrays with horizontally distributed phased array elements.

[0067] 2.2 Rapid Receiving Beamforming Based on Artificial Intelligence Multimodal and Multidimensional Fusion

[0068] The receiving beamforming of ultrasonic transceiver phased arrays can also identify the specific type of tissue based on AI imaging pattern comparison. The upper-level processing system of the equipment issues optimized receiving AI beamforming for the specific tissue being detected, controlling the phased array to operate in the most suitable ab channel for detecting that tissue, thereby reducing the scanning time of a single phased array. The same principle and working method are not limited to the one-dimensional linear phased arrays described in the figure, but also applicable to improving the imaging scanning speed of 1.5-dimensional and 2-dimensional phased arrays with phased array elements distributed laterally.

[0069] In one specific embodiment of the present invention, the system operation method includes the following steps:

[0070] S1: Improve the overall phase consistency of the receiving path of the ultrasonic-optical fusion precision receiving front-end according to the technical solution of the present invention;

[0071] S2: Perform an initial ultrasound-photoacoustic fusion scan to collect sample information data and send it to the device cloud;

[0072] S3: The collected data is preprocessed with imaging features and then identified and compared to obtain the most similar type of sample. The main method used is Siamese neural network for feature identification and registration.

[0073] S4: The background sends the minimum power consumption strategy of this type to the beamforming network and the photoacoustic emission strategy to the pulsed laser to achieve efficient ultrasonic and photoacoustic transceiver beamforming with multi-dimensional fusion of artificial intelligence.

[0074] S5: Perform precise multimodal scanning on the sample, upload the sample information data collected by the scan, and send it to the cloud;

[0075] S6: The collected data is preprocessed for imaging and then compared and identified. The data is then input into the network for training. After detail interpolation, correction and multimodal imaging fusion of similar experience images, more accurate fused imaging results of the samples are obtained.

[0076] Image detail interpolation adds Gaussian noise, and during denoising, a reversible Markov chain reduces the Gaussian distribution back to the original distribution. The diffusion model consists of latent variable models for both the forward and reverse processes. The forward process is defined by the following Markov chain:

[0077]

[0078] Where, x t It follows a Gaussian distribution over time t, with a mean of 1 / 2. Variance σ t (x t )=(1-α t )I, where α t It is a learnable variable that changes over time.

[0079] Learning diffusion by predicting the original data x0 can be viewed as a process of predicting noise. The optimization problem is expressed as follows:

[0080]

[0081] Where ε is the input noise, ε θ (x t Let ,t) represent the predicted noise at time t, and finally, image data interpolation is performed. A cross-modal attention mechanism is used during image registration.

[0082]

[0083] Among them, c i and p j The features of c and p at positions i and j, θ(.), Both G(.) and f(.) are linear embeddings, and f(.) = exp(.). In the equation, f(.) computes a scalar representing the position, c. i and p j The correlation between the features. Result y i It is a normalized summary of the features at all positions of P, weighted by their correlation with the cross-modal features at position i. Therefore, by y i The resulting matrix Y integrates nonlocal information from each position from P to C.

[0084] Furthermore, the technical solution of this invention achieves a comprehensive improvement in the phase consistency of the receiving path of the ultrasonic-optical-acoustic fusion precision receiving front-end through the phase consistency screening of components in the ultrasonic transceiver phased array, the phase consistency screening of capacitors and resistors in the ultrasonic transceiver phased array and pulsed laser module, the phase consistency screening of capacitors and resistors in the optical-acoustic fusion receiving front-end, and the refined design of PCB wiring phase consistency.

[0085] Using a fusion imaging device that incorporates a precision receiving front-end for ultrasound-photoacoustic fusion and improves the phase consistency of the receiving path, ultrasound B-mode and photoacoustic initial scans are performed on the sample to obtain a series of initial scan data on the ultrasonic time difference and photoacoustic propagation time of different surfaces on the corresponding cross section of the transducer array.

[0086] The data series was sent to the device's cloud platform, and a Siamese neural network was used for feature recognition and comparison. The loss for determining the most similar type of sample was:

[0087]

[0088] in, This is used to calculate the Euclidean distance between two sample features X1 and X2. P represents the feature dimension of the sample, Y is the label indicating whether the two samples match, Y=1 means the two samples are similar or match, Y=0 means they do not match, m is the set threshold, and N is the number of samples.

[0089] The most suitable ultrasonic transceiver beamforming strategy and photoacoustic transceiver strategy for this type of sample are delivered from the cloud to the beamforming network and pulsed laser, realizing rapid transceiver beamforming with multi-dimensional integration of artificial intelligence.

[0090] Furthermore, by using a beamforming network and components in an ultrasonic transceiver phased array to form an ultrasonic phased array, beam focusing and deflection are achieved, enabling precise control of a series of targeted imaging parameters such as the focusing depth, beam angle, focal column position, and focal spot size of the phased array. This allows for various high-precision fan-shaped scanning, linear scanning, dynamic depth scanning, and continuous Doppler blood flow velocity acquisition functions required for sample imaging.

[0091] The obtained precise scanning data is uploaded and sent to the device cloud. The collected data is preprocessed for imaging and then identified and compared. The data is then input into the network for training. After detail interpolation, correction and multimodal fusion of similar experience images, a more accurate multifunctional fusion imaging result is obtained.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A portable cardiac multimodal intelligent imaging system, characterized in that: The system includes an ultrasonic transceiver phased array and pulsed laser module, an optical-acoustic fusion receiving front end, a phased array beamforming network module, and an artificial intelligence optical-acoustic multi-dimensional fusion cloud module that are connected in sequence for data transmission. The components in the ultrasonic transceiver phased array undergo phase consistency screening: Place the batch of tested components in an environment with a temperature of 22~27℃ and a relative humidity of 40~55%, and apply rated frequency, power signal and load to keep them in operation for 24 hours to achieve a stable working state. The test environment is maintained at 22~27℃ and relative humidity at 40~55%, with no detectable vibration or air turbulence. The vector network analyzer is powered on and preheated for ≥10 minutes. After the working display readings are stable, the calibration tests of two-port open circuit, short circuit and load are completed in sequence using calibration parts with appropriate frequency and interface. Select array elements or RC test fixtures according to the function of the components and test them at a predetermined frequency. Number each component and record its measured impedance and phase parameters. After the test is completed, the components are clustered and sorted according to the set numerical accuracy. At the same time, components whose phase difference from the average value of the same type and batch of components is ≥90° are removed. Components with N different phase numerical clusters are obtained. According to the required number of array elements and components, take them from the component group of each phase value cluster. If the number of components in the component group of the phase value cluster is sufficient, obtain the required number of array elements or the capacitors and resistors in the ultrasonic transceiver phased array and pulsed laser module, and the photoacoustic fusion receiving front end. If insufficient, take them from the component group of other value clusters, or repeat the above steps to expand the number of available components. Phase consistency screening of capacitors and resistors in the ultrasonic transceiver phased array and pulsed laser module, and the photoacoustic fusion receiving front end: Select the RC phase test fixture and obtain the capacitors and resistors in the phase-consistent ultrasonic transceiver phased array and pulsed laser module and the optoacoustic fusion receiver front-end required for constructing the transceiver front-end according to the aforementioned phase consistency screening method. The ultrasonic transceiver phased array completes a cycle from 1 to... n After the full-channel scanning signal is transmitted, the reflected signal from the detected tissue returns to the full-channel transceiver phased array to generate image data. The upper-level processing system identifies and distinguishes the specific type of tissue based on AI imaging pattern comparison, and issues optimized ultrasonic transmission AI beamforming for the detected tissue to control the phased array to detect the corresponding tissue. a - b Channel operation reduces the single phased array scan time and the pulsed laser emission second pulse parameter.

2. A portable cardiac multimodal intelligent imaging method based on the system of claim 1, characterized in that: The method includes the following steps: S1: Improve the overall phase consistency of the receiving path of the ultrasonic-optical fusion precision receiving front-end; S2: Perform an initial ultrasound-photoacoustic fusion scan to collect sample information data and send it to the device cloud; S3: The collected data is preprocessed through imaging features and then identified and compared to obtain the most similar type of sample. The Siamese neural network is used for feature identification and registration. S4: The background sends the minimum power consumption strategy of this type to the beamforming network and the photoacoustic emission strategy to the pulsed laser to achieve efficient ultrasonic and photoacoustic transceiver beamforming with multi-dimensional fusion of artificial intelligence. S5: Perform precise multimodal scanning on the sample, upload the sample information data collected by the scan, and send it to the cloud; S6: The collected data is preprocessed for imaging and then identified and compared. The identified and compared data is then input into the diffusion model for training. After detail interpolation, correction and multimodal imaging fusion of similar experience images, the fused imaging result of the sample is obtained.

3. The portable cardiac multimodal intelligent imaging method according to claim 2, characterized in that: Gaussian noise is added to the image detail interpolation, and during the denoising process, the Gaussian distribution is reduced to the original distribution using a reversible Markov chain. The diffusion model consists of latent variable models for two processes: a forward process and a reverse process. The forward process is defined by the following Markov chain: in, It is in time t Gaussian distribution, conditional distribution It follows a Gaussian distribution with a mean of 1 / 2. The covariance matrix is ,in It is a learnable variable that changes over time; By predicting the original data Learning diffusion can be viewed as a process of predicting noise; the optimization problems are expressed as follows: in, For input noise, for t The prediction noise at each time step is eliminated, and image data interpolation is finally completed; a cross-modal attention mechanism is used in the image registration process. in, and yes and In position and Features ， and They are all linear embeddings. =exp(.); In the equation, Calculate a scalar representing position. and The correlation between features; results yes Normalized summaries of features for all locations, through their relationship with location i Correlation weighting of cross-modal features; by The matrix formed Integration from arrive Non-local information at each location.

Images