A portable intelligent non-alcoholic fatty liver disease detection method and system

Abstract

The present invention relates to the field of data processing, and proposes a portable intelligent non-alcoholic fatty liver disease detection method and system. The attenuation coefficient and backscattering coefficient are optimized and extracted through a linear least squares method and a one-dimensional convolutional neural network, further improving the accuracy of detection. Quantitative relationship analysis is then performed through a nonlinear regression model, and ultrasonic fat fraction is quantitatively analyzed based on an energy absorption method, saving energy required for detection and meeting the requirements of portable detection. Different ultrasonic coefficients are combined to improve the accuracy of analysis, and a liver comprehensive verification model is designed for simulation verification processing, realizing intelligent verification and intelligent evaluation, avoiding dependence on professional knowledge, and making the detection results more consistent with actual conditions. The optimized results are visualized and displayed, making the detection results more intuitive and clearer, and improving the convenience of detection. The present invention improves the accuracy and convenience of the non-alcoholic fatty liver disease detection method.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a portable intelligent non-alcoholic fatty liver disease detection method and system. Background Art

[0002] As one of the most common liver diseases, non-alcoholic fatty liver disease is very important for early detection and quantitative assessment. However, current detection of non-alcoholic fatty liver disease often relies on expensive and complex imaging equipment, such as large magnetic resonance imaging equipment and ultrasound equipment. These expensive and complex imaging equipment require professional operation, resulting in obvious limitations in the regional popularity and convenience of non-alcoholic fatty liver disease detection. Therefore, the portability of non-alcoholic fatty liver disease detection has become an important link.

[0003] In the existing technology, non-alcoholic fatty liver disease detection is usually based on ultrasound detection to obtain ultrasound data for quantitative analysis. However, existing ultrasound detection methods often obtain ultrasound imaging data based on the shear stress method, which results in high energy required for overall detection and insufficient clarity of ultrasound imaging. For portable devices, it is difficult to maintain the purpose of long-term use. For operators of portable devices, the lack of intelligence makes it impossible to intuitively and clearly obtain test results and change trends, so that the detection still relies on the operator's professional knowledge, affecting the accuracy of portable detection.

[0004] Therefore, how to design a non-alcoholic fatty liver disease detection method that can meet the requirements of portable detection while improving the accuracy and intelligence of detection has become an urgent problem to be solved. Summary of the Invention

[0005] Based on this, the present invention proposes a portable intelligent non-alcoholic fatty liver disease detection method and system, which optimizes and extracts the attenuation coefficient and backscattering coefficient through linear least squares method and one-dimensional convolutional neural network to obtain more accurate liver fat status information, further improving the accuracy of detection, and then designs a nonlinear regression model for quantitative relationship analysis, and quantitatively analyzes the ultrasonic fat fraction based on the energy absorption method, saving the energy required for detection and meeting the requirements of portable detection. It also combines different ultrasonic coefficients and improves the accuracy of analysis, and designs a liver comprehensive verification model for simulation verification processing, realizing intelligent verification and intelligent evaluation of analysis results, avoiding dependence on professional technology and professional knowledge, making the test results more in line with actual conditions, and visually displaying the optimized results, making the test results more intuitive and clearer, and improving the convenience of detection. The present invention improves the accuracy and convenience of the non-alcoholic fatty liver disease detection method.

[0006] The present invention proposes a portable intelligent non-alcoholic fatty liver disease detection method, comprising:

[0007] Obtaining basic ultrasonic signal parameters and extracting ultrasonic comprehensive coefficients, wherein the ultrasonic comprehensive coefficients include attenuation coefficients and backscatter coefficients, and the extracted ultrasonic comprehensive coefficients are based on a linear least squares method and a one-dimensional convolutional neural network;

[0008] Performing a quantitative relationship analysis based on the ultrasound comprehensive coefficient to obtain an ultrasound fat score, wherein the quantitative relationship analysis is based on an energy absorption method and the ultrasound fat score is calculated based on a nonlinear regression model;

[0009] Perform B-mode ultrasound imaging processing according to a field programmable gate array architecture to obtain a target basic ultrasound image;

[0010] Performing simulation verification processing according to a comprehensive liver verification model to obtain a target optimized ultrasound image, wherein the comprehensive liver verification model includes a basic simulation model and a multi-coefficient verification model;

[0011] A visual evaluation report is generated according to the ultrasound comprehensive coefficient, the ultrasound fat fraction, and the target optimized ultrasound image. The visual evaluation report is based on an image segmentation model, and the image segmentation model includes morphological correction optimization and connected domain analysis optimization.

[0012] In summary, according to the above-mentioned portable intelligent non-alcoholic fatty liver disease detection method, the attenuation coefficient and backscattering coefficient are optimized and extracted by linear least squares method and one-dimensional convolutional neural network to obtain more accurate liver fat status information, thereby further improving the accuracy of detection. Then, a nonlinear regression model is designed for quantitative relationship analysis, and the ultrasonic fat fraction is quantitatively analyzed based on the energy absorption method, which saves the energy required for detection and meets the requirements of portable detection. Different ultrasonic coefficients are combined to improve the accuracy of analysis. A liver comprehensive verification model is designed for simulation verification processing, which realizes intelligent verification and intelligent evaluation of analysis results, avoids dependence on professional technology and professional knowledge, and makes the test results more in line with actual conditions. The optimized results are visualized and processed, making the test results more intuitive and clearer, thereby improving the convenience of detection. The present invention improves the accuracy and convenience of the non-alcoholic fatty liver disease detection method. Specifically, basic ultrasound signal parameters are obtained and ultrasound comprehensive coefficients are extracted. The ultrasound comprehensive coefficients include attenuation coefficients and backscatter coefficients. The extracted ultrasound comprehensive coefficients are based on linear least squares method and one-dimensional convolutional neural network to obtain more accurate liver fat status information, further improving the accuracy of detection. Quantitative relationship analysis is performed based on the ultrasound comprehensive coefficients to obtain ultrasound fat score. The quantitative relationship analysis is based on energy absorption method. The ultrasound fat score is calculated based on a nonlinear regression model. The ultrasound fat score is quantitatively analyzed based on the energy absorption method, which saves the energy required for detection and meets the requirements of portable detection. Different ultrasound coefficients are combined to improve the accuracy of analysis. B-mode ultrasound synthesis is performed according to the field programmable gate array architecture. Image processing is performed to obtain a target basic ultrasound image, and simulation verification processing is performed according to a liver comprehensive verification model to obtain a target optimized ultrasound image. The liver comprehensive verification model includes a basic simulation model and a multi-coefficient verification model, which realizes intelligent verification and intelligent evaluation of analysis results, avoids dependence on professional technology and professional knowledge, and makes the test results more in line with actual conditions. A visual evaluation report is performed according to the ultrasonic comprehensive coefficient, the ultrasonic fat score and the target optimized ultrasound image. The visual evaluation report is based on an image segmentation model, and the image segmentation model includes morphological correction optimization and connected domain analysis optimization, which makes the test results more intuitive and clearer, and improves the convenience of detection. The present invention improves the accuracy and convenience of the non-alcoholic fatty liver disease detection method.

[0013] Furthermore, the step of obtaining basic ultrasonic signal parameters and extracting ultrasonic comprehensive coefficients specifically includes:

[0014] Acquiring basic ultrasonic signal parameters, and performing data interception on a depth range in the basic ultrasonic signal parameters to determine a depth axis region;

[0015] Analyzing the basic ultrasonic signal parameters according to a sliding window to obtain envelope signal parameters, adjusting the sliding window according to target dynamics, solving the envelope signal parameters in the sliding window in the depth axis region, determining a depth value at the center of the sliding window in the depth axis region, and obtaining a two-dimensional matrix according to the depth value and the frequency value of the basic ultrasonic signal parameters;

[0016] Estimating the attenuation coefficient and the backscattering coefficient of the depth value and the frequency value according to the weak scattering Born approximation, and then solving and extracting them according to the linear least squares method to obtain the attenuation coefficient and the backscattering coefficient;

[0017] The attenuation coefficient and backscattering coefficient are input into a one-dimensional convolutional neural network for coefficient enhancement extraction.

[0018] Furthermore, the step of inputting the attenuation coefficient and the backscattering coefficient into a one-dimensional convolutional neural network to perform coefficient enhancement extraction specifically includes:

[0019] Performing a time-axis-depth conversion on the basic ultrasound signal parameters to intercept the time-axis information of the RF signal segment in the basic ultrasound signal parameters, and performing region of interest tracking adjustment based on a cross-correlation algorithm to dynamically adjust the sliding window of the time-axis region;

[0020] Inputting the attenuation coefficient and the backscatter coefficient into a one-dimensional convolutional neural network, wherein the one-dimensional convolutional neural network is constructed based on historical relevant detection parameters of the detection target;

[0021] Extracting basic ultrasonic features using depthwise separable convolution, dividing the basic ultrasonic features into multiple feature groups of different scales according to feature scale, performing channel attention enhancement processing on each feature group to obtain multiple attention-enhanced feature groups of different scales, and fusing the attention-enhanced feature groups step by step in order from low to high feature scales to obtain multi-scale ultrasonic features;

[0022] The multi-scale ultrasonic features are respectively input into the attenuation coefficient branch and the backscattering coefficient branch for coefficient extraction to obtain the enhanced attenuation coefficient and the enhanced backscattering coefficient respectively.

[0023] Furthermore, the step of performing quantitative relationship analysis based on the ultrasound comprehensive coefficient to obtain the ultrasound fat fraction specifically includes:

[0024] The ultrasound comprehensive coefficients were smoothed using Savitsky-Golay processing to exclude respiratory motion artifacts;

[0025] According to the sound field depth distribution in the region of interest in the basic ultrasonic signal parameters, the ultrasonic comprehensive coefficient is corrected by depth weighting;

[0026] The original value and interaction term of the ultrasound comprehensive coefficient after depth weighted correction are input into the nonlinear regression model to obtain the ultrasound fat score. The specific algorithm of the nonlinear regression model is as follows:

[0027] ,

[0028] ,

[0029] Among them, UDFF represents the raw ultrasound fat fraction, 、 、 represents the fitting coefficient, AC represents the attenuation coefficient in the ultrasonic comprehensive coefficient, BSC represents the backscattering coefficient in the ultrasonic comprehensive coefficient, (AC·BSC) represents the interaction term of the ultrasonic comprehensive coefficient, c represents a constant term, represents the ultrasound fat fraction after standardized mapping, represents the mapping coefficient, Represents a mapping constant.

[0030] Furthermore, the step of performing simulation verification processing based on the liver comprehensive verification model to obtain a target optimized ultrasound image specifically includes:

[0031] Performing simulation verification processing according to a comprehensive liver verification model, wherein the comprehensive liver verification model includes a basic simulation model and a multi-coefficient verification model;

[0032] The basic simulation model includes a basic scatterer model and a global attenuation constraint. The basic scatterer model includes regular scatterers and diffuse scatterers. The distribution of the regular scatterers is based on a chi-square distribution rule, the amplitude of the regular scatterers is based on a uniform distribution rule, and the distribution and amplitude of the diffuse scatterers are both based on a uniform distribution rule. The regular scatterers are used to adjust the isotropy of the spatial distribution of scattering points of biological soft tissue to map the one-dimensional scattering point distribution to a three-dimensional space. The distribution mapping is based on a Hilbert space filling curve.

[0033] The multi-coefficient verification model includes a basic scatterer model, a global attenuation constraint, and a backscattering constraint. The basic scatterer model of the multi-coefficient verification model is an elastic sphere model. The elastic sphere model adjusts the scatterer size and distribution based on the intervention theory model, and calculates the scattering intensity of a single elastic sphere model. The number density of the basic scatterer model of the multi-coefficient verification model is calculated based on liver biological characteristic information. The scattering intensity is linearly superimposed based on the scattering intensities of all elastic sphere models, and the scattering intensity is response convolved with the spatial pulse of the ultrasonic transducer to obtain an ultrasonic radio frequency echo signal.

[0034] Simulation verification and optimization are performed based on the ultrasonic radio frequency echo signal to obtain a target optimized ultrasonic image.

[0035] Furthermore, the step of generating a visual evaluation report based on the ultrasound comprehensive coefficient, the ultrasound fat fraction, and the target optimized ultrasound image specifically includes:

[0036] Preprocessing the target optimized ultrasound image, wherein the preprocessing includes speckle noise suppression processing and contrast enhancement processing, wherein the speckle noise suppression processing is based on non-local mean filtering, and the contrast enhancement processing is based on an adaptive histogram equalization algorithm with local contrast limitation;

[0037] performing segmentation processing on the pre-processed target optimized ultrasound image according to the image segmentation model to obtain a liver region mask, and then performing interference structure removal processing on the liver region mask, wherein the interference structures include vascular structures and mass structures;

[0038] Post-processing the liver region mask, the post-processing including morphological correction optimization and connected domain analysis optimization, performing a closing operation according to the morphological correction optimization algorithm to fill the void area, and then performing maximum area retention according to the connected domain analysis optimization to obtain an optimized liver region segmentation result;

[0039] Performing heatmap color mapping on the liver region optimization segmentation result according to the ultrasound fat score and the ultrasound comprehensive coefficient, and then converting the liver region optimization segmentation result into a color layer through transparency fusion to overlay it on the target optimized ultrasound image to obtain a visual display result;

[0040] The visualization results are parameterized and evaluated according to the ultrasonic fat fraction and the ultrasonic comprehensive coefficient to obtain a visualization evaluation report.

[0041] The present invention proposes a portable intelligent non-alcoholic fatty liver disease detection system, comprising:

[0042] A coefficient extraction module is used to obtain basic ultrasonic signal parameters and extract ultrasonic comprehensive coefficients, wherein the ultrasonic comprehensive coefficients include attenuation coefficients and backscatter coefficients, and the extracted ultrasonic comprehensive coefficients are based on a linear least squares method and a one-dimensional convolutional neural network;

[0043] a quantitative relationship analysis module, configured to perform a quantitative relationship analysis based on the ultrasound comprehensive coefficient to obtain an ultrasound fat score, wherein the quantitative relationship analysis is based on an energy absorption method and the ultrasound fat score is calculated based on a nonlinear regression model;

[0044] an ultrasound imaging module, configured to perform B-mode ultrasound imaging processing based on a field programmable gate array architecture to obtain a target basic ultrasound image;

[0045] A simulation verification module, configured to perform simulation verification processing based on a comprehensive liver verification model to obtain a target optimized ultrasound image, wherein the comprehensive liver verification model includes a basic simulation model and a multi-coefficient verification model;

[0046] A visualization display module is used to generate a visualization evaluation report based on the ultrasound comprehensive coefficient, the ultrasound fat score and the target optimized ultrasound image, wherein the visualization evaluation report is based on an image segmentation model, and the image segmentation model includes morphological correction optimization and connected domain analysis optimization.

[0047] Furthermore, the ultrasonic imaging module specifically includes: an ultrasonic transducer, a front-end signal transceiver processor, and a back-end ultrasonic image processor;

[0048] The front-end signal transceiver processor includes a transmitting front-end unit, a receiving front-end unit, a beam processing unit, and a signal processing unit. The transmitting front-end unit is used to generate high-voltage excitation pulses, focus digital electronics, beam scanning adjustment, and aperture size control. The receiving front-end unit is used for front-end amplification of echo signals, analog-to-digital conversion, and high-speed transmission. The beam processing unit is used for digital signal processing, variable aperture reception, dynamic tracking, and beam synthesis. The signal processing unit is used for gain control, dynamic filtering, envelope detection, secondary sampling, logarithmic compression, coordinate conversion processing, and transmits the final generated signal to the back-end ultrasonic image processor.

[0049] The present invention also provides a storage medium storing one or more programs, which, when executed by a processor, implement the portable intelligent non-alcoholic fatty liver disease detection method as described above.

[0050] The present invention further provides a computer device, comprising a memory and a processor, wherein:

[0051] The memory is used to store computer programs;

[0052] When the processor is used to execute the computer program stored in the memory, the portable intelligent non-alcoholic fatty liver disease detection method as described above is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a flow chart of the portable intelligent non-alcoholic fatty liver disease detection method proposed in the first embodiment of the present invention;

[0054] Figure 2 This is a schematic diagram of the structure of a portable intelligent non-alcoholic fatty liver disease detection system proposed in the second embodiment of the present invention.

[0055] The following specific embodiments will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION

[0056] To facilitate understanding of the present invention, the present invention will be described more fully below with reference to the accompanying drawings. The drawings illustrate several embodiments of the present invention. However, the present invention may be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and comprehensive understanding of the present invention.

[0057] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be an intermediate element. When an element is referred to as being "connected to" another element, it may be directly connected to the other element or there may be an intermediate element. The terms "vertical," "horizontal," "left," "right," and similar expressions used herein are for illustrative purposes only.

[0058] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one skilled in the art to which this invention pertains. The terms used in this specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0059] See also Figure 1 , which is a flow chart of a portable intelligent non-alcoholic fatty liver disease detection method proposed in the first embodiment of the present invention, the portable intelligent non-alcoholic fatty liver disease detection method includes steps S01 to S05, wherein:

[0060] Step S01: Obtain basic ultrasonic signal parameters and extract ultrasonic comprehensive coefficients;

[0061] It should be noted that in this embodiment, the ultrasonic comprehensive coefficient includes the attenuation coefficient and the backscatter coefficient. The extraction of the ultrasonic comprehensive coefficient is based on the linear least squares method and the one-dimensional convolutional neural network to obtain the basic ultrasonic signal parameters, and the depth range in the basic ultrasonic signal parameters is intercepted to determine the depth axis area;

[0062] Analyzing the basic ultrasonic signal parameters according to a sliding window to obtain envelope signal parameters, adjusting the sliding window according to target dynamics, solving the envelope signal parameters in the sliding window in the depth axis region, determining a depth value at the center of the sliding window in the depth axis region, and obtaining a two-dimensional matrix according to the depth value and the frequency value of the basic ultrasonic signal parameters;

[0063] Estimating the attenuation coefficient and the backscattering coefficient of the depth value and the frequency value according to the weak scattering Born approximation, and then solving and extracting them according to the linear least squares method to obtain the attenuation coefficient and the backscattering coefficient;

[0064] The principle formula for estimating the attenuation coefficient and backscattering coefficient is as follows:

[0065] ,

[0066] in, represents the weak scattering Born approximation, Represents the combined effect of electrical excitation and transducer, represents the diffraction effect value, represents the cumulative attenuation coefficient, represents the backscatter coefficient, Indicates the frequency value, Indicates the depth value;

[0067] Inputting the attenuation coefficient and the backscattering coefficient into a one-dimensional convolutional neural network to perform coefficient enhancement extraction;

[0068] Performing a time-axis-depth conversion on the basic ultrasound signal parameters to intercept the time-axis information of the RF signal segment in the basic ultrasound signal parameters, and performing region of interest tracking adjustment based on a cross-correlation algorithm to dynamically adjust the sliding window of the time-axis region;

[0069] Inputting the attenuation coefficient and the backscatter coefficient into a one-dimensional convolutional neural network, wherein the one-dimensional convolutional neural network is constructed based on historical relevant detection parameters of the detection target;

[0070] Extracting basic ultrasonic features using depthwise separable convolution, dividing the basic ultrasonic features into multiple feature groups of different scales according to feature scale, performing channel attention enhancement processing on each feature group to obtain multiple attention-enhanced feature groups of different scales, and fusing the attention-enhanced feature groups step by step in order from low to high feature scales to obtain multi-scale ultrasonic features;

[0071] The multi-scale ultrasonic features are respectively input into the attenuation coefficient branch and the backscattering coefficient branch for coefficient extraction to obtain the enhanced attenuation coefficient and the enhanced backscattering coefficient respectively.

[0072] Step S02: performing quantitative relationship analysis based on the ultrasound comprehensive coefficient to obtain the ultrasound fat fraction;

[0073] It should be noted that in this embodiment, the quantitative relationship analysis is based on the energy absorption method, the ultrasound fat fraction is calculated based on a nonlinear regression model, and the ultrasound comprehensive coefficient is subjected to Savitzky-Golay smoothing to eliminate respiratory motion artifacts;

[0074] According to the sound field depth distribution in the region of interest in the basic ultrasonic signal parameters, the ultrasonic comprehensive coefficient is corrected by depth weighting;

[0075] The original value and interaction term of the ultrasound comprehensive coefficient after depth weighted correction are input into the nonlinear regression model to obtain the ultrasound fat score. The specific algorithm of the nonlinear regression model is as follows:

[0076] ,

[0077] ,

[0078] Among them, UDFF represents the raw ultrasound fat fraction, 、 、 represents the fitting coefficient, AC represents the attenuation coefficient in the ultrasonic comprehensive coefficient, BSC represents the backscattering coefficient in the ultrasonic comprehensive coefficient, (AC·BSC) represents the interaction term of the ultrasonic comprehensive coefficient, c represents a constant term, represents the ultrasound fat fraction after standardized mapping, represents the mapping coefficient, Represents a mapping constant.

[0079] Step S03: performing B-mode ultrasound imaging processing according to a field programmable gate array architecture to obtain a target basic ultrasound image;

[0080] It should be noted that the B-mode ultrasound imaging process in this embodiment is based on an ultrasound imaging module, which specifically includes: an ultrasound transducer, a front-end signal transceiver processor, and a back-end ultrasound image processor;

[0081] The front-end signal transceiver processor includes a transmitting front-end unit, a receiving front-end unit, a beam processing unit, and a signal processing unit. The transmitting front-end unit is used to generate high-voltage excitation pulses, focus digital electronics, beam scanning adjustment, and aperture size control. The receiving front-end unit is used for front-end amplification of echo signals, analog-to-digital conversion, and high-speed transmission. The beam processing unit is used for digital signal processing, variable aperture reception, dynamic tracking, and beam synthesis. The signal processing unit is used for gain control, dynamic filtering, envelope detection, secondary sampling, logarithmic compression, coordinate conversion processing, and transmits the final generated signal to the back-end ultrasonic image processor.

[0082] Step S04: performing simulation verification processing according to the liver comprehensive verification model to obtain a target optimized ultrasound image;

[0083] It should be noted that in this embodiment, the liver comprehensive verification model includes a basic simulation model and a multi-coefficient verification model, and the simulation verification process is performed according to the liver comprehensive verification model, which includes a basic simulation model and a multi-coefficient verification model;

[0084] The basic simulation model includes a basic scatterer model and a global attenuation constraint. The basic scatterer model includes regular scatterers and diffuse scatterers. The distribution of the regular scatterers is based on a chi-square distribution rule, the amplitude of the regular scatterers is based on a uniform distribution rule, and the distribution and amplitude of the diffuse scatterers are both based on a uniform distribution rule. The regular scatterers are used to adjust the isotropy of the spatial distribution of scattering points of biological soft tissue to map the one-dimensional scattering point distribution to a three-dimensional space. The distribution mapping is based on a Hilbert space filling curve.

[0085] The multi-coefficient verification model includes a basic scatterer model, a global attenuation constraint, and a backscattering constraint. The basic scatterer model of the multi-coefficient verification model is an elastic sphere model. The elastic sphere model adjusts the scatterer size and distribution based on the intervention theory model, and calculates the scattering intensity of a single elastic sphere model. The number density of the basic scatterer model of the multi-coefficient verification model is calculated based on liver biological characteristic information. The scattering intensity is linearly superimposed based on the scattering intensities of all elastic sphere models, and the scattering intensity is response convolved with the spatial pulse of the ultrasonic transducer to obtain an ultrasonic radio frequency echo signal.

[0086] Simulation verification and optimization are performed based on the ultrasonic radio frequency echo signal to obtain a target optimized ultrasonic image.

[0087] Step S05: Producing a visual evaluation report based on the ultrasound comprehensive coefficient, ultrasound fat fraction, and target optimized ultrasound image;

[0088] It should be noted that in this embodiment, the visual evaluation report is based on an image segmentation model, which includes morphological correction optimization and connected domain analysis optimization. The target optimized ultrasound image is preprocessed, and the preprocessing includes speckle noise suppression processing and contrast enhancement processing. The speckle noise suppression processing is based on non-local mean filtering, and the contrast enhancement processing is based on an adaptive histogram equalization algorithm with local contrast limitation.

[0089] performing segmentation processing on the pre-processed target optimized ultrasound image according to the image segmentation model to obtain a liver region mask, and then performing interference structure removal processing on the liver region mask, wherein the interference structures include vascular structures and mass structures;

[0090] Post-processing the liver region mask, the post-processing including morphological correction optimization and connected domain analysis optimization, performing a closing operation according to the morphological correction optimization algorithm to fill the void area, and then performing maximum area retention according to the connected domain analysis optimization to obtain an optimized liver region segmentation result;

[0091] Performing heatmap color mapping on the liver region optimization segmentation result according to the ultrasound fat score and the ultrasound comprehensive coefficient, and then converting the liver region optimization segmentation result into a color layer through transparency fusion to overlay it on the target optimized ultrasound image to obtain a visual display result;

[0092] The visualization results are parameterized and evaluated according to the ultrasonic fat fraction and the ultrasonic comprehensive coefficient to obtain a visualization evaluation report.

[0093] In summary, according to the above-mentioned portable intelligent non-alcoholic fatty liver disease detection method, the attenuation coefficient and backscattering coefficient are optimized and extracted by linear least squares method and one-dimensional convolutional neural network to obtain more accurate liver fat status information, thereby further improving the accuracy of detection. Then, a nonlinear regression model is designed for quantitative relationship analysis, and the ultrasonic fat fraction is quantitatively analyzed based on the energy absorption method, which saves the energy required for detection and meets the requirements of portable detection. Different ultrasonic coefficients are combined to improve the accuracy of analysis. A liver comprehensive verification model is designed for simulation verification processing, which realizes intelligent verification and intelligent evaluation of analysis results, avoids dependence on professional technology and professional knowledge, and makes the test results more in line with actual conditions. The optimized results are visualized and processed, making the test results more intuitive and clearer, thereby improving the convenience of detection. The present invention improves the accuracy and convenience of the non-alcoholic fatty liver disease detection method. Specifically, basic ultrasound signal parameters are obtained and ultrasound comprehensive coefficients are extracted. The ultrasound comprehensive coefficients include attenuation coefficients and backscatter coefficients. The extracted ultrasound comprehensive coefficients are based on linear least squares method and one-dimensional convolutional neural network to obtain more accurate liver fat status information, further improving the accuracy of detection. Quantitative relationship analysis is performed based on the ultrasound comprehensive coefficients to obtain ultrasound fat score. The quantitative relationship analysis is based on energy absorption method. The ultrasound fat score is calculated based on a nonlinear regression model. The ultrasound fat score is quantitatively analyzed based on the energy absorption method, which saves the energy required for detection and meets the requirements of portable detection. Different ultrasound coefficients are combined to improve the accuracy of analysis. B-mode ultrasound synthesis is performed according to the field programmable gate array architecture. Image processing is performed to obtain a target basic ultrasound image, and simulation verification processing is performed according to a liver comprehensive verification model to obtain a target optimized ultrasound image. The liver comprehensive verification model includes a basic simulation model and a multi-coefficient verification model, which realizes intelligent verification and intelligent evaluation of analysis results, avoids dependence on professional technology and professional knowledge, and makes the test results more in line with actual conditions. A visual evaluation report is performed according to the ultrasonic comprehensive coefficient, the ultrasonic fat score and the target optimized ultrasound image. The visual evaluation report is based on an image segmentation model, and the image segmentation model includes morphological correction optimization and connected domain analysis optimization, which makes the test results more intuitive and clearer, and improves the convenience of detection. The present invention improves the accuracy and convenience of the non-alcoholic fatty liver disease detection method.

[0094] See also Figure 2 , which is a schematic diagram of the structure of a portable intelligent non-alcoholic fatty liver disease detection system proposed in the second embodiment of the present invention, the system includes:

[0095] A coefficient extraction module 10 is used to obtain basic ultrasonic signal parameters and extract ultrasonic comprehensive coefficients, wherein the ultrasonic comprehensive coefficients include attenuation coefficients and backscatter coefficients, and the extracted ultrasonic comprehensive coefficients are based on a linear least squares method and a one-dimensional convolutional neural network;

[0096] a quantitative relationship analysis module 20 for performing a quantitative relationship analysis based on the ultrasound comprehensive coefficient to obtain an ultrasound fat score, wherein the quantitative relationship analysis is based on an energy absorption method and the ultrasound fat score is calculated based on a nonlinear regression model;

[0097] An ultrasound imaging module 30 is configured to perform B-mode ultrasound imaging processing according to a field programmable gate array architecture to obtain a target basic ultrasound image;

[0098] A simulation verification module 40 is configured to perform simulation verification processing based on a comprehensive liver verification model to obtain a target optimized ultrasound image, wherein the comprehensive liver verification model includes a basic simulation model and a multi-coefficient verification model;

[0099] The visualization display module 50 is used to generate a visualization evaluation report based on the ultrasound comprehensive coefficient, the ultrasound fat fraction and the target optimized ultrasound image. The visualization evaluation report is based on an image segmentation model, which includes morphological correction optimization and connected domain analysis optimization.

[0100] The present invention also provides a computer storage medium having one or more programs stored thereon, which, when executed by a processor, implement the above-mentioned portable intelligent non-alcoholic fatty liver disease detection method.

[0101] The present invention also proposes a computer device comprising a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to implement the above-mentioned portable intelligent non-alcoholic fatty liver disease detection method.

[0102] Those skilled in the art will appreciate that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device), or in conjunction with such instruction execution system, apparatus, or device. For purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transmit a program for use by an instruction execution system, apparatus, or device, or in conjunction with such instruction execution system, apparatus, or device.

[0103] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting, or processing it in another suitable manner as necessary, and then storing it in a computer memory.

[0104] It should be understood that various components of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the aforementioned embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one or a combination of the following technologies known in the art may be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0105] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0106] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A portable intelligent non-alcoholic fatty liver disease detection method, characterized in that: include: Obtaining basic ultrasonic signal parameters and extracting ultrasonic comprehensive coefficients, wherein the ultrasonic comprehensive coefficients include attenuation coefficients and backscatter coefficients, and the extracted ultrasonic comprehensive coefficients are based on a linear least squares method and a one-dimensional convolutional neural network; The step of obtaining basic ultrasonic signal parameters and extracting ultrasonic comprehensive coefficients specifically includes: Acquiring basic ultrasonic signal parameters, and performing data interception on a depth range in the basic ultrasonic signal parameters to determine a depth axis region; Analyzing the basic ultrasonic signal parameters according to a sliding window to obtain envelope signal parameters, adjusting the sliding window according to target dynamics, solving the envelope signal parameters in the sliding window in the depth axis region, determining a depth value at the center of the sliding window in the depth axis region, and obtaining a two-dimensional matrix according to the depth value and the frequency value of the basic ultrasonic signal parameters; Estimating the attenuation coefficient and the backscattering coefficient of the depth value and the frequency value according to the weak scattering Born approximation, and then solving and extracting them according to the linear least squares method to obtain the attenuation coefficient and the backscattering coefficient; Inputting the attenuation coefficient and the backscattering coefficient into a one-dimensional convolutional neural network to perform coefficient enhancement extraction; The principle formula for estimating the attenuation coefficient and backscattering coefficient is as follows: ， in, represents the weak scattering Born approximation, Represents the combined effect of electrical excitation and transducer, represents the diffraction effect value, represents the cumulative attenuation coefficient, represents the backscatter coefficient, Indicates the frequency value, Indicates the depth value; The step of inputting the attenuation coefficient and the backscattering coefficient into a one-dimensional convolutional neural network to perform coefficient enhancement extraction specifically includes: Performing a time-axis-depth conversion on the basic ultrasound signal parameters to intercept the time-axis information of the RF signal segment in the basic ultrasound signal parameters, and performing region of interest tracking adjustment based on a cross-correlation algorithm to dynamically adjust the sliding window of the time-axis region; Inputting the attenuation coefficient and the backscatter coefficient into a one-dimensional convolutional neural network, wherein the one-dimensional convolutional neural network is constructed based on historical relevant detection parameters of the detection target; Extracting basic ultrasonic features using depthwise separable convolution, dividing the basic ultrasonic features into multiple feature groups of different scales according to feature scale, performing channel attention enhancement processing on each feature group to obtain multiple attention-enhanced feature groups of different scales, and fusing the attention-enhanced feature groups step by step in order from low to high feature scales to obtain multi-scale ultrasonic features; Inputting the multi-scale ultrasonic features into the attenuation coefficient branch and the backscattering coefficient branch respectively for coefficient extraction to obtain the enhanced attenuation coefficient and the enhanced backscattering coefficient respectively; Performing a quantitative relationship analysis based on the ultrasound comprehensive coefficient to obtain an ultrasound fat score, wherein the quantitative relationship analysis is based on an energy absorption method and the ultrasound fat score is calculated based on a nonlinear regression model; Perform B-mode ultrasound imaging processing according to a field programmable gate array architecture to obtain a target basic ultrasound image; Performing simulation verification processing according to a comprehensive liver verification model to obtain a target optimized ultrasound image, wherein the comprehensive liver verification model includes a basic simulation model and a multi-coefficient verification model; A visual evaluation report is generated according to the ultrasound comprehensive coefficient, the ultrasound fat fraction, and the target optimized ultrasound image. The visual evaluation report is based on an image segmentation model, and the image segmentation model includes morphological correction optimization and connected domain analysis optimization.

2. The portable intelligent non-alcoholic fatty liver disease detection method according to claim 1, characterized in that: The step of performing quantitative relationship analysis based on the ultrasound comprehensive coefficient to obtain the ultrasound fat fraction specifically includes: The ultrasound comprehensive coefficients were smoothed using Savitsky-Golay processing to exclude respiratory motion artifacts; According to the sound field depth distribution in the region of interest in the basic ultrasonic signal parameters, the ultrasonic comprehensive coefficient is corrected by depth weighting; The original value and interaction term of the ultrasound comprehensive coefficient after depth weighted correction are input into the nonlinear regression model to obtain the ultrasound fat score. The specific algorithm of the nonlinear regression model is as follows: ， ， Among them, UDFF represents the raw ultrasound fat fraction, 、 、 represents the fitting coefficient, AC represents the attenuation coefficient in the ultrasonic comprehensive coefficient, BSC represents the backscattering coefficient in the ultrasonic comprehensive coefficient, (AC·BSC) represents the interaction term of the ultrasonic comprehensive coefficient, c represents a constant term, represents the ultrasound fat fraction after standardized mapping, represents the mapping coefficient, Represents a mapping constant.

3. The portable intelligent non-alcoholic fatty liver disease detection method according to claim 1, characterized in that: The step of performing simulation verification processing based on the liver comprehensive verification model to obtain a target optimized ultrasound image specifically includes: Performing simulation verification processing according to a comprehensive liver verification model, wherein the comprehensive liver verification model includes a basic simulation model and a multi-coefficient verification model; The basic simulation model includes a basic scatterer model and a global attenuation constraint. The basic scatterer model includes regular scatterers and diffuse scatterers. The distribution of the regular scatterers is based on a chi-square distribution rule, the amplitude of the regular scatterers is based on a uniform distribution rule, and the distribution and amplitude of the diffuse scatterers are both based on a uniform distribution rule. The regular scatterers are used to adjust the isotropy of the spatial distribution of scattering points of biological soft tissue to map the one-dimensional scattering point distribution to a three-dimensional space. The distribution mapping is based on a Hilbert space filling curve. The multi-coefficient verification model includes a basic scatterer model, a global attenuation constraint, and a backscattering constraint. The basic scatterer model of the multi-coefficient verification model is an elastic sphere model. The elastic sphere model adjusts the scatterer size and distribution based on the intervention theory model, and calculates the scattering intensity of a single elastic sphere model. The number density of the basic scatterer model of the multi-coefficient verification model is calculated based on liver biological characteristic information. The scattering intensity is linearly superimposed based on the scattering intensities of all elastic sphere models, and the scattering intensity is response convolved with the spatial pulse of the ultrasonic transducer to obtain an ultrasonic radio frequency echo signal. Simulation verification and optimization are performed based on the ultrasonic radio frequency echo signal to obtain a target optimized ultrasonic image.

4. The portable intelligent non-alcoholic fatty liver disease detection method according to claim 1, characterized in that: The step of generating a visual evaluation report based on the ultrasound comprehensive coefficient, the ultrasound fat fraction, and the target optimized ultrasound image specifically includes: Preprocessing the target optimized ultrasound image, wherein the preprocessing includes speckle noise suppression processing and contrast enhancement processing, wherein the speckle noise suppression processing is based on non-local mean filtering, and the contrast enhancement processing is based on an adaptive histogram equalization algorithm with local contrast limitation; performing segmentation processing on the pre-processed target optimized ultrasound image according to the image segmentation model to obtain a liver region mask, and then performing interference structure removal processing on the liver region mask, wherein the interference structures include vascular structures and mass structures; Post-processing the liver region mask, the post-processing including morphological correction optimization and connected domain analysis optimization, performing a closing operation according to the morphological correction optimization algorithm to fill the void area, and then performing maximum area retention according to the connected domain analysis optimization to obtain an optimized liver region segmentation result; Performing heatmap color mapping on the liver region optimization segmentation result according to the ultrasound fat score and the ultrasound comprehensive coefficient, and then converting the liver region optimization segmentation result into a color layer through transparency fusion to overlay it on the target optimized ultrasound image to obtain a visual display result; The visualization results are parameterized and evaluated according to the ultrasonic fat fraction and the ultrasonic comprehensive coefficient to obtain a visualization evaluation report.

5. A portable intelligent non-alcoholic fatty liver disease detection system, characterized in that: include: A coefficient extraction module is used to obtain basic ultrasonic signal parameters and extract ultrasonic comprehensive coefficients, wherein the ultrasonic comprehensive coefficients include attenuation coefficients and backscatter coefficients, and the extracted ultrasonic comprehensive coefficients are based on a linear least squares method and a one-dimensional convolutional neural network; The step of obtaining basic ultrasonic signal parameters and extracting ultrasonic comprehensive coefficients specifically includes: Acquiring basic ultrasonic signal parameters, and performing data interception on a depth range in the basic ultrasonic signal parameters to determine a depth axis region; Analyzing the basic ultrasonic signal parameters according to a sliding window to obtain envelope signal parameters, adjusting the sliding window according to target dynamics, solving the envelope signal parameters in the sliding window in the depth axis region, determining a depth value at the center of the sliding window in the depth axis region, and obtaining a two-dimensional matrix according to the depth value and the frequency value of the basic ultrasonic signal parameters; Estimating the attenuation coefficient and the backscattering coefficient of the depth value and the frequency value according to the weak scattering Born approximation, and then solving and extracting them according to the linear least squares method to obtain the attenuation coefficient and the backscattering coefficient; Inputting the attenuation coefficient and the backscattering coefficient into a one-dimensional convolutional neural network to perform coefficient enhancement extraction; The principle formula for estimating the attenuation coefficient and backscattering coefficient is as follows: ， in, represents the weak scattering Born approximation, Represents the combined effect of electrical excitation and transducer, represents the diffraction effect value, represents the cumulative attenuation coefficient, represents the backscatter coefficient, Indicates the frequency value, Indicates the depth value; The step of inputting the attenuation coefficient and the backscattering coefficient into a one-dimensional convolutional neural network to perform coefficient enhancement extraction specifically includes: Performing a time-axis-depth conversion on the basic ultrasound signal parameters to intercept the time-axis information of the RF signal segment in the basic ultrasound signal parameters, and performing region of interest tracking adjustment based on a cross-correlation algorithm to dynamically adjust the sliding window of the time-axis region; Inputting the attenuation coefficient and the backscatter coefficient into a one-dimensional convolutional neural network, wherein the one-dimensional convolutional neural network is constructed based on historical relevant detection parameters of the detection target; Extracting basic ultrasonic features using depthwise separable convolution, dividing the basic ultrasonic features into multiple feature groups of different scales according to feature scale, performing channel attention enhancement processing on each feature group to obtain multiple attention-enhanced feature groups of different scales, and fusing the attention-enhanced feature groups step by step in order from low to high feature scales to obtain multi-scale ultrasonic features; Inputting the multi-scale ultrasonic features into the attenuation coefficient branch and the backscattering coefficient branch respectively for coefficient extraction to obtain the enhanced attenuation coefficient and the enhanced backscattering coefficient respectively; a quantitative relationship analysis module, configured to perform a quantitative relationship analysis based on the ultrasound comprehensive coefficient to obtain an ultrasound fat score, wherein the quantitative relationship analysis is based on an energy absorption method and the ultrasound fat score is calculated based on a nonlinear regression model; an ultrasound imaging module, configured to perform B-mode ultrasound imaging processing based on a field programmable gate array architecture to obtain a target basic ultrasound image; A simulation verification module, configured to perform simulation verification processing based on a comprehensive liver verification model to obtain a target optimized ultrasound image, wherein the comprehensive liver verification model includes a basic simulation model and a multi-coefficient verification model; A visualization display module is used to generate a visualization evaluation report based on the ultrasound comprehensive coefficient, the ultrasound fat score and the target optimized ultrasound image, wherein the visualization evaluation report is based on an image segmentation model, and the image segmentation model includes morphological correction optimization and connected domain analysis optimization.

6. The portable intelligent non-alcoholic fatty liver disease detection system according to claim 5, characterized in that: The ultrasonic imaging module specifically includes: an ultrasonic transducer, a front-end signal transceiver processor, and a back-end ultrasonic image processor; The front-end signal transceiver processor includes a transmitting front-end unit, a receiving front-end unit, a beam processing unit, and a signal processing unit. The transmitting front-end unit is used to generate high-voltage excitation pulses, focus digital electronics, beam scanning adjustment, and aperture size control. The receiving front-end unit is used for front-end amplification of echo signals, analog-to-digital conversion, and high-speed transmission. The beam processing unit is used for digital signal processing, variable aperture reception, dynamic tracking, and beam synthesis. The signal processing unit is used for gain control, dynamic filtering, envelope detection, secondary sampling, logarithmic compression, coordinate conversion processing, and transmits the final generated signal to the back-end ultrasonic image processor.

7. A storage medium, characterized in that: The storage medium stores one or more programs, which, when executed by the processor, implement the portable intelligent non-alcoholic fatty liver disease detection method according to any one of claims 1 to 4.

8. A computer device, characterized in that: The computer device comprises a memory and a processor, wherein: The memory is used to store computer programs; When the processor is used to execute the computer program stored in the memory, it implements the portable intelligent non-alcoholic fatty liver disease detection method according to any one of claims 1 to 4.

Images