A three-dimensional imaging and interactive method of ultrasonic detector

Abstract

The present invention discloses a three-dimensional imaging and interaction method of an ultrasonic detector, comprising the following steps: obtaining a fitted probe shape; determining the probe pressure through feedback information; obtaining multi-level and multi-angle ultrasonic signals; evaluating the stability and clarity of each signal based on the multi-frequency ultrasonic signal; obtaining an optimized ultrasonic signal using adaptive filtering and beamforming processing; obtaining the geometric and material properties of the target area using inverse signal processing and time-frequency analysis, and adjusting the ultrasonic emission parameters to achieve dynamic focusing; reconstructing the structure and obtaining a three-dimensional image based on the dynamic focusing parameters and the optimized ultrasonic signal; identifying key features in the image based on the generated three-dimensional image; and obtaining targeted diagnostic auxiliary information based on the key features in the marked three-dimensional image. The three-dimensional imaging and interaction method of the ultrasonic detector of the present invention can automatically identify lesions, provide auxiliary diagnosis, and enhance the user interaction experience.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a three-dimensional imaging and interaction method of an ultrasonic detector. Background Art

[0002] In current medical practice, ultrasound detectors are widely used diagnostic tools that can provide real-time, non-invasive medical images and are essential for the diagnosis and monitoring of a variety of diseases. Although traditional ultrasound systems have achieved remarkable clinical application results, there are still some limitations and challenges, especially in terms of adaptability, image quality and user interaction. With the rapid development of medical technology, the requirements for ultrasound detection systems are also increasing, and a new generation of ultrasound detectors is urgently needed to meet higher clinical needs. First, the shape and hardness of traditional ultrasound probes limit their application on different curved surfaces and complex anatomical structures. For irregular body surfaces or deep tissues, fixed-shaped probes are difficult to achieve a good fit with the surface being detected, which may cause image blur or signal loss, affecting the accuracy and reliability of diagnosis. Therefore, it has become an urgent need to develop an ultrasound probe system that can adapt to different curved surfaces and provide optimized contact. Secondly, as the requirements for detail resolution and depth perception in medical diagnosis continue to increase, the limitations of traditional ultrasound technology in multi-level and multi-angle imaging capabilities are gradually emerging. A technology that can provide richer depth information and high-resolution images is needed to more accurately detect and analyze tiny structures and complex tissues in the body. In addition, most of the existing ultrasound detection systems rely on the operator's experience and skills, which to some extent limits the universality and standardization of ultrasound detection. For non-professionals, it may be a challenge to operate complex ultrasound equipment for accurate diagnosis. Finally, with the rise of precision medicine and personalized treatment, there is a growing demand for more intelligent and automated diagnostic tools. However, the existing ultrasound detection systems still have room for improvement in terms of intelligence, especially in terms of automatic identification of lesions, providing auxiliary diagnosis, and enhancing user interaction experience. Therefore, although ultrasound detection technology has made significant progress in the past few decades, it still faces a series of challenges, and technological innovation is needed to further improve its application efficiency and popularity in modern medical diagnosis. Summary of the invention

[0003] In order to solve the technical problems existing in the above background technology, one of the purposes of the present invention is to provide a three-dimensional imaging and interaction method of an ultrasonic detector, comprising the following steps:

[0004] S0, based on the surface features of the object being tested, a deformable ultrasonic probe driven by a deep learning algorithm is used to obtain a probe shape that fits various curved surfaces;

[0005] S1, by real-time monitoring of the contact between the probe and the detected surface, feedback information of the contact force and area is obtained to determine the probe pressure;

[0006] S2, based on the probe pressure and surface feature information, uses multi-frequency ultrasonic emission technology to adjust the emission frequency to meet the detection needs of different depths and tissue types, and obtains multi-level and multi-angle ultrasonic signals;

[0007] S3, based on the multi-frequency ultrasonic signals at different depths and angles, the characteristics of the multi-frequency ultrasonic signals in the time and frequency domains are determined using Fourier transform and wavelet transform, and the stability and clarity of each signal are evaluated;

[0008] S4, based on the signal stability and clarity evaluation results, adaptive filtering and beamforming are used to perform directional enhancement and interference suppression on the original ultrasonic signal to obtain an optimized ultrasonic signal;

[0009] S5, using the optimized ultrasonic signal, inverse signal processing and time-frequency analysis to obtain the geometric and material characteristics of the target area, and adjust the ultrasonic emission parameters, including the focal depth, focal position and size of the focal area, to achieve dynamic focusing;

[0010] S6, reconstructing the three-dimensional structure of the detected object according to the dynamic focusing parameters and the optimized ultrasonic signal to obtain a continuous and complete three-dimensional image;

[0011] S7, identifying key features in the image according to the generated continuous three-dimensional images, and marking and classifying the key features;

[0012] S8, based on the key features in the marked three-dimensional image, the features in the image are matched with the known pathological features in the clinical knowledge base through a feature matching method to obtain targeted diagnostic auxiliary information.

[0013] The preferred technical solution of the present invention is that, in S0, a convolutional neural network is used to capture and analyze the surface of the object to be detected in real time to identify the geometric features of the surface, including curvature, texture and irregular shape; through the piezoelectric driver in the micromechanical structure, the control parameters are adjusted to drive the probe material to deform, and the shape of the probe is adjusted to achieve precise fit with the surface to be detected, thereby obtaining a probe shape that fits different curved surfaces.

[0014] The preferred technical solution of the present invention is that in S1, when changes in contact force and contact area are detected, an analog signal is obtained through a sensor, the analog signal includes feedback information of the contact force and area, and is converted into a digital signal; the converted digital signal is received by a controller, and the digital signal is analyzed according to a preset PID control algorithm to determine whether it is necessary to adjust the pressure applied to the probe.

[0015] The preferred technical solution of the present invention is that, in S2, information on the patient's specified tissue type and detection depth is obtained, including the acoustic impedance of the tissue, the expected detection depth, and the anatomical characteristics of the target area; according to the detection requirements, if the requirement is high-resolution shallow imaging, a high-frequency ultrasonic transmission frequency is selected;

[0016] The contact between the probe and the detection surface is simulated by the acoustic simulation system to evaluate the coupling effect of the sound wave under specified conditions. If the simulation results show that the acoustic wave coupling effect is lower than the preset effect threshold, the probe pressure is increased until the optimal coupling state is reached.

[0017] The preferred technical solution of the present invention is that, in S3, the amplitude spectrum of the multi-frequency ultrasonic signal is obtained by fast Fourier transform, and if the amplitude spectrum shows that the signal has a high energy peak in the specified frequency band, it is determined that the signal has high energy concentration in the frequency band;

[0018] The fast Fourier transform algorithm is used to analyze the phase spectrum of the signal. If the phase spectrum is continuous and changes smoothly, it is considered that the stability of the signal at the corresponding frequency is higher than the preset stability threshold. The fast Fourier transform noise reduction algorithm is used to process the signal. If the signal clarity is improved after the noise level is reduced, it is determined that the noise reduction processing is effective. The attenuation characteristics of different frequency components are analyzed by fast Fourier transform. If the low-frequency component decays slower than the high-frequency component, it is judged that the propagation distance of the low-frequency signal is longer than the propagation distance of the high-frequency signal.

[0019] The preferred technical solution of the present invention is that, in S4, the adaptive filter is initialized by using the minimum mean square error algorithm, and the initial filter coefficients are set according to the statistical characteristics of the original ultrasonic signal.

[0020] The preferred technical solution of the present invention is that, in S5, an ultrasonic detector is used to receive the reflected echo signal to obtain original time-space data, which contains preliminary sound wave reflection information of the target area.

[0021] The preferred technical solution of the present invention is that, in S6, surface reconstruction technology is used to integrate data acquired from multiple angles, and the three-dimensional structure of the object being detected is reconstructed through the time delay and intensity difference of the ultrasonic signal to obtain a continuous and complete three-dimensional image.

[0022] The preferred technical solution of the present invention is that in S7, the key features include tissue boundaries, texture patterns and abnormal areas, the classification includes locating tumors, cysts or other structural abnormalities, and the key features are described according to their shape, size and position.

[0023] The preferred technical solution of the present invention is that, in S8, a feature extraction algorithm is used to analyze the located region of interest, extract morphological, texture, intensity and dynamic change features, determine key feature points in the image and mark abnormalities.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] The present invention provides a three-dimensional imaging and interaction method of an ultrasonic detector. In terms of technical solutions, by realizing automatic adjustment of the probe morphology to ensure the best fit with different body surfaces and anatomical structures, the imaging quality and diagnostic accuracy are significantly improved; by dynamically adjusting the probe pressure and ultrasonic parameters, the present invention can adapt to various depths and tissue types, capture clear multi-level and multi-angle images, and thus provide more comprehensive diagnostic information for clinicians. In addition, by using advanced image processing and reconstruction technology, the present invention can accurately identify and describe key features such as tissue boundaries and abnormal areas from the acquired images, further enhancing the reliability of diagnosis. By matching image features with a clinical knowledge base, the present invention can provide doctors with targeted diagnostic auxiliary information and greatly improve diagnostic efficiency. Overall, the present invention provides a more efficient, accurate and user-friendly ultrasonic detection solution for medical diagnosis, meets the needs of modern medicine for high-quality diagnostic tools, and has obvious intelligence. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0027] Figure 1 This is a flow chart of a three-dimensional imaging and interaction method of an ultrasonic detector according to Embodiment 1;

[0028] Figure 2 This is a flow chart of a method for obtaining multi-level and multi-angle ultrasonic signals according to Embodiment 1;

[0029] Figure 3 This is a flow chart of a method for identifying key features in an image and marking classification according to Embodiment 1. DETAILED DESCRIPTION

[0030] The present invention is further described below in conjunction with examples. These examples are only used to illustrate the present invention and are not intended to limit the scope of the present invention. The experimental methods in the following examples that do not specify specific conditions are usually carried out according to conventional conditions in the art or according to the conditions recommended by the manufacturer; the raw materials, reagents, etc. used, unless otherwise specified, are all raw materials and reagents that can be obtained from commercial channels such as conventional markets. Any non-substantial changes and substitutions made by those skilled in the art on the basis of the present invention belong to the scope of protection claimed by the present invention.

[0031] Embodiment 1

[0032] like Figure 1-3 This embodiment provides a three-dimensional imaging and interaction method of an ultrasonic detector, which specifically includes the following steps:

[0033] S0, according to the surface features of the object to be detected, uses a deformable ultrasonic probe driven by a deep learning algorithm. Through the flexibility of the probe material and the adjustment of the micromechanical structure, a probe shape that fits various curved surfaces is obtained.

[0034] The convolutional neural network is used to capture and analyze the surface of the object to be inspected in real time, and to identify the geometric features of the surface, including curvature, texture, and irregular shapes. The piezoelectric driver in the micromechanical structure is used to adjust the control parameters to drive the probe material to deform, adjust the shape of the probe, achieve precise fit with the surface to be inspected, and obtain a probe shape that fits different curved surfaces. The sensor integrated in the probe monitors the material deformation and ultrasonic propagation effect in real time, and obtains real-time data on the current probe shape and ultrasonic transmission efficiency. If the sensor data shows that the ultrasonic transmission efficiency is lower than expected, the control parameters of the piezoelectric driver are adjusted to optimize the probe shape to improve the ultrasonic transmission efficiency. The current shape of the probe, sensor data, and ultrasonic image information are displayed through the user interface to obtain the working status of the probe. If the data displayed on the user interface does not meet the preset detection standard, the convolutional neural network is used again to analyze the sensor data and determine the shape adjustment strategy. The probe performance and ultrasonic image quality are continuously monitored. If the fit between the probe shape and the object to be inspected is not optimal, the surface features are re-identified and analyzed according to the convolutional neural network. Through the closed-loop feedback mechanism, the surface of the object to be inspected is dynamically adapted to the changes, and the probe shape is continuously optimized until the best ultrasonic transmission effect is achieved.

[0035] For example, when detecting a lump at a depth of 2 cm on the surface of human skin, the initial shape of the probe is a flat shape, which is suitable for most flat surfaces. The target value of ultrasonic propagation efficiency is 95%. The skin surface is captured in real time using a convolutional neural network, which identifies the geometric features of the skin surface, including a slight bend with a curvature of 0.01 cm^-1, medium texture complexity, and some small irregular shapes. The piezoelectric driver in the micromechanical structure adjusts the probe shape according to the CNN analysis results to achieve precise fit with the skin. The probe is deformed from a flat shape to a slightly curved shape to fit the skin surface with a curvature of 0.01 cm^-1. The built-in sensor in the probe monitors the deformation of the material and the ultrasonic propagation effect in real time. The initial measurement shows that the ultrasonic propagation efficiency is 85%, which is lower than the target value of 95%. According to the sensor feedback, the control parameters of the piezoelectric driver are adjusted to further optimize the probe shape. After adjustment, the probe shape fits the skin better and the ultrasonic propagation efficiency is increased to 92%. The current shape of the probe, sensor data and ultrasonic image information are displayed through the user interface. After the probe shape is adjusted, the ultrasonic propagation efficiency is 92%, which still does not reach the target value of 95%. CNN analyzes the sensor data again and determines a new morphology adjustment strategy. Based on the CNN analysis, the probe shape is adjusted again, and the final ultrasonic transmission efficiency reaches the target value of 95%. The system continuously monitors the probe performance and ultrasonic image quality to ensure that the fit between the probe shape and the skin surface remains optimal. Through the closed-loop feedback mechanism, the probe shape dynamically adapts to the slight changes on the skin surface, maintaining the ultrasonic transmission efficiency above 95%, and successfully detecting the mass 2 cm below the skin.

[0036] S1, by real-time monitoring of the contact situation between the probe and the surface being tested, feedback information on the contact force and area is obtained, and the pressure applied by the probe is automatically adjusted using a closed-loop control method to determine the optimal probe pressure.

[0037] When changes in contact force and contact area are detected, an analog signal is obtained through the sensor, which includes feedback information on contact force and area and is converted into a digital signal. The converted digital signal is received by the controller, and the digital signal is analyzed according to the preset PID control algorithm to determine whether the pressure applied to the probe needs to be adjusted. If the digital signal analysis results show that the pressure needs to be adjusted, a corresponding instruction is issued to guide the increase or decrease of the pressure on the probe. The new contact situation after the pressure change is captured again by the sensor and converted into a digital signal by the data acquisition system to form a closed-loop feedback system. The adjusted contact data is re-analyzed to determine whether the optimal probe pressure setting has been achieved. If the optimal pressure has not been achieved, the output instruction is adjusted and the probe pressure is adjusted again. The real-time contact force, contact area and probe pressure adjustment status are displayed through the human-machine interface, and the option of adjusting the control parameters is provided, allowing the operator to manually intervene and optimize according to the probe performance and real-time data; according to the real-time monitoring data and feedback data, the control parameters are adjusted, and the long-term operation data of the probe is analyzed to identify potential performance improvement points, and the parameters of the control algorithm are automatically or manually adjusted to achieve optimal pressure control.

[0038] For example, the target contact force is 0.5N and the initial contact force is 0.2N. Because the contact between the probe and the skin is not complete, the PID control parameter proportional coefficient P = 0.1, integral coefficient I = 0.01, and differential coefficient D = 0.005. The initial contact force detected by the pressure sensor is 0.2N, the analog signal of the sensor is converted into a digital signal, and the controller receives the information that the contact force is 0.2N. The controller calculates the pressure adjustment amount through the PID algorithm. Since the actual contact force is lower than the target contact force, the pressure needs to be increased. Error = target contact force - actual contact force = 0.5N-0.2N = 0.3N, increased pressure = P error + I × error integral + D × error change rate. If the error integral in the current cycle is 0.03N and the error change rate is 0.1N / s, the increased pressure = 0.1×0.3+0.01×0.03+0.005×0.1=0.0335. Pressure adjustment Based on the results of the PID algorithm, the control system issues a command to increase the probe pressure by 0.0335N, and the new probe pressure becomes 0.2N+0.0335N=0.2335N. Feedback and re-analysis The new contact situation is captured by the sensor again and converted into a digital signal, and the controller receives the contact force information of 0.2335N. The PID algorithm is reapplied, and it is found that the contact force is still lower than the target value, so adjustment continues. Human-machine interface interaction The human-machine interface shows that the current contact force is 0.2335N, which has not reached the target contact force. The operator can choose to manually adjust the PID parameters or allow the system to continue to adjust automatically. Long-term performance optimization As the system continues to be used, the controller collects long-term operation data and analyzes performance improvement points, such as automatically adjusting the PID control parameters according to the changes in contact force in actual operation to achieve faster response and more stable control performance.

[0039] S2, based on the optimal probe pressure and surface feature information, uses multi-frequency ultrasonic emission technology to adjust the emission frequency to meet the detection needs of different depths and tissue types, and obtains multi-level and multi-angle ultrasonic signals.

[0040] Obtain information about the patient's specified tissue type and detection depth, including the acoustic impedance of the tissue, the expected detection depth, and the anatomical features of the target area. According to the detection requirements, if the requirement is high-resolution shallow imaging, select a high-frequency ultrasonic emission frequency. Use the acoustic simulation system to simulate the contact between the probe and the detection surface, and evaluate the coupling effect of the sound wave under specified conditions. If the simulation results show that the acoustic wave coupling effect is lower than the preset effect threshold, increase the probe pressure until the optimal coupling state is reached. According to the acoustic impedance characteristics of the tissue, adjust the beam focusing parameters of the ultrasound device, including changing the focusing depth and focusing range. If the tissue impedance is higher than the preset impedance threshold, change the focusing mode to enhance the penetration of the sound wave and the quality of the reflected signal. Calculate the average propagation time of the ultrasonic signal in different tissues, and adjust the pulse repetition frequency accordingly. If the propagation time is longer than the preset time threshold, reduce the pulse repetition frequency to avoid signal superposition. According to the size and shape of the target structure, calculate and adjust the optimal pulse width through the acoustic imaging algorithm to improve the resolution of fine structures. Implement real-time gain adjustment for echo signals at different depths to ensure balanced reception of the signal. For weak signal areas, increase the gain to improve signal quality. After obtaining the preliminary ultrasound image, the time gain compensation software module is used to adjust the uneven brightness of the image in the depth direction. If the brightness of the deep image is lower than the preset brightness threshold, the time gain compensation value is increased in the deep area. The dynamic range of the ultrasound device is adjusted. If the image shows oversaturated or dark areas, the dynamic range setting is optimized to improve the image quality. The acquired ultrasound signal is analyzed using an image processing algorithm. If there is noise in the image, the algorithm is applied to reduce the noise and improve the image clarity through edge enhancement.

[0041] For example, to obtain ultrasound imaging information of the patient's muscle tissue, the expected detection depth is 2 cm. It is known that the acoustic impedance of the muscle is about 7 MRayls. For high-resolution shallow imaging, a high-frequency ultrasound probe with a transmission frequency of 10 MHz is selected. The contact between the probe and the skin surface is simulated by the acoustic simulation system. The preliminary simulation shows that the acoustic wave coupling effect is not ideal when the probe pressure is 10 kPa. The probe pressure is gradually increased to 30 kPa until the coupling effect meets the preset effect threshold. According to the acoustic impedance characteristics of the muscle tissue, the sound beam focusing parameters of the ultrasound device are adjusted. If the focusing depth is set to 5 cm, the muscle tissue can be better focused. When the tissue impedance is detected to be higher than the preset impedance threshold of 8 MRayls, the focusing mode is switched from the standard mode to the deep focusing mode to enhance the penetration of the sound wave in the deep tissue and the quality of the reflected signal. In different tissues, the average propagation time of the ultrasonic signal is calculated to be 13 microseconds. Based on this, the pulse repetition frequency is adjusted to 5 kHz. When the propagation time is found to be longer than the preset time threshold of 15 microseconds, the repetition frequency is reduced to 4 kHz to avoid signal superposition. According to the size and shape of the target muscle fibers, the optimal pulse width was calculated to be 1 microsecond, which will help improve the resolution of the fine structure of the muscle fibers. During imaging, the echo signals at different depths were adjusted in real time. For the signals returned from a depth of 2 cm, the gain was adjusted to 20 dB to ensure balanced reception of the signal. In the weak signal area at a depth of 5 cm, the gain was adjusted to 25 dB to improve the signal quality. After obtaining the preliminary image, the brightness of the image in the depth direction was adjusted by the time gain compensation software module. If the brightness of the deep image was lower than the preset brightness threshold of 80%, the TGC value was increased to 95% in the deep area. The dynamic range of the ultrasound device was adjusted, and it was found that the image was too dark at a depth of 5 cm. The dynamic range was adjusted from 60 dB to 70 dB to improve the image quality. Finally, the acquired ultrasonic signal was analyzed by the image processing algorithm, and the noise level in the image was detected to be 5%. The noise reduction algorithm was applied to reduce the noise ratio to 2%, and the edge enhancement algorithm was used to improve the image clarity, and the visible edge contrast of the target muscle fiber was increased from the original 10% to 30%.

[0042] S3, based on the multi-frequency ultrasonic signals at different depths and angles, uses Fourier transform and wavelet transform to determine the characteristics of the multi-frequency ultrasonic signals in the time and frequency domains and evaluate the stability and clarity of each signal.

[0043] The amplitude spectrum of the multi-frequency ultrasonic signal is obtained by fast Fourier transform. If the amplitude spectrum shows that the signal has a high energy peak in the specified frequency band, it is judged that the signal has high energy concentration in this frequency band. The phase spectrum of the signal is analyzed by the fast Fourier transform algorithm. If the phase spectrum is continuous and changes smoothly, it is considered that the stability of the signal at the corresponding frequency is higher than the preset stability threshold. The continuous wavelet transform result of the signal is obtained. Through time-frequency resolution analysis, if the frequency component of the signal at the specified time point is clear, it is determined that the clarity of the signal at this time point is higher than the preset clarity threshold. The energy distribution diagram of the signal is obtained by discrete wavelet transform. If the energy distribution is uniform and has a local energy peak, it is judged that there is a mutation point or non-stationary feature in the signal. According to the discrete wavelet transform results, the time-frequency characteristics of the signal are extracted, including the instantaneous frequency and the local waveform. If the time-frequency characteristics show regular changes, it is considered that the characteristics of the signal in the dynamic change process are obvious. The signal is processed by the fast Fourier transform denoising algorithm. If the signal clarity is improved after the noise level is reduced, it is determined that the denoising process is effective. The attenuation characteristics of different frequency components are analyzed by fast Fourier transform. If the low-frequency component decays slower than the high-frequency component, it is judged that the propagation distance of the low-frequency signal is longer than that of the high-frequency signal. The group delay of the signal is calculated to evaluate the stability of the signal in the time domain. If the change amplitude of the group delay within a predetermined time is less than the preset amplitude threshold, it is determined that the signal has good time domain stability. The signal is compressed by the compression algorithm of wavelet transform. If the compressed signal still maintains the original characteristics, it is considered that the signal has good redundancy and compressibility.

[0044] For example, to detect the internal structure of the liver and look for possible abnormal tissue or tumors, the ultrasound frequency range is 1MHz to 5MHz, which is suitable for deep tissue imaging. The preset stability threshold is a phase change of less than 5 degrees / second, the preset clarity threshold is a frequency resolution of less than 0.1MHz in time-frequency analysis, and the preset amplitude threshold is a change amplitude of less than 0.5dB. The amplitude spectrum of the ultrasound signal is obtained using fast Fourier transform, and a high energy peak is found at 3MHz with an amplitude of -20dB, indicating that the signal has high energy concentration in the 3MHz frequency band. Continue to use the FFT algorithm to analyze the phase spectrum of the signal. In the 3MHz frequency band, the phase is continuous and changes smoothly, and the maximum change rate is 3 degrees / second, which is lower than the preset stability threshold, indicating that the signal has high stability in this frequency band. Perform continuous wavelet transform to analyze the time-frequency resolution of the signal. At the second second of the signal, the 3MHz frequency component is very clear, and the frequency resolution is 0.05MHz, which is higher than the preset clarity threshold, indicating that the signal has good clarity at this time point. The energy distribution diagram of the signal is obtained by applying discrete wavelet transform. The energy distribution is uniform, but there is an obvious local energy peak at the second second, indicating that the signal may have mutation or non-stationary characteristics at this point. According to the DWT results, the instantaneous frequency and local waveform of the signal are extracted. The instantaneous frequency and local waveform show regular changes at the second second, indicating that the signal has obvious characteristics in the dynamic change process. The signal is processed by FFT denoising algorithm. After denoising, the noise level of the signal is reduced from -30dB to -50dB, and the signal clarity is significantly improved. The attenuation characteristics of different frequency components are analyzed. The attenuation rate of the low-frequency component 1MHz is -0.2dB / cm, and the attenuation rate of the high-frequency component 5MHz is -1dB / cm, indicating that the propagation distance of the low-frequency signal is farther than that of the high-frequency signal. The group delay of the signal is calculated. During the observation period of 5 seconds, the group delay changes less than 0.2 seconds, which is less than the preset amplitude threshold, indicating that the signal has good time domain stability. The compression algorithm of wavelet transform is used to compress the signal data. The compressed signal retains all key features, and the compression ratio reaches 2:1, indicating that the signal has good redundancy and compressibility.

[0045] S4, based on the signal stability and clarity evaluation results, uses adaptive filtering and beamforming to perform directional enhancement and interference suppression on the original ultrasonic signal to obtain an optimized ultrasonic signal.

[0046] The adaptive filter is initialized by the minimum mean square error algorithm, and the initial filter coefficients are set according to the statistical characteristics of the original ultrasonic signal. If the filter convergence speed does not meet the real-time processing requirements, the step size parameter of the minimum mean square error algorithm is adjusted to increase the convergence speed in the iteration process to adapt to the dynamically changing signal environment. If the filter stability is insufficient, resulting in performance fluctuations, a regularization method is introduced or the weight factor is adjusted to improve the robustness of the filter when adjusting the coefficients. According to the tracking ability of the adaptive filter, if it cannot adapt to the changes in signal characteristics in time, the algorithm is adjusted, including introducing a variable step size or switching to a recursive least squares algorithm. The filtering effect is evaluated based on the error performance. If the mean square error is greater than the preset error threshold, the filter parameter configuration is optimized, including adjusting the filter length and step size to reduce the error. If the computational complexity is higher than the preset complexity threshold, affecting real-time performance, the affine projection algorithm is used or the filter order is reduced to reduce the computational burden. Through beamforming technology, the time delay of each element of the array is adjusted to form an optimized beam pattern in the target direction and enhance the ultrasonic signal in a specific direction. If the sidelobe level is too high and causes interference, the weight matrix in the beamforming algorithm is refined to suppress the sidelobe level, enhance the mainlobe signal, and reduce interference. Based on the results obtained from the spatial filtering capability, if the interference in the non-target direction is not completely suppressed, the beamforming weight matrix is ​​further adjusted to improve the selectivity of specific spatial signals and optimize the clarity and stability of the signal.

[0047] For example, ultrasonic imaging technology is used to detect the internal structure of the heart and find abnormal tissues. The ultrasonic frequency is 2.5MHz, which is suitable for deep tissue imaging. The minimum mean square error algorithm step size is 0.01. In order to ensure a good balance between convergence speed and stability, the preset error threshold is -30dB for evaluating the filtering effect. The preset complexity threshold is that the calculation time does not exceed 10ms to meet the requirements of real-time processing. The adaptive filter is initialized using the minimum mean square error algorithm, and the initial filter coefficients are set according to the statistical characteristics of the cardiac ultrasonic signal. After the first application, it was found that the filter converged slowly and failed to reach a stable state within the expected time. The step size parameter of the minimum mean square error algorithm was adjusted from 0.01 to 0.02 to speed up the convergence speed. It was observed that the filter performance fluctuated, and the regularization method was introduced, and the weight factor was adjusted to 0.001 to improve the robustness of the filter. The filter failed to adapt to the rapid changes in signal characteristics in time, and a variable step size mechanism was introduced, with the step size range set to 0.01 to 0.05 to improve the tracking effect. The measured mean square error was -28dB, which was higher than the preset error threshold. The filter length was increased from 32 to 64 and the step size was adjusted to 0.015 to reduce the error. The calculation time of the adjusted filter was 12ms, which exceeded the preset complexity threshold. The affine projection algorithm was used to replace the minimum mean square error algorithm to reduce the calculation time to 8ms. In order to enhance the signal in a specific area of ​​the heart, the time delay of the ultrasound array was adjusted to form an optimized beam pattern. The signal in the target area was enhanced and the sidelobe level was reduced from -25dB to -35dB. The weight matrix in the beamforming algorithm was further refined to suppress the sidelobe level. The mainlobe signal was further enhanced and the sidelobe level was reduced to -40dB. Interference in the non-target direction was not completely suppressed. The beamforming weight matrix was further adjusted to improve the selectivity of specific spatial signals. The clarity and stability of the signal were optimized, and abnormal tissue in the heart was successfully identified.

[0048] S5, through the optimized ultrasonic signal, adopts inverse signal processing and time-frequency analysis to obtain the geometric and material characteristics of the target area, and adjusts the ultrasonic emission parameters, including the focal depth, focal position and the size of the focal area, to achieve dynamic focusing.

[0049] An ultrasonic detector is used to receive the reflected echo signal and obtain the original time-space data, which contains the preliminary acoustic reflection information of the target area. If the acoustic signal quality is poor, inverse signal processing technology, including denoising and signal enhancement, is used to improve the readability and reliability of the signal. Based on the improved acoustic signal, the geometric position and boundaries of the target area are determined by wavelet transform to obtain preliminary focal depth and focal position information. Computer simulation is used to simulate the ultrasonic propagation path and reflection characteristics under different focal depths and focal positions, determine the optimal focal area size, and obtain the clearest imaging effect. According to the simulation results, the ultrasonic emission parameters, including focal depth, position, and focal area size, are adjusted to achieve precise dynamic focusing on the target area. According to the geometric shape and material characteristics complexity of the target area, the sound beam shape and waveform are further adjusted to adapt to the specific reflection and absorption characteristics of the target area. By adjusting the emission frequency and pulse length, the resolution and penetration depth are balanced to ensure that the image clarity is maintained while detecting the target area. According to the characteristics and motion characteristics of the target material, the pulse repetition frequency and sound pressure settings are optimized to capture dynamically changing signals and reduce motion artifacts. When receiving signals, dynamic receiving and focusing technology is applied to perform real-time focusing processing on the reflected signals of sound waves at different depths to improve the spatial resolution and contrast of imaging.

[0050] For example, when using an ultrasonic detector to receive a reflected echo signal, suppose the raw data shows a strong echo peak at 3 microseconds, which may indicate that there is a reflective interface about 25 mm from the probe. The average speed of sound in the medium, i.e., 5000 m / s, is used in the calculation process. If the acoustic signal is interfered by background noise, the signal-to-noise ratio may be only 5:1. By using inverse signal processing techniques, such as Wiener filtering, the signal-to-noise ratio can be increased to 10:1, making the echo signal clearer and easier to identify. In the application of wavelet transform, it may be found that the boundary of the target area is most obvious at the decomposition level of wavelet scale 2. If the mean of the wavelet coefficients corresponding to this scale is 8, and the mean of the coefficients of the noise level is 1, it can be considered that this is a significant feature of the target boundary. In computer simulation, in order to determine the optimal focal area size, it may be found that when the focal depth is 30 mm, the ultrasonic beam width of 2 mm can obtain the highest reflection signal intensity, thereby providing the clearest imaging effect. When adjusting the ultrasonic emission parameters, assuming that the final selected focal depth is 30 mm, the size of the focal area is adjusted to a width of 2 mm to ensure that the reflected signal in the target area is maximized. In view of the complexity of the target area, it may be necessary to adjust the beam shape to an ellipse and adjust the waveform to adapt to the specific reflection characteristics. For example, assuming that the target object has a strong absorption of sound waves with a frequency of 5 MHz, adjusting the emission frequency to 2 MHz may reduce energy loss. When balancing resolution and penetration depth, it is found that the combination of a transmission frequency of 2 MHz and a pulse length of 5 microseconds can provide sufficient penetration depth while maintaining good resolution. If the target material moves at a speed of 1 mm per second, optimizing the pulse repetition frequency to 1 kHz and setting the sound pressure to 1 MPa can effectively capture this dynamic change while reducing motion artifacts caused by high-speed movement. When receiving signals, the dynamic receiving focusing technology can automatically adjust the depth of the receiving focus according to the propagation time of the sound wave, so as to achieve the focusing of the reflected signal at a depth of about 35 mm with a propagation time of 5 microseconds, and correspondingly improve the spatial resolution and contrast of the image.

[0051] S6, based on dynamic focusing parameters and optimized ultrasonic signals, uses surface reconstruction technology to integrate data acquired from multiple angles, reconstructs the three-dimensional structure of the object being tested through the time delay and intensity difference of the ultrasonic signal, and obtains a continuous and complete three-dimensional image.

[0052] According to the type of object to be inspected and the expected imaging depth, the ultrasonic frequency is determined to balance the imaging resolution and penetration depth. The pulse repetition frequency matching the ultrasonic frequency is set to control the imaging speed and time resolution to meet the requirements of dynamic focusing technology. The focus position of the probe array is gradually adjusted to optimize the structures at different depths to obtain imaging results with clarity greater than the preset second clarity threshold. The time delay of the ultrasonic wave in different media is measured, and the actual path of the ultrasonic wave in the medium is calculated in combination with the known sound speed to provide the data required for image reconstruction. The intensity difference data of the ultrasonic wave in different media is obtained, the attenuation coefficient of the medium is analyzed, and the contrast and details in the imaging are corrected. The object to be inspected is scanned from different directions, and the data from different angles are integrated to provide comprehensive information for three-dimensional reconstruction. The image reconstruction algorithm, including back-projection, iterative reconstruction or model-based reconstruction algorithm, is used to integrate the time delay and intensity difference data to generate the preliminary three-dimensional structure of the object to be inspected. The preliminary three-dimensional image is post-processed, including denoising, image enhancement and edge detection operations, to improve the image clarity and quality. The acoustic impedance matching technology is applied to ensure the effective transmission of ultrasonic energy, and a continuous and complete three-dimensional image is obtained through the final optimization process.

[0053] For example, to detect the structure of knee cartilage and identify possible damage or degeneration, the expected imaging depth is about 2mm to 5mm in cartilage thickness, but deeper tissue needs to be penetrated to obtain a complete knee joint image, so the imaging depth is set to 0 to 10cm, and the ultrasonic frequency is selected to be 7.5MHz, which is a frequency commonly used for shallow tissue imaging and can provide sufficient resolution and moderate penetration depth. The pulse repetition frequency is set to 2kHz, which can meet the needs of dynamic focusing and ensure sufficient imaging speed and time resolution. The focus position of the probe array is gradually adjusted to focus from the epidermis to the deeper cartilage and other knee joint structures. Through dynamic adjustment, the obtained cartilage imaging clarity reaches a signal-to-noise ratio of -15dB, which is higher than the preset second clarity threshold of -20dB. The time delay of ultrasound is measured at different depths. It is known that the speed of sound in human tissue is about 1540m / s. In the cartilage area at a depth of 5cm, the time delay of ultrasound is about 6.49μs. The intensity difference of ultrasound in normal and damaged cartilage is recorded, and it is found that the ultrasound attenuation coefficient in the damaged cartilage area is significantly higher than that in the normal cartilage, providing contrast information of the damaged area. Ultrasonic scanning of the cartilage was performed from the front, side, and back of the knee, and a comprehensive ultrasound data set covering the entire knee cartilage was obtained. An iterative reconstruction algorithm was used to integrate the time delay and intensity difference data obtained from different angles to generate a preliminary three-dimensional structure of the cartilage, and a preliminary three-dimensional image of the knee cartilage and its surrounding structures was produced, which clearly showed the thickness and damaged area of ​​the cartilage. The preliminary three-dimensional image was post-processed by denoising, image enhancement, and edge detection to improve the clarity and quality of the image, making the edge of the cartilage damage area more obvious. Acoustic impedance matching technology was applied to ensure the effective transmission of ultrasonic energy in the cartilage tissue, and the final optimization processing was performed. After optimization, a continuous and complete three-dimensional image of the knee cartilage was obtained, which clearly showed the contrast between the normal and damaged areas.

[0054] S7, based on the generated continuous three-dimensional images, identify key features in the images, including tissue boundaries, texture patterns, and abnormal areas, and mark and classify the key features, including locating tumors, cysts, or other structural abnormalities, and describing the key features according to their shapes, sizes, and locations.

[0055] Preprocess the continuous three-dimensional image data, including noise reduction and contrast adjustment. Use edge detection algorithms, including Sobel or Canny operators, to identify the boundaries of tissues in the image and mark them as potential key feature boundaries. Apply gray-level co-occurrence matrix to analyze the texture pattern in the image. If there is an area with a texture difference greater than a preset difference threshold from the surrounding tissue, it is judged as an abnormal area. Combined with the Otsu algorithm, the abnormal area is segmented from the background tissue to obtain the rough shape and position of the key feature. Morphological operations, including corrosion and dilation, are performed on the segmented features to morphologically refine the segmented features and determine the boundaries of the key features. Volume rendering technology is used to calculate the three-dimensional volume and maximum diameter of the segmented features to determine the size of the key features. The random forest algorithm is applied to classify the features based on the multi-dimensional information of the shape, size, texture and boundary of the features to identify tumors, cysts or other structural abnormalities. The absorption pattern of the contrast agent of the key features is determined by time-intensity curve analysis to provide additional information about their biological properties. The marking and classification of key features are verified and supplemented by combining CT and MRI image information.

[0056] For example, when processing a batch of three-dimensional image data of magnetic resonance imaging, the data is a continuous thin-layer scan with a resolution of 256x256 pixels and a layer spacing of 1mm for each layer. In the preprocessing stage, a Gaussian filter can be used to reduce the noise of the image, setting the standard deviation to 0, and then adjusting the contrast through histogram equalization to improve the visual difference between tissues. Next, the edges are detected by applying the Sobel operator, and when the edge strength exceeds 40, it is regarded as a potential key feature boundary. In terms of texture analysis, the gray-level co-occurrence matrix is ​​used to calculate texture features such as contrast or homogeneity, and the difference threshold is set to 5 to identify abnormal areas with significant differences. Using the Otsu algorithm to segment the abnormal area with high contrast with the background tissue, it may be found that a volume of 15cm 3 , an abnormal area with a maximum diameter of 3 cm. The feature is eroded and expanded, possibly using a structural element with a radius of 1 mm, followed by expansion using a structural element of the same size to refine the boundary. In the classification stage, the random forest algorithm may classify the abnormal area as a malignant tumor with an accuracy of 85% based on multi-dimensional features including shape, size, and texture. Time-intensity curve analysis showed that the area rapidly enhanced within the first 2 minutes after contrast injection, followed by gradual washout to baseline levels within the next 5 minutes, suggesting that it may be a highly vascular tumor. Finally, by combining multimodal image information, such as superposition analysis of CT and MRI, this finding was confirmed and the tumor was determined to be located in the left lobe of the liver.

[0057] S8, based on the key features and abnormalities in the marked three-dimensional image, the features in the image are matched with the known pathological features in the clinical knowledge base through feature matching method to obtain targeted diagnostic auxiliary information.

[0058] High-resolution CT scanning equipment is used to obtain three-dimensional image data of patients by setting specified acquisition parameters and using contrast agents. The acquired images are preprocessed, including noise reduction and correction. Segmentation technology is used to process the preprocessed images, determine the region of interest (ROI), calibrate the location of key features and abnormal areas, and locate the feature extraction. The region of interest (ROI) is the location of the tumor. The located region of interest is analyzed using a feature extraction algorithm to extract morphological, texture, intensity and dynamic change features, determine key feature points in the image and mark abnormalities. Relevant pathological classifications and known pathological features are retrieved through the clinical knowledge base to obtain clinical information that matches the marked feature attributes, providing reference data for the feature matching algorithm. A feature matching algorithm based on the support vector machine algorithm is used to compare the similarity between the extracted image features and the pathological features in the knowledge base, and the Mahalanobis distance measurement method is used to determine the best match. The results generated by the matching algorithm are analyzed for fault tolerance and matching accuracy, and the algorithm parameters are adjusted to optimize the matching effect to obtain high-precision diagnostic auxiliary information. According to the feature matching results, the confidence is evaluated and combined with the patient's clinical symptoms to generate standardized diagnostic auxiliary information to provide practical decision support for doctors. A user interface with ease of use and strong interactivity is designed to allow doctors to view and manually adjust the matching results, and optimize the feature matching algorithm and the accuracy of the output results through interface feedback.

[0059] For example, CT scans are used to detect brain tumors and identify their types. A high-resolution scanning mode of 512x512 pixels is used. Iodine-based contrast agents are used to enhance the contrast between the tumor and surrounding tissues. In order to effectively remove noise, the signal-to-noise ratio is set to 20dB. A CT scan of the patient's brain is performed, and iodine-based contrast agents are used to obtain detailed three-dimensional image data of the patient's brain. The acquired CT images are subjected to noise reduction and correction processing. Gaussian filters are used for noise reduction, and geometric correction technology is used to correct image distortion. The signal-to-noise ratio of the image is increased to 25dB, exceeding the preset threshold. Using threshold-based image segmentation technology, the region of interest ROI, that is, the tumor location, is determined, and the tumor area is successfully calibrated, with a size of 1.5cmx1.2cm. The feature extraction algorithm is used to analyze the tumor area, extract morphological, texture, intensity and dynamic change features, and identify the irregular edges, heterogeneous texture features and enhanced contrast of the tumor. Through the clinical knowledge base, known pathological features that match the extracted features are retrieved to obtain clinical information on multiple types of brain tumors similar to tumor features. Using the support vector machine algorithm, the image features are compared with the pathological features in the knowledge base based on the Mahalanobis distance, and the best match is determined to be a glioma, with a matching accuracy of 95%. The fault tolerance and accuracy of the matching results are analyzed, and the algorithm parameters are adjusted. By fine-tuning the kernel function parameters of the support vector machine algorithm, the matching accuracy is increased to 98%. Combined with the feature matching results, confidence assessment and clinical symptoms, diagnostic auxiliary information is generated. The generated diagnostic report indicates a high-confidence glioma diagnosis and provides the specific location, size and possible type of the tumor. The user interface is designed to allow doctors to view the matching results and make manual adjustments. The interface provides intuitive image display, result interpretation and adjustment options, allowing doctors to verify and fine-tune the results according to their own clinical experience. Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit the scope of protection of the present invention. Although the present invention is described in detail with reference to the preferred embodiment, ordinary technicians in the field should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the essence and scope of the technical solution of the present invention.

Claims

1. A three-dimensional imaging and interaction method of an ultrasonic detector, characterized in that: The following steps are involved: S0, based on the surface features of the object being tested, a deformable ultrasonic probe driven by a deep learning algorithm is used to obtain a probe shape that fits various curved surfaces; S1, by real-time monitoring of the contact between the probe and the detected surface, feedback information of the contact force and area is obtained to determine the probe pressure; S2, based on the probe pressure and surface feature information, uses multi-frequency ultrasonic emission technology to adjust the emission frequency to meet the detection needs of different depths and tissue types, and obtains multi-level and multi-angle ultrasonic signals; S3, based on the multi-frequency ultrasonic signals at different depths and angles, the characteristics of the multi-frequency ultrasonic signals in the time and frequency domains are determined using Fourier transform and wavelet transform, and the stability and clarity of each signal are evaluated; S4, based on the signal stability and clarity evaluation results, adaptive filtering and beamforming are used to perform directional enhancement and interference suppression on the original ultrasonic signal to obtain an optimized ultrasonic signal; S5, using the optimized ultrasonic signal, inverse signal processing and time-frequency analysis to obtain the geometric and material characteristics of the target area, and adjust the ultrasonic emission parameters, including the focal depth, focal position and size of the focal area, to achieve dynamic focusing; S6, reconstructing the three-dimensional structure of the detected object according to the dynamic focusing parameters and the optimized ultrasonic signal to obtain a continuous and complete three-dimensional image; S7, identifying key features in the image according to the generated continuous three-dimensional images, and marking and classifying the key features; S8, according to the key features in the marked three-dimensional image, by using a feature matching method, matching the features in the image with known pathological features in a clinical knowledge base to obtain targeted diagnostic auxiliary information; In S0, a convolutional neural network is used to capture and analyze real-time images of the surface of the object being inspected, and to identify the geometric features of the surface, including curvature, texture, and irregular shapes. The piezoelectric driver in the micromechanical structure is used to adjust the control parameters to drive the probe material to deform, and the shape of the probe is adjusted to achieve precise fit with the surface being inspected, thereby obtaining a probe shape that fits different curved surfaces.

2. The three-dimensional imaging and interaction method of the ultrasonic detector according to claim 1, characterized in that: In S1, when changes in contact force and contact area are detected, an analog signal is acquired through the sensor, the analog signal includes feedback information of the contact force and area, and is converted into a digital signal; the converted digital signal is received by the controller, and the digital signal is analyzed according to a preset PID control algorithm to determine whether the pressure applied to the probe needs to be adjusted.

3. The three-dimensional imaging and interaction method of the ultrasonic detector according to claim 1, characterized in that: In S2, information about the patient's specified tissue type and detection depth is obtained, including the acoustic impedance of the tissue, the expected detection depth, and the anatomical characteristics of the target area; according to the detection requirements, if the requirement is high-resolution shallow imaging, a high-frequency ultrasonic transmission frequency is selected; The contact between the probe and the detection surface is simulated by the acoustic simulation system to evaluate the coupling effect of the sound wave under specified conditions. If the simulation results show that the acoustic wave coupling effect is lower than the preset effect threshold, the probe pressure is increased until the optimal coupling state is reached.

4. The three-dimensional imaging and interaction method of the ultrasonic detector according to claim 1, characterized in that: In S3, an amplitude spectrum of the multi-frequency ultrasonic signal is obtained by fast Fourier transform, and if the amplitude spectrum shows that the signal has a high energy peak in the specified frequency band, it is determined that the signal has high energy concentration in the frequency band; The phase spectrum of the signal is analyzed using the fast Fourier transform algorithm. If the phase spectrum is continuous and changes smoothly, it is considered that the stability of the signal at the corresponding frequency is higher than the preset stability threshold. The signal is processed using the fast Fourier transform noise reduction algorithm. If the signal clarity is improved after the noise level is reduced, it is determined that the noise reduction processing is effective. The attenuation characteristics of different frequency components are analyzed by fast Fourier transform. If the low-frequency component decays slower than the high-frequency component, it is judged that the propagation distance of the low-frequency signal is longer than the propagation distance of the high-frequency signal.

5. The three-dimensional imaging and interaction method of the ultrasonic detector according to claim 1, characterized in that: In S4, the adaptive filter is initialized using the minimum mean square error algorithm, and the initial filter coefficients are set according to the statistical characteristics of the original ultrasonic signal.

6. The three-dimensional imaging and interaction method of the ultrasonic detector according to claim 1, characterized in that: In S5, an ultrasonic detector is used to receive the reflected echo signal to obtain original time-space data, which contains preliminary sound wave reflection information of the target area.

7. The three-dimensional imaging and interaction method of the ultrasonic detector according to claim 1, characterized in that: In S6, surface reconstruction technology is used to integrate data acquired from multiple angles. The three-dimensional structure of the object being tested is reconstructed through the time delay and intensity difference of the ultrasonic signal to obtain a continuous and complete three-dimensional image.

8. The three-dimensional imaging and interaction method of the ultrasonic detector according to claim 1, characterized in that: In S7, the key features include tissue boundaries, texture patterns, and abnormal areas, and the classification includes locating tumors, cysts, or other structural abnormalities, and describing the key features based on their shape, size, and location.

9. The three-dimensional imaging and interaction method of the ultrasonic detector according to claim 1, characterized in that: In S8, feature extraction algorithms are used to analyze the located regions of interest, extract morphological, texture, intensity and dynamic change features, determine key feature points in the image and mark anomalies.

Images