AI model training method and effective radiation area measuring method and system

Abstract

The invention provides a training method and device of an AI model and a measuring method and system of an effective radiation area of a treatment head of ultrasonic physiotherapy equipment. When the AI model is trained, the treatment head outputs ultrasonic waves, the ultrasonic waves are transmitted to the surface of the sound absorption medium, a thermal imaging video is recorded through the thermal infrared imager, and the thermal imaging video comprises a dynamic process that the temperature of the surface of the sound absorption medium rises along with time when the treatment head of the ultrasonic physiotherapy equipment outputs the ultrasonic waves; extracting the instantaneous temperature change rate of each frame in the thermal imaging video; generating a temperature change rate heat map sequence according to the instantaneous temperature change rate of each frame, and using the temperature change rate heat map sequence as input data in a training sample; and using the boundary mask of the effective radiation area of the treatment head as label data in the training sample. According to the invention, the effective radiation area of the treatment head of the ultrasonic physiotherapy equipment can be measured more efficiently and more intelligently, and manual operation is greatly reduced.

Description

Technical Field

[0001] One or more embodiments of the present invention relate to network communication technologies, and in particular, to a method and device for training an AI (Artificial Intelligence) model, and a method and system for measuring the effective radiation area of a treatment head of an ultrasonic physiotherapy device. Background Art

[0002] An ultrasonic physiotherapy device is a medical instrument that uses the physical properties of ultrasonic waves (usually in the frequency range of 1 - 3 MHz) to treat diseases. The applications of ultrasonic physiotherapy devices include: generating micro heat in tissues using ultrasonic waves to promote local blood circulation, relieve pain and muscle spasms; producing a micro massage effect through high-frequency vibration to loosen adhesions and promote the dissipation of inflammation; enhancing cell membrane permeability through micro bubble oscillation to accelerate metabolism and repair. Therefore, in clinical applications, ultrasonic physiotherapy devices can be used for, for example, pain management, such as the treatment of arthritis, tendinitis, soft tissue injuries, etc., and can also be used for tissue repair to accelerate wound healing and recovery after fractures.

[0003] A core component of an ultrasonic physiotherapy device is a treatment head, which converts electrical energy into high-frequency mechanical vibrations (ultrasonic waves) and transmits them through the skin to human tissues to treat patients.

[0004] When using an ultrasonic physiotherapy device for treatment, the treatment head usually contacts the human body. To avoid burning the patient, it is necessary to measure the effective radiation area of the treatment head. However, the treatment head in an ultrasonic physiotherapy device usually consists of an ultrasonic transducer, and the ultrasonic transducer consists of a piezoelectric sensitive element equipped with a metal end face. The effective radiation area of the treatment head refers to the area where the ultrasonic transducer actually generates an effective energy distribution in the medium. Since the entire surface of the piezoelectric element in the treatment head does not vibrate with the same amplitude, the effective radiation area of the ultrasonic waves emitted by the treatment head is not equal to the geometric projection area of the piezoelectric element, that is, the effective radiation area of the treatment head may be larger or smaller than the geometric projection area. Therefore, an efficient method is needed to measure the effective radiation area of the treatment head of an ultrasonic physiotherapy device. Summary of the Invention

[0005] One or more embodiments of the present invention describe a method and device for training an AI model, and a method and system for measuring the effective radiation area of a treatment head of an ultrasonic physiotherapy device, which can measure the effective radiation area of the treatment head of an ultrasonic physiotherapy device more efficiently and intelligently, and greatly reduce manual operations.

[0006] According to a first aspect, there is provided a method for training an AI model, where the AI model is used to measure the effective radiation area of a treatment head of an ultrasonic physiotherapy device, and the training method includes:

[0007] Generate multiple sets of training samples; wherein, each set of training samples includes input data and the label data corresponding to the input data;

[0008] Input each set of training samples into the AI model to be trained to train the AI model;

[0009] Among them, the method for generating training samples includes:

[0010] Orient the treatment head of the ultrasonic physiotherapy device towards the surface of the sound-absorbing medium, and the treatment head is at a preset distance from the surface of the sound-absorbing medium;

[0011] Adjust the focal length of the infrared thermal imager so that the focal length of the infrared thermal imager is aligned with the radiation area of the treatment head of the ultrasonic physiotherapy device;

[0012] Start the treatment head, and the treatment head outputs ultrasonic waves, and the ultrasonic waves are transmitted to the surface of the sound-absorbing medium;

[0013] Record a thermal imaging video through the infrared thermal imager, and the thermal imaging video includes the dynamic process of the temperature on the surface of the sound-absorbing medium increasing with time when the treatment head of the ultrasonic physiotherapy device outputs ultrasonic waves;

[0014] Extract the instantaneous temperature change rate of each frame in the thermal imaging video;

[0015] Generate a sequence of temperature change rate heat maps according to the instantaneous temperature change rate of each frame, and use the sequence of temperature change rate heat maps as the input data in the training samples; and

[0016] Use the boundary mask of the effective radiation area of the treatment head as the label data in the training samples.

[0017] The preset distance between the treatment head and the surface of the sound-absorbing medium is 0.3 cm;

[0018] and / or

[0019] The sound-absorbing medium is a tissue-mimicking material with the difference between the acoustic impedance and the acoustic impedance of human tissue less than a predetermined value;

[0020] and / or

[0021] The sound-absorbing medium includes a tissue-mimicking hydrogel with the difference between the acoustic impedance and the acoustic impedance of human tissue less than a predetermined value;

[0022] and / or

[0023] The treatment head outputs short-time ultrasonic pulses to reduce the influence of heat diffusion.

[0024] During the generation process of different sets of training samples:

[0025] The types of treatment heads used are different; and / or

[0026] The types of infrared thermal imagers used are different; and / or

[0027] The powers used when the treatment head generates ultrasonic waves are different; and / or

[0028] The types of sound-absorbing media are different to increase the diversity of training samples.

[0029] After recording the thermal imaging video by the infrared thermal imager and before extracting the instantaneous temperature change rate of each frame in the thermal imaging video, it further includes: for the thermal imaging video, segmenting the thermal imaging video according to a preset time window and sliding step length to obtain at least two frame sequences;

[0030] Then, extracting the instantaneous temperature change rate of each frame in the thermal imaging video includes: for each frame sequence, extracting the instantaneous temperature change rate of each frame in the frame sequence;

[0031] Then, generating a sequence of temperature change rate heat maps according to the instantaneous temperature change rate of each frame includes: generating a sequence of temperature change rate heat maps for the instantaneous temperature change rate of each frame in each frame sequence;

[0032] Then, using the sequence of temperature change rate heat maps as the input data in the training samples includes: using each sequence of temperature change rate heat maps generated for each frame sequence as the input data in a separate set of training samples, so as to generate at least two sets of input data in the training samples for at least two frame sequences of a thermal imaging video;

[0033] Then, using the boundary mask of the effective radiation area of the treatment head of the ultrasonic physiotherapy device as the label data in the training samples includes: using the boundary mask of the effective radiation area of the treatment head used when forming the thermal imaging video as the label data in the at least two sets of training samples.

[0034] The input data in each set of training samples further includes: the calibration parameters of the infrared thermal imager, and this calibration coefficient indicates the number of pixels corresponding to each unit area when using this infrared thermal imager;

[0035] The label data in each set of training samples further includes: the effective radiation area of the treatment head of the ultrasonic physiotherapy device.

[0036] According to a second aspect, there is provided a method for measuring the effective radiation area of the treatment head of an ultrasonic physiotherapy device, and this method includes:

[0037] Orient the first treatment head of the ultrasonic physiotherapy device to be measured towards the surface of the sound-absorbing medium, and the first treatment head is at a preset distance from the surface of the sound-absorbing medium; the difference between the acoustic impedance of the sound-absorbing medium and the acoustic impedance of human tissue is less than a predetermined value;

[0038] Adjust the focal length of the infrared thermal imager so that the focal length of the first infrared thermal imager is aligned with the radiation area of the first treatment head of the ultrasonic physiotherapy device;

[0039] Start the first treatment head, and output ultrasonic waves from the first treatment head. The ultrasonic waves are transmitted to the surface of the sound-absorbing medium;

[0040] Record a thermal imaging video through the first infrared thermal imager. The thermal imaging video includes the dynamic process of the temperature rising with time on the surface of the sound-absorbing medium when the first treatment head of the ultrasonic physiotherapy device outputs ultrasonic waves;

[0041] Extract the instantaneous temperature change rate of each frame in the thermal imaging video;

[0042] Generate a sequence of thermal maps of temperature change rates according to the instantaneous temperature change rate of each frame;

[0043] Input the sequence of thermal maps of temperature change rates into the AI model, where the AI model is obtained by using the training method in the embodiments of the present invention;

[0044] Obtain the boundary mask of the effective radiation area of the first treatment head output by the AI model;

[0045] Multiply the number of pixels corresponding to the boundary mask of the effective radiation area of the first treatment head by the first calibration parameter of the first infrared thermal imager, so as to obtain the effective radiation area of the first treatment head of the ultrasonic physiotherapy device to be measured; wherein, the first calibration coefficient indicates the number of pixels corresponding to each unit area when using the first infrared thermal imager.

[0046] The AI model performs the step of "multiplying the number of pixels corresponding to the boundary mask of the effective radiation area of the first treatment head by the first calibration parameter of the first infrared thermal imager, so as to obtain the effective radiation area of the first treatment head of the ultrasonic physiotherapy device to be measured";

[0047] The method further includes: the AI model outputs the effective radiation area of the first treatment head of the ultrasonic physiotherapy device to be measured.

[0048] The treatment head outputs short-time ultrasonic pulses.

[0049] According to the third aspect, there is provided a training device for an AI model. The AI model is used to measure the effective radiation area of the treatment head of an ultrasonic physiotherapy device. The training device includes:

[0050] A training sample generation module, configured to generate multiple groups of training samples; wherein, each group of training samples includes input data and the label data corresponding to the input data;

[0051] A training execution module, configured to input each group of training samples into the AI model to be trained for training the AI model;

[0052] Among them, the training sample generation module is used to execute:

[0053] Receive the thermal imaging video recorded by the infrared thermal imager, where the thermal imaging video includes the dynamic process of the temperature on the surface of the sound absorption medium increasing with time when the treatment head of the ultrasonic physiotherapy device outputs ultrasonic waves;

[0054] Extract the instantaneous temperature change rate of each frame in the thermal imaging video;

[0055] Generate a sequence of temperature change rate heat maps according to the instantaneous temperature change rate of each frame, and use the sequence of temperature change rate heat maps as the input data in the training samples; and

[0056] Use the boundary mask of the effective radiation area of the treatment head as the label data in the training samples.

[0057] According to the fourth aspect, a measurement system for the effective radiation area of the treatment head of an ultrasonic physiotherapy device is provided. The system includes: an infrared thermal imager, a computer processing module, an AI model, and a calculation module;

[0058] The first treatment head of the ultrasonic physiotherapy device to be measured faces the surface of the sound absorption medium, and the first treatment head is at a preset distance from the surface of the sound absorption medium; the difference between the acoustic impedance of the sound absorption medium and the acoustic impedance of human tissue is less than a predetermined value; after the first treatment head is activated, it outputs ultrasonic waves, and the ultrasonic waves are transmitted to the surface of the sound absorption medium;

[0059] The focal length of the infrared thermal imager is adjusted to be aligned with the radiation area of the first treatment head of the ultrasonic physiotherapy device; the infrared thermal imager is used to record a thermal imaging video after the first treatment head is activated, and the thermal imaging video includes the dynamic process of the temperature on the surface of the sound absorption medium increasing with time when the first treatment head of the ultrasonic physiotherapy device outputs ultrasonic waves;

[0060] A computer processing module, configured to extract the instantaneous temperature change rate of each frame in the thermal imaging video; generate a sequence of temperature change rate heat maps according to the instantaneous temperature change rate of each frame; input the sequence of temperature change rate heat maps into the AI model; among them, the AI model is trained by using the training device in the embodiments of the present invention;

[0061] The AI model is used to output the boundary mask of the effective radiation area of the first treatment head of the ultrasonic physiotherapy device;

[0062] A calculation module, configured to multiply the number of pixels corresponding to the boundary mask of the effective radiation area of the first treatment head by the first calibration parameter of the first infrared thermal imager to obtain the effective radiation area of the first treatment head of the ultrasonic physiotherapy device to be measured; wherein, the first calibration coefficient indicates the number of pixels corresponding to each unit area when using the first infrared thermal imager.

[0063] Thus, each embodiment of the present invention has at least the following beneficial effects:

[0064] 1. In the embodiments of the present invention, an AI model is used to learn how to determine the temperature gradient boundary, that is, to extract the boundary mask of the effective radiation area from a sequence of thermal maps of the temperature change rate (i.e., a kind of dynamic thermal gradient data). Thus, in the subsequent actual measurement process, the information of the effective radiation area output by the AI model can be used to replace the information of the effective radiation area obtained by the time-consuming, laborious and complex manual interpretation or hydrophone point-by-point scanning in the prior art. Therefore, the intelligence and efficiency are greatly improved. For example, when using a hydrophone in the prior art, it is necessary to scan the sound field point by point, which takes a long time, depends on manual operation and is prone to errors. In the embodiments of the present invention, a single thermal imaging video can capture the temperature change rate distribution of the entire sound field area, and the dynamic gradient boundary, that is, the boundary mask of the effective radiation area, can be directly identified by training the AI model, without manual interpretation, which is efficient and intelligent.

[0065] 2. In the embodiments of the present invention, a dynamic thermal gradient boundary is used. The dynamic thermal gradient boundary refers to determining the boundary of the ultrasonic energy distribution in the medium by analyzing the rate of change of temperature over time (dT / dt) rather than the static temperature value. The dynamic thermal gradient boundary in the embodiments of the present invention has dynamics, that is, it focuses on the instantaneous effect of temperature change rather than the thermal equilibrium in the steady state. Moreover, the dynamic thermal gradient boundary in the embodiments of the present invention has gradient sensitivity, that is, it distinguishes the effective sound field area from the non-effective sound field area by the speed of temperature change (the steepness of the gradient). Therefore, the sound intensity of the ultrasonic wave can be better corresponding by means of temperature.

[0066] 3. The embodiments of the present invention better eliminate the influence of thermal diffusion. Considering that heat diffuses over time and with the thermal conductivity of the medium, resulting in the thermal imaging boundary (i.e., temperature distribution) being larger than the sound field boundary, which makes it impossible to accurately reflect the sound field boundary. To solve this problem, instead of using static temperature values to determine the boundary in the embodiments of the present invention, a sequence of thermal maps of the temperature change rate (i.e., a kind of dynamic thermal gradient data) is used to replace the conventional static temperature values, so as to capture the instantaneous effect of sound energy input. The region with the maximum temperature change rate is closer in time to the instantaneous distribution of sound energy. Specifically, when ultrasonic waves act on the medium, the temperature in the sound energy input region rises rapidly (manifested as a high temperature change rate dT / dt), while in the surrounding regions, due to the attenuation of sound intensity and thermal diffusion, the temperature changes slowly (manifested as a low temperature change rate dT / dt). On the other hand, in the embodiments of the present invention, through the dynamic thermal gradient boundary, i.e., the boundary mask of the effective radiation area of the treatment head (such as 5% - 10% corresponding to the peak), the coverage range of a more effective sound field can be determined.

[0067] 4. In the embodiments of the present invention, for a thermal imaging video, the thermal imaging video is segmented according to a preset time window and sliding step length to obtain at least two frame sequences. Subsequently, the at least two frame sequences can be used to generate the input data in multiple groups of training samples, without the need to repeat a large number of complete sample acquisition processes such as turning on the treatment head, turning on the infrared thermal imager, and recording the thermal imaging video. By turning on the treatment head once and recording the thermal imaging video once, a large number of training samples, such as several hundred groups, can be obtained, reducing the dependence on a large amount of labeled data, and the implementation method is more convenient.

[0068] 5. In the embodiments of the present invention, data enhancement can be performed on multiple frame sequences formed from the same thermal imaging video, such as spatial enhancement and / or temporal enhancement, etc., so as to expand one frame sequence into multiple frame sequences, respectively corresponding to the input data in multiple groups of training samples. Therefore, there is no need to repeat a large number of steps of turning on the treatment head and recording the thermal imaging video. By turning on the treatment head once and recording the thermal imaging video once, a large number of multiple groups of training samples can be obtained, reducing the dependence on a large amount of labeled data, and the implementation method is more convenient.

[0069] 6. The infrared thermal imager used in the embodiments of the present invention is easier to obtain and easier to operate compared to the hydrophone used in the prior art, reducing the implementation cost and difficulty.

[0070] 7. In the embodiments of the present invention, there is no need to place the infrared thermal imager in a specific working environment. For example, there is no need to use a test container filled with water in the prior art, and there is no need for an approximate free-field condition with the water temperature at a specific temperature, such as 22°±3°. Therefore, detection can be carried out in a non-laboratory environment. In actual business applications, detection in a non-laboratory environment is often required. For example, when an ultrasonic physiotherapy device is sold and used in a hospital, it may be necessary to regularly detect whether the effective radiation area of the treatment head of the ultrasonic physiotherapy device meets the requirements in the hospital, so as to continuously verify safety and further ensure safety. Another example is that in many scenarios, non-professional testers are required to perform the measurement process. Since the method for measuring the effective radiation area of the treatment head in the embodiments of the present invention is suitable for execution in a non-laboratory environment, the development of the business is greatly expanded. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0072] Figure 1 is a flowchart of a method for training an AI model in an embodiment of the present invention.

[0073] Figure 2 is a flowchart of a method for generating multiple sets of training samples using a thermal imaging video in an embodiment of the present invention.

[0074] Figure 3 is a flowchart of a method for measuring the effective radiation area of the treatment head of an ultrasonic physiotherapy device in an embodiment of the present invention.

[0075] Figure 4 is a schematic structural diagram of a training device for an AI model in an embodiment of the present invention.

[0076] Figure 5 is a schematic structural diagram of a system for measuring the effective radiation area of the treatment head of an ultrasonic physiotherapy device in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0077] The following will describe the solutions provided by the present invention in conjunction with the drawings.

[0078] First of all, it should be noted that the terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The singular forms "a", "the" and "said" used in the embodiments of the present invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0079] It should be understood that the term " / and" used herein is only a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A / and B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0080] In the prior art, the method for measuring the effective radiation area of the treatment head of an ultrasonic physiotherapy device includes: using a hydrophone to measure the beam cross-sectional area, and then calculating the effective radiation area using the beam cross-sectional area. This involves a large number of professional and cumbersome operation processes for professional instruments such as hydrophones, as well as complex calculation processes for calculating the effective radiation area using the beam cross-sectional area. It can be seen that the measurement method of the prior art has at least the following disadvantages:

[0081] 1. To measure the effective radiation area of the treatment head of an ultrasonic physiotherapy device, the prior art needs to use professional measurement tools such as hydrophones for measurement. Therefore, additional professional hardware tools are required to assist during the measurement process, and professional testers are needed to perform the measurement process, increasing the implementation cost and technical difficulty.

[0082] 2. During the detection process of the prior art, the ultrasonic physiotherapy device and the hydrophone need to be placed in an effective working environment, that is, a specific working environment is required for testing. For example, a test container filled with water needs to be prepared, the treatment head of the ultrasonic physiotherapy device and the hydrophone need to be immersed in water, and all measurements need to be carried out under approximate free-field conditions with the water temperature at a specific temperature, such as 22°±3°. In addition, professional tools such as hydrophones are required, and professional measurement operations need to be performed using the hydrophone. Therefore, it is actually only suitable for detection in a laboratory and not suitable for detection in a non-laboratory environment. In actual business applications, detection in a non-laboratory environment is often required. For example, when the ultrasonic physiotherapy device is sold and used in a hospital, it may be necessary to regularly detect whether the effective radiation area of the treatment head of the ultrasonic physiotherapy device meets the requirements in the hospital, so as to continuously verify safety and further ensure safety; again, in many scenarios, non-professional testers are required to perform the measurement process. It can be seen that the measurement method of the effective radiation area of the treatment head in the prior art greatly restricts the development of the business.

[0083] 3. During the measurement process of the prior art, a hydrophone is used for measurement, and the measurement accuracy highly depends on the use of the hydrophone. If the tester uses the hydrophone incorrectly or the measurement accuracy of the hydrophone is inaccurate, it may lead to inaccurate sampling data during a large number of repeated samplings, resulting in the inability to accurately detect the effective radiation area of the treatment head of the ultrasonic physiotherapy device.

[0084] 4. When using a hydrophone in the prior art, it is necessary to scan the sound field point by point, which is time-consuming and relies on manual operation, and is prone to errors, resulting in more time-consuming for the measurement process of the effective radiation area of the treatment head of the ultrasonic physiotherapy device, which is not conducive to improving the work efficiency of the detection personnel.

[0085] In view of various problems of the prior art, in the embodiments of the present invention, an artificial intelligence model is used to measure the effective radiation area of the treatment head, so as to solve various problems of using a hydrophone in the prior art in a more automated, intelligent, and simple operation form.

[0086] In the embodiments of the present invention, the applicant considers that ultrasonic waves will generate a thermal effect in a medium, and the area where the temperature rises is related to the absorption of sound energy. That is to say, within a fixed medium and time, the temperature rise is positively correlated with the sound intensity, and the heat distribution can indirectly reflect the sound field energy distribution. Therefore, in the embodiments of the present invention, an infrared thermal imager with simpler operation is used, and an AI model is trained based on AI technology and infrared temperature measurement technology, and then the AI model is used to predict the effective radiation area of the treatment head of the ultrasonic physiotherapy device.

[0087] Based on this, in an embodiment of the present invention, a training method of the AI model is proposed. The AI model is used to measure the effective radiation area of the treatment head of the ultrasonic physiotherapy device. See Figure 1 , and the training method includes:

[0088] Step 101: Generate multiple groups of training samples; where each group of training samples includes input data and the label data corresponding to the input data;

[0089] Step 103: Input each group of training samples into the AI model to be trained to train the AI model;

[0090] Among them, in step 101, the generation method of the training samples includes:

[0091] Step 1011: Orient the treatment head of the ultrasonic physiotherapy device towards the surface of the sound-absorbing medium, and the treatment head is at a preset distance from the surface of the sound-absorbing medium;

[0092] Step 1013: Adjust the focal length of the infrared thermal imager so that the focal length of the infrared thermal imager is aligned with the radiation area of the treatment head of the ultrasonic physiotherapy device;

[0093] Step 1015: Activate the treatment head to output ultrasonic waves, which are transmitted to the surface of the sound-absorbing medium.

[0094] Step 1017: Record a thermal imaging video through an infrared thermal imager. The thermal imaging video includes the dynamic process of the temperature on the surface of the sound-absorbing medium increasing with time when the treatment head of the ultrasonic physiotherapy device outputs ultrasonic waves.

[0095] Step 1019: Extract the instantaneous temperature change rate of each frame in the thermal imaging video.

[0096] Step 1021: Generate a sequence of thermal maps of temperature change rates based on the instantaneous temperature change rate of each frame, and use the sequence of thermal maps of temperature change rates as input data in the training samples.

[0097] Step 1023: Use the boundary mask of the effective radiation area of the treatment head as the label data in the training samples.

[0098] Ultrasonic waves will produce a thermal effect in the medium. The area where the temperature rises is related to the absorption of sound energy. That is to say, at a fixed medium and time, the temperature rise is positively correlated with the sound intensity, and the thermal distribution can indirectly reflect the sound field energy distribution. Therefore, in Figure 1 the method of the embodiment of the present invention shown, the following idea is adopted to train the AI model: under short-term and controllable heat diffusion conditions, the temperature gradient boundary can effectively approximate the sound intensity threshold boundary, and use the AI model to learn how to determine the boundary contour of the temperature gradient boundary (such as the 5% of the peak value where the gradient drops). During subsequent actual measurements, combined with the information of the temperature gradient boundary output by the AI model, that is Figure 1 the "boundary mask of the effective radiation area" in, after conversion, the effective radiation area of the treatment head can be obtained. Therefore, in the above Figure 1 shown process, the core role of the AI model is: to extract the boundary contour of the effective radiation area from the sequence of thermal maps of temperature change rates (i.e., a kind of dynamic thermal gradient data), so that the information of the effective radiation area output by the AI model can be used to replace the information of the effective radiation area obtained by time-consuming, laborious and complex manual interpretation or point-by-point scanning of hydrophones in the prior art during subsequent actual measurement processes. Therefore, the intelligence and efficiency are greatly improved.

[0099] Combined with the above Figure 1 shown process, a more detailed method for training the AI model in the embodiment of the present invention is described.

[0100] In an embodiment of the present invention, in step 1011, the preset distance between the treatment head and the surface of the sound-absorbing medium can be 0.3 cm.

[0101] In an embodiment of the present invention, the sound-absorbing medium is a tissue-mimicking material with the difference between its acoustic impedance and that of human tissue being less than a predetermined value. For example, the sound-absorbing medium includes a tissue-mimicking hydrogel with the difference between its acoustic impedance and that of human tissue being less than a predetermined value.

[0102] Because the shorter the time, the stronger the positive correlation between temperature rise and sound intensity, and the more accurately the heat distribution can reflect the sound field energy distribution. Therefore, in an embodiment of the present invention, in step 1015, the treatment head outputs short-time ultrasonic pulses to reduce the influence of heat diffusion. For example, the treatment head outputs short-time ultrasonic pulses with a pulse width of 10 ms.

[0103] To improve the training effect of the AI model, data augmentation processing can be performed on the input data in the training samples in the following ways:

[0104] Spatial augmentation: Random rotation (-10° to 10°), translation (±5%), scaling (0.9 to 1.1 times); simulating treatment head offset or uneven medium surface.

[0105] Noise injection: Adding Gaussian noise (simulating environmental temperature fluctuations); performing local occlusion (simulating accidental occlusion during operation).

[0106] Temporal augmentation: Randomly intercepting time segments of the thermal image sequence (such as randomly selecting 20 frames from 30 frames).

[0107] In an embodiment of the present invention, in order to achieve data balance and diversity during the training process of the AI model, during the generation process of different groups of training samples:

[0108] 1. The types of treatment heads used are different;

[0109] To generate multiple groups of training samples, a planar treatment head is used when generating the first group of training samples, a focused treatment head is used when generating the second group of training samples, a treatment head with one frequency is used when generating the third group of training samples, and a treatment head with another frequency is used when generating the fourth group of training samples.

[0110] 2. The types of infrared thermal imagers used are different;

[0111] 3. The power used by the treatment head when generating ultrasonic waves is different;

[0112] For example, when generating the fifth group of training samples, the power of the treatment head is set to 1 W / cm2, and when generating the sixth group of training samples, the power of the treatment head is set to 1.5 W / cm2, etc.

[0113] 4. The types of sound-absorbing media are different, such as tissue-mimicking hydrogels, media containing impurities or bubbles, etc., to increase the diversity of training samples.

[0114] Attention can be paid to the balance of the sample proportions of different types of treatment heads and different types of sound absorption media.

[0115] In an embodiment of the present invention, input data in a set of training samples can be generated by using a thermal imaging video recorded by an infrared thermal imager.

[0116] In order to improve the training efficiency, input data in more sets of training samples can also be generated by using a thermal imaging video recorded by an infrared thermal imager. At this time, referring to Figure 2 , after recording the thermal imaging video by the infrared thermal imager in step 1017 and before extracting the instantaneous temperature change rate of each frame in the thermal imaging video in step 1019, it further includes:

[0117] Step 1018: For the thermal imaging video, segment the thermal imaging video according to a preset time window and sliding step length to obtain at least two frame sequences;

[0118] In step 1018, for example, a thermal imaging video includes a total of 100 working video frames of the treatment head, the length of the time window is 10 frames (capturing a dynamic thermal gradient map of 0.1 seconds), and the sliding step length is 1 frame. Then, these 100 frames of the thermal imaging video can be segmented into 91 frame sequences in total, which can actually correspond to the input data in 91 sets of training samples.

[0119] Of course, if more sets of training samples are to be obtained, data augmentation can also be performed on the 91 frame sequences formed by the same thermal imaging video respectively, such as the above-mentioned spatial augmentation and / or temporal augmentation, etc., so as to expand one frame sequence into multiple frame sequences, corresponding to the input data in multiple sets of training samples respectively. In this way, these 100 frames of the thermal imaging video can form far more than 91 frame sequences in total, that is, the subsequent corresponding input data in far more than 91 sets (such as dozens to hundreds of sets) of training samples. Therefore, if the training process is to be further simplified, it is not necessary to repeatedly execute a large number of steps of turning on the treatment head and recording the thermal imaging video to collect samples, but use the above Figure 2 As shown, by turning on the treatment head once / few times and collecting and recording the thermal imaging video once / few times, a large number of training samples can be obtained, and the implementation method is more convenient.

[0120] Correspondingly, the above step 1019 is specifically step A: For each frame sequence, extract the instantaneous temperature change rate of each frame in the frame sequence;

[0121] Correspondingly, the above-mentioned step 1021 is specifically step B: generating a sequence of thermal maps of temperature change rates based on the instantaneous temperature change rate of each frame in the frame sequence, and using the sequence of thermal maps of temperature change rates generated for each frame sequence as the input data in a separate set of training samples, so as to generate the input data in at least two sets of training samples for at least two frame sequences of a thermal imaging video;

[0122] Correspondingly, the above-mentioned step 1023 is specifically step C: using the boundary mask of the effective radiation area of the treatment head used when forming the thermal imaging video as the label data in the at least two sets of training samples.

[0123] In an embodiment of the present invention, the method for obtaining the boundary mask of the effective radiation area of the treatment head, which is the label data in each set of training samples, may include: obtaining the sound intensity distribution through the hydrophone scanning method in the prior art, and generating a binary mask (1 represents the effective area, 0 represents the invalid area) with a threshold of 5% to 10% of the sound intensity peak as the label data.

[0124] In an embodiment of the present invention, the trained AI model can learn and output the boundary mask of the effective radiation area of the treatment head, and the AI model cannot directly output the effective radiation area of the treatment head. During subsequent measurements, the boundary mask of the effective radiation area of the treatment head output by the AI model can be converted (using the calibration parameters of the infrared thermal imager, and this calibration coefficient indicates the number of pixels corresponding to each unit area when using this infrared thermal imager, such as 1 pixel = 0.1 mm 2 ) to obtain the effective radiation area of the treatment head.

[0125] However, in another embodiment of the present invention, it may also involve an end-to-end system, embedding the calibration parameters of the infrared thermal imager into the AI model. In this way, the AI model can further learn and directly output the effective radiation area of the treatment head. At this time, when training the AI model, the input data in each set of training samples further includes: the calibration parameters of the infrared thermal imager, and this calibration coefficient indicates the number of pixels corresponding to each unit area when using this infrared thermal imager;

[0126] The label data in each set of training samples further includes: the effective radiation area of the treatment head of the ultrasonic physiotherapy device.

[0127] For the AI model, the backbone network can adopt a lightweight U-Net structure, which is suitable for embedded deployment and has a moderate number of parameters. Additionally, 3D convolutional layers or LSTM modules can be added to capture the temporal features of the heat map sequence for outputting the boundary mask of the effective radiation area of the treatment head of the ultrasonic physiotherapy device. In the embodiments of the present invention, the types of loss functions of the AI model include: Dice Loss + boundary-sensitive loss. Further, a fully connected layer can be additionally provided in the AI model. The input of this fully connected layer is the number of mask pixels corresponding to the above-mentioned boundary mask and the calibration parameters of the infrared thermal imager, and the output of this fully connected layer is the effective radiation area of the treatment head of the ultrasonic physiotherapy device.

[0128] After multiple rounds of training until the AI model converges, this AI model can be used to measure the effective radiation area of the treatment head of an ultrasonic physiotherapy device. Refer to Figure 3 , for the treatment head (for ease of description, denoted as treatment head 1) of the ultrasonic physiotherapy device to be measured (for ease of description, denoted as ultrasonic physiotherapy device 1), the method for measuring its effective radiation area includes:

[0129] Step 301: Orient the treatment head 1 of the ultrasonic physiotherapy device 1 towards the surface of the sound-absorbing medium, with the treatment head 1 at a preset distance from the surface of the sound-absorbing medium, such as 0.3 cm; the difference between the acoustic impedance of the sound-absorbing medium and the acoustic impedance of human tissue is less than a predetermined value;

[0130] Step 303: Adjust the focal length of the currently used infrared thermal imager (for ease of description, denoted as infrared thermal imager 1) so that the focal length of the infrared thermal imager 1 is aligned with the radiation area of the treatment head 1;

[0131] Step 305: Start the treatment head 1 to output ultrasonic waves, and these ultrasonic waves are radiated onto the surface of the sound-absorbing medium;

[0132] Step 307: Record a thermal imaging video through the infrared thermal imager 1, and this thermal imaging video includes the dynamic process of the temperature on the surface of the sound-absorbing medium increasing with time when the treatment head 1 of the ultrasonic physiotherapy device 1 is working;

[0133] Step 309: Extract the instantaneous temperature change rate of each frame in this thermal imaging video;

[0134] Step 311: Generate a sequence of temperature change rate heat maps based on the instantaneous temperature change rate of each frame;

[0135] Step 313: Input this sequence of temperature change rate heat maps into the AI model, where this AI model is trained by the method of any embodiment of the present invention;

[0136] Step 315: Obtain the boundary mask of the effective radiation area of the treatment head 1 of the ultrasonic physiotherapy device 1 output by this AI model;

[0137] Step 317: Multiply the number of pixels corresponding to the boundary mask of the effective radiation area of the treatment head 1 by the calibration parameter 1 of the infrared thermal imager 1 to obtain the effective radiation area of the treatment head 1 of the ultrasonic physiotherapy device 1 to be measured; wherein, the calibration parameter 1 indicates the number of pixels corresponding to each unit area when using the infrared thermal imager 1.

[0138] In the embodiment of the present invention, the boundary mask is a binary image output by an AI model. The white area (pixel value = 1) represents the effective radiation area of the treatment head 1, and the black area (pixel value = 0) represents the non-effective radiation area of the treatment head 1. For example, the total number of white pixels in the boundary mask of the effective radiation area of the treatment head 1 is 5000, which directly reflects the coverage range of the effective radiation area in the image. The calibration parameter 1 of the infrared thermal imager 1 is the pixel resolution of the infrared thermal imager 1, which can be calibrated through experiments, that is, the actual physical area corresponding to a single pixel. For example, the calibration parameter 1 is: 1 pixel = 0.01 mm2. At this time, the effective radiation area of the treatment head 1 = the number of pixels corresponding to the boundary mask of the effective radiation area output by the AI model * the calibration parameter 1 of the infrared thermal imager 1 (i.e., the actual area of a single pixel) = (5000 × 0.01) / 100 = 0.5 cm2.

[0139] In an embodiment of the present invention, if the trained AI model can not only output the boundary mask of the effective radiation area of the treatment head, but also further directly output the effective radiation area of the treatment head, then in the above step 317, the AI model will perform the processing of this step 317, and the AI model will output the effective radiation area of the treatment head 1 of the ultrasonic physiotherapy device 1 to be measured.

[0140] In the prior art, when using a hydrophone for measurement, the measurement must be performed in a specific working environment with an ultrasonic physiotherapy device, and remote measurement cannot be achieved. However, in the embodiment of the present invention, because the AI model is used to perform the measurement, the process from step 309 to step 317 can be remotely executed.

[0141] In the embodiment of the present invention, the calibration method for the calibration parameter of the infrared thermal imager may include:

[0142] 1. Place a standard template with a known size (such as a square calibration plate with a side length of 1 cm) in the field of view of the infrared thermal imager;

[0143] 2. Calculate the number of pixels of the standard template in the infrared thermal imager image (such as 100 × 100 pixels);

[0144] 3. The actual area of a single pixel = the actual area of the template / the number of template pixels = (1 cm × 1 cm) / 10,000 pixels = 0.0001 cm2 / pixel, that is, 1 pixel = 0.01 mm2.

[0145] As can be seen, each embodiment of the present invention has at least the following beneficial effects:

[0146] 1. In the embodiments of the present invention, an AI model is used to learn how to determine the temperature gradient boundary, that is, to extract the boundary mask of the effective radiation area from a sequence of thermal maps of temperature change rates (i.e., a kind of dynamic thermal gradient data). Thus, in the subsequent actual measurement process, the information of the effective radiation area output by the AI model can be used to replace the information of the effective radiation area obtained by the time-consuming, laborious and complex manual interpretation or point-by-point scanning of hydrophones in the prior art. Therefore, the intelligence and efficiency are greatly improved. For example, in the prior art, when using a hydrophone, it is necessary to scan the sound field point by point, which takes a long time, depends on manual operation, and is prone to errors. In the embodiments of the present invention, a single thermal imaging video can capture the distribution of temperature change rates in the entire sound field area. By training the AI model to directly identify the dynamic gradient boundary, that is, the boundary mask of the effective radiation area, manual interpretation is not required, which is efficient and intelligent.

[0147] 2. In the embodiments of the present invention, a dynamic thermal gradient boundary is used. The dynamic thermal gradient boundary refers to determining the boundary of the distribution of ultrasonic energy in a medium by analyzing the rate of change of temperature over time (dT / dt) rather than static temperature values. The dynamic thermal gradient boundary in the embodiments of the present invention has dynamics, that is, it focuses on the instantaneous effect of temperature change rather than the thermal equilibrium in the steady state. Moreover, the dynamic thermal gradient boundary in the embodiments of the present invention has gradient sensitivity, that is, it distinguishes the effective sound field area from the non-effective sound field area by the speed of temperature change (the steepness of the gradient). Therefore, the sound intensity of ultrasonic waves can be better corresponded by means of temperature.

[0148] 3. The embodiments of the present invention better eliminate the influence of heat diffusion. Considering that heat will diffuse over time and the thermal conductivity of the medium, resulting in the thermal imaging boundary (i.e., the temperature distribution) being larger than the sound field boundary, which will not accurately reflect the sound field boundary. To solve this problem, instead of using static temperature values to determine the boundary in the embodiments of the present invention, a sequence of thermal maps of temperature change rates (i.e., a kind of dynamic thermal gradient data) is used to replace the conventional static temperature values, so as to capture the instantaneous effect of sound energy input. The maximum area of the temperature change rate is closer to the instantaneous distribution of sound energy in time. Specifically, when ultrasonic waves act on the medium, the temperature in the sound energy input area will rise rapidly (manifested as a high temperature change rate dT / dt), while the temperature in the surrounding area changes slowly due to sound intensity attenuation and heat diffusion (manifested as a low temperature change rate dT / dt). On the other hand, in the embodiments of the present invention, through the dynamic thermal gradient boundary, that is, the boundary mask of the effective radiation area of the treatment head (such as 5% - 10% corresponding to the peak value), the coverage range of a more effective sound field can be determined.

[0149] 4. In an embodiment of the present invention, for a thermal imaging video, the thermal imaging video is segmented according to a preset time window and a sliding step size to obtain at least two frame sequences. Subsequently, the at least two frame sequences can be used to generate input data in multiple groups of training samples. There is no need to repeat a large number of complete sample collection processes such as turning on the treatment head, turning on the infrared thermal imager, and recording the thermal imaging video. By turning on the treatment head once and collecting and recording the thermal imaging video once, a large number of, for example, hundreds of groups of training samples can be obtained, which reduces the dependence on a large amount of labeled data and is more convenient to implement.

[0150] 5. In an embodiment of the present invention, data enhancement, such as spatial enhancement and / or temporal enhancement, can be performed separately on multiple frame sequences formed by the same thermal imaging video, so that one frame sequence is expanded into multiple frame sequences, which respectively correspond to the input data in multiple groups of training samples. Therefore, there is no need to repeat a large number of steps of turning on the treatment head and recording the thermal imaging video. A large number of multiple groups of training samples can be obtained by turning on the treatment head once and recording the thermal imaging video once, which reduces the dependence on a large amount of labeled data and makes the implementation method simpler.

[0151] 6. The infrared thermal imager used in the embodiment of the present invention is easier to obtain and operate than the hydrophone used in the prior art, which reduces the implementation cost and difficulty.

[0152] 7. In the embodiments of the present invention, there is no need to place the infrared thermal imager in a specific working environment. For example, there is no need to use a test container filled with water in the prior art, and there is no need for the water temperature to be at a specific temperature, such as 22°±3°, which is close to the free field condition. Therefore, detection can be performed in a non-laboratory environment. In actual business applications, it is often necessary to perform detection in a non-laboratory environment. For example, when an ultrasonic therapy device is sold and used in a hospital, it may be necessary to regularly test the effective radiation area of ​​the treatment head of the ultrasonic therapy device in the hospital to see if it meets the requirements, so as to continuously verify safety and further ensure safety. For example, in many scenarios, non-professional testers are required to perform the measurement process. Because the method for measuring the effective radiation area of ​​the treatment head in the embodiment of the present invention is suitable for execution in a non-laboratory environment, it greatly expands the development of the business.

[0153] In one embodiment of the present invention, a training device for an AI model is proposed, see Figure 4 The AI ​​model is used to measure the effective radiation area of ​​the treatment head of the ultrasonic therapy device. The training device includes:

[0154] The training sample generation module 401 is configured to generate multiple sets of training samples; wherein each set of training samples includes input data and label data corresponding to the input data;

[0155] The training execution module 402 is configured to input each group of training samples into the AI model to be trained for training the AI model;

[0156] Among them, the training sample generation module 401 is used to execute:

[0157] Receive the thermal imaging video recorded by the infrared thermal imager, where the thermal imaging video includes the dynamic process of the temperature on the surface of the sound absorption medium increasing with time when the treatment head of the ultrasonic physiotherapy device outputs ultrasonic waves;

[0158] Extract the instantaneous temperature change rate of each frame in the thermal imaging video;

[0159] Generate a sequence of temperature change rate heat maps according to the instantaneous temperature change rate of each frame, and use the sequence of temperature change rate heat maps as the input data in the training samples; and

[0160] Use the boundary mask of the effective radiation area of the treatment head as the label data in the training samples.

[0161] An embodiment of the present invention proposes a measurement system for the effective radiation area of the treatment head of an ultrasonic physiotherapy device. Refer to Figure 5 , this system includes: an infrared thermal imager 501, a computer processing module 502, an AI model 503, and a calculation module 504;

[0162] The first treatment head of the ultrasonic physiotherapy device to be measured faces the surface of the sound absorption medium, and the first treatment head is at a preset distance from the surface of the sound absorption medium; the difference between the acoustic impedance of the sound absorption medium and the acoustic impedance of human tissue is less than a predetermined value; after the first treatment head is activated, it outputs ultrasonic waves, and the ultrasonic waves are transmitted to the surface of the sound absorption medium;

[0163] The focal length of the infrared thermal imager 501 is adjusted to align with the radiation area of the first treatment head of the ultrasonic physiotherapy device; the infrared thermal imager 501 is used to record a thermal imaging video after activating the first treatment head, and the thermal imaging video includes the dynamic process of the temperature on the surface of the sound absorption medium increasing with time when the first treatment head of the ultrasonic physiotherapy device outputs ultrasonic waves;

[0164] The computer processing module 502 is configured to extract the instantaneous temperature change rate of each frame in the thermal imaging video; generate a sequence of temperature change rate heat maps according to the instantaneous temperature change rate of each frame; input the sequence of temperature change rate heat maps into the AI model 503; among them, the AI model is trained by using the device described in claim 9;

[0165] The AI model 503 is used to output the boundary mask of the effective radiation area of the first treatment head of the ultrasonic physiotherapy device;

[0166] The calculation module 504 is configured to multiply the number of pixels corresponding to the boundary mask of the effective radiation area of the first treatment head by the first calibration parameter of the first infrared thermal imager to obtain the effective radiation area of the first treatment head of the ultrasonic physiotherapy device to be measured; wherein, the first calibration coefficient indicates the number of pixels corresponding to each unit area when using the first infrared thermal imager.

[0167] In an embodiment of the system of the present invention, the calculation module 504 may be integrated in the AI model 503.

[0168] In an embodiment of the system of the present invention, the preset distance between the treatment head and the surface of the sound-absorbing medium is 0.3 cm.

[0169] In an embodiment of the system of the present invention, the sound-absorbing medium includes a tissue-mimicking hydrogel with a difference between the acoustic impedance and the acoustic impedance of human tissue less than a predetermined value.

[0170] In an embodiment of the system of the present invention, the treatment head outputs short-time ultrasonic pulses.

[0171] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed in a computer, the computer is made to execute the method in any one of the embodiments in the specification.

[0172] An embodiment of the present invention provides a computing device, including a memory and a processor. An executable code is stored in the memory. When the processor executes the executable code, the method in any one of the embodiments in the specification is implemented.

[0173] It can be understood that the structure schematically shown in the embodiments of the present invention does not constitute a specific limitation on the device of the embodiments of the present invention. In other embodiments of the specification, the above device may include more or fewer components than shown in the figures, or combine certain components, or split certain components, or have different component arrangements. The components shown in the figures may be implemented in hardware, software, or a combination of software and hardware.

[0174] Each embodiment in the present invention is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and the differences between each embodiment and other embodiments are emphasized. In particular, for the device embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiments.

[0175] Those skilled in the art should be able to realize that in one or more of the above examples, the functions described in the present invention can be implemented by hardware, software, add-ons, or any combination thereof. When implemented using software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium.

[0176] The specific embodiments described above have further elaborated on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made on the basis of the technical solutions of the present invention should be included within the protection scope of the present invention.

Claims

1. The training method of the AI ​​model is characterized by: The AI ​​model is used to measure the effective radiation area of ​​the treatment head of the ultrasonic physiotherapy device. The training method includes: Generate multiple sets of training samples; each set of training samples includes input data and label data corresponding to the input data; Inputting each group of training samples into the AI ​​model to be trained to train the AI ​​model; The method for generating training samples includes: Directing a treatment head of the ultrasonic therapy device toward the surface of the sound-absorbing medium, with the treatment head being at a preset distance from the surface of the sound-absorbing medium; Adjusting the focus of the infrared thermal imager so that the focus of the infrared thermal imager is aligned with the radiation area of ​​the treatment head of the ultrasonic physiotherapy device; The treatment head is started, and the treatment head outputs ultrasonic waves, which are transmitted to the surface of the sound absorbing medium; Recording a thermal imaging video by an infrared thermal imager, the thermal imaging video includes a dynamic process of the temperature of the surface of the sound absorbing medium increasing over time when the treatment head of the ultrasonic therapy device outputs ultrasonic waves; Extract the instantaneous temperature change rate of each frame in the thermal imaging video; generating a temperature change rate heat map sequence according to the instantaneous temperature change rate of each frame, and using the temperature change rate heat map sequence as input data in a training sample; and The boundary mask of the effective radiation area of ​​the treatment head is used as the label data in the training sample.

2. The method according to claim 1, characterized in that The preset distance between the treatment head and the surface of the sound absorbing medium is 0.3 cm; and / or The sound absorbing medium is a tissue-mimicking material whose difference between the acoustic impedance and the acoustic impedance of human tissue is less than a predetermined value; and / or The sound absorbing medium includes a tissue-mimicking hydrogel whose acoustic impedance has a difference with that of human tissue that is less than a predetermined value; and / or The treatment head outputs short-duration ultrasonic pulses.

3. The method according to claim 1, characterized in that In the process of generating different groups of training samples: The type of treatment tip used is different; and / or The type of thermal imaging camera used is different; and / or The treatment tip generates ultrasound waves at different powers; and / or The types of sound absorbing media are different to increase the diversity of training samples.

4. The method according to claim 1, characterized in that: After recording the thermal imaging video by the infrared thermal imager and before extracting the instantaneous temperature change rate of each frame in the thermal imaging video, the method further includes: segmenting the thermal imaging video according to a preset time window and a sliding step size to obtain at least two frame sequences; Then, the extracting the instantaneous temperature change rate of each frame in the thermal imaging video includes: for each frame sequence, extracting the instantaneous temperature change rate of each frame in the frame sequence; Then, generating a temperature change rate heat map sequence according to the instantaneous temperature change rate of each frame includes: generating a temperature change rate heat map sequence for the instantaneous temperature change rate of each frame in each frame sequence; Then, the using of the temperature change rate heat map sequence as input data in the training sample includes: using each temperature change rate heat map sequence generated for each frame sequence as input data in a separate set of training samples, thereby generating input data in at least two sets of training samples for at least two frame sequences of a thermal imaging video; Then, using the boundary mask of the effective radiation area of ​​the treatment head as the label data in the training samples includes: using the boundary mask of the effective radiation area of ​​the treatment head used when forming the thermal imaging video as the label data in the at least two groups of training samples.

5. The method according to claim 1, characterized in that The input data of each set of training samples further includes: calibration parameters of the infrared thermal imager, the calibration coefficient indicating the number of pixels per unit area when the infrared thermal imager is used; The label data of each set of training samples further includes: the effective radiation area of ​​the treatment head of the ultrasonic therapy device.

6. A method for measuring the effective radiation area of ​​a treatment head of an ultrasonic therapy device, characterized in that: The method includes: The first treatment head of the ultrasonic therapy device to be measured is directed toward the surface of the sound absorbing medium, and the first treatment head is at a preset distance from the surface of the sound absorbing medium; the difference between the acoustic impedance of the sound absorbing medium and the acoustic impedance of human tissue is less than a predetermined value; Adjusting the focal length of the infrared thermal imager so that the focal length of the first infrared thermal imager is aligned with the radiation area of ​​the first treatment head of the ultrasonic physiotherapy device; Starting the first treatment head, so that the first treatment head outputs ultrasonic waves, and the ultrasonic waves are transmitted to the surface of the sound absorbing medium; Recording a thermal imaging video by a first infrared thermal imager, the thermal imaging video includes a dynamic process of the temperature of the surface of the sound absorbing medium increasing over time when the first treatment head of the ultrasonic therapy device outputs ultrasonic waves; Extract the instantaneous temperature change rate of each frame in the thermal imaging video; Generate a temperature change rate heat map sequence according to the instantaneous temperature change rate of each frame; Inputting the temperature change rate heat map sequence into an AI model, wherein the AI ​​model is trained using the method described in any one of claims 1 to 6; Obtaining a boundary mask of an effective radiation area of ​​the first treatment head output by the AI ​​model; The effective radiation area of ​​the first treatment head of the ultrasonic therapy device to be measured is obtained by multiplying the number of pixels corresponding to the boundary mask of the effective radiation area of ​​the first treatment head by the first calibration parameter of the first infrared thermal imager; wherein the first calibration coefficient indicates the number of pixels corresponding to each unit area when the first infrared thermal imager is used.

7. The method according to claim 6, characterized in that The AI ​​model executes the step of "multiplying the number of pixels corresponding to the boundary mask of the effective radiation area of ​​the first treatment head by the first calibration parameter of the first infrared thermal imager to obtain the effective radiation area of ​​the first treatment head of the ultrasonic physiotherapy device to be measured"; The method further includes: the AI ​​model outputs the effective radiation area of ​​the first treatment head of the ultrasonic therapy device to be measured.

8. The method according to claim 6, characterized in that The treatment head outputs short-duration ultrasonic pulses to reduce the influence of heat diffusion.

9. A training device for an AI model, characterized in that: The AI ​​model is used to measure the effective radiation area of ​​the treatment head of the ultrasonic physiotherapy device. The training device includes: A training sample generation module is configured to generate multiple groups of training samples; wherein each group of training samples includes input data and label data corresponding to the input data; A training execution module, configured to input each group of training samples into the AI ​​model to be trained to train the AI ​​model; Among them, the training sample generation module is used to perform: receiving a thermal imaging video recorded by an infrared thermal imager, wherein the thermal imaging video includes a dynamic process of the temperature of the surface of the sound absorbing medium increasing over time when the treatment head of the ultrasonic therapy device outputs ultrasonic waves; Extract the instantaneous temperature change rate of each frame in the thermal imaging video; generating a temperature change rate heat map sequence according to the instantaneous temperature change rate of each frame, and using the temperature change rate heat map sequence as input data in a training sample; and The boundary mask of the effective radiation area of ​​the treatment head is used as the label data in the training sample.

10. A system for measuring the effective radiation area of ​​a treatment head of an ultrasonic therapy device, characterized in that: The system includes: an infrared thermal imager, a computer processing module, an AI model and a computing module; The first treatment head of the ultrasonic therapy device to be measured faces the surface of the sound absorbing medium, and the first treatment head is at a preset distance from the surface of the sound absorbing medium; the difference between the acoustic impedance of the sound absorbing medium and the acoustic impedance of human tissue is less than a predetermined value; the first treatment head outputs ultrasonic waves after being activated, and the ultrasonic waves are transmitted to the surface of the sound absorbing medium; The focus of the infrared thermal imager is adjusted to align with the radiation area of ​​the first treatment head of the ultrasonic therapy device; the infrared thermal imager is used to record a thermal imaging video after the first treatment head is started, and the thermal imaging video includes a dynamic process of the temperature of the surface of the sound absorbing medium increasing over time when the first treatment head of the ultrasonic therapy device outputs ultrasonic waves; A computer processing module configured to extract the instantaneous temperature change rate of each frame in the thermal imaging video; generate a temperature change rate heat map sequence according to the instantaneous temperature change rate of each frame; and input the temperature change rate heat map sequence into an AI model; wherein the AI ​​model is trained using the device described in claim 9; An AI model, used to output a boundary mask of an effective radiation area of ​​a first treatment head of an ultrasonic therapy device; The calculation module is configured to obtain the effective radiation area of ​​the first treatment head of the ultrasonic therapy device to be measured by multiplying the number of pixels corresponding to the boundary mask of the effective radiation area of ​​the first treatment head by the first calibration parameter of the first infrared thermal imager; wherein the first calibration coefficient indicates the number of pixels corresponding to each unit area when the first infrared thermal imager is used.

Images