Loading method of ultrasound device, ultrasound device, and storage medium

Abstract

The application provides a loading method and device of an ultrasonic equipment, electronic equipment and a storage medium. The loading method of the ultrasonic equipment comprises the following steps: determining selected options, wherein all the options are organized as a hierarchical architecture, and there is at least one first option in all the options, and the number of next-level options of each first option is greater than 1; for each model, determining an evaluation parameter of the model according to the same option in the selected option and all the options related to the model, wherein there are a plurality of first models in all the models, and at least two first models each include different next-level options of the same first option in all the options related to the first models respectively, and each evaluation parameter corresponds to a level at which all the options related to the model are located respectively; for each model, determining a priority of the model according to the evaluation parameter of the model; and loading at least one model with the highest priority. The application can improve the operation efficiency of the ultrasonic equipment.

Description

Technical Field

[0001] This invention relates to the field of ultrasonic equipment technology, and more specifically, to an ultrasonic equipment loading method, ultrasonic equipment, storage medium, and computer program product. Background Technology

[0002] With the development of technology, deep learning models are increasingly being applied to various aspects of medicine. For example, by utilizing deep learning models, the functionality of ultrasound equipment can be enhanced, leading to a greater variety of functionalities available for ultrasound equipment. In practical use, ultrasound equipment can invoke the appropriate deep learning model for ultrasound imaging based on the user's needs for medical diagnosis.

[0003] Each function has a corresponding deep learning model, thus requiring different combinations of deep learning models to be run for different functions. For example, when performing ultrasound imaging on the heart and liver, different functions need to be selected to call the corresponding deep learning models for ultrasound imaging.

[0004] Each deep learning model requires video memory resources from the ultrasound device, and video memory resources are limited. If deep learning models are not loaded properly, the ultrasound imaging efficiency of the device will be reduced. On the other hand, if all deep learning models are loaded every time, the user experience will be extremely poor due to the long waiting time when switching between functions. Summary of the Invention

[0005] The present invention was proposed in view of the above-mentioned problems.

[0006] According to a first aspect of the present invention, a loading method for an ultrasound device is provided. The method includes: in response to a user's first selection operation, determining selected options, wherein all options are organized in a hierarchical structure, at least one first option exists among all options, and the number of next-level options for each first option is greater than one; for each model, determining evaluation parameters for the model based on the selected options and common options among all options related to the model, wherein multiple first models exist among all models, and at least two first models each include different next-level options of the same first option in their respective related options, and each evaluation parameter corresponds to the hierarchy of all options related to the model; for each model, determining the priority of the model based on the evaluation parameters of the model; and loading the at least one model with the highest priority.

[0007] For example, loading at least one model with the highest priority includes: accumulating the cache space required by the models according to the order of priority from high to low, until the accumulated sum is greater than or equal to a first space threshold; when the accumulated sum is greater than the first space threshold, loading all models except the last model among the models whose cache space has been accumulated; when the accumulated sum is equal to the first space threshold, loading all models whose cache space has been accumulated.

[0008] For example, determining the evaluation parameters of the model based on the selected options and the same options among all options related to the model includes: determining the evaluation parameters of the model corresponding to each first level based on the selected options and the same options among the options related to the model at the first level, wherein the first level includes the level in which each of the selected options is located, and for each level in the first level, when the options related to the model at that level do not include any selected options, the evaluation parameter of the model corresponding to that level is determined to be 0.

[0009] For example, the determination of the evaluation parameters of the model is also based on the unselected option among all the options related to the model.

[0010] For example, the method further includes: updating the selected option in response to the user's first selection operation, and updating the loaded model according to the updated option.

[0011] For example, updating the loaded models according to the updated options includes: determining at least one model with the highest current priority among all models according to the updated options; when the determined model is different from the loaded models, unloading the loaded models that are different from the determined models; and loading models other than the ununloaded models among the determined models based on the ununloaded models among the loaded models.

[0012] For example, the method further includes: determining a weight coefficient for each level, wherein the weight coefficient of the previous level is greater than or equal to the weight coefficient of the next level, and the determination of the evaluation parameters of the model is also based on the weight coefficient.

[0013] For example, determining the weight coefficients for each level includes: acquiring multimodal data, which includes the subject's physiological data, the selected options, the operating status of the ultrasound device, and / or user identity information; inputting the multimodal data into a deep learning model to output a probability distribution of the weight coefficients for each level; and determining the weight coefficients for each level based on the probability distribution.

[0014] For example, the determination of the evaluation parameters of the model is also based on the frequency of selection of the model-related options within a preset time period.

[0015] For example, the method further includes updating the selected frequency in response to a user's first selection operation.

[0016] According to a second aspect of the present invention, a loading device for an ultrasonic device is also provided, comprising:

[0017] The option determination module is used to determine the selected option in response to the user's first selection operation, wherein all options are organized in a hierarchical architecture, there is at least one first option among all options, and the number of next-level options of each first option is greater than 1;

[0018] The evaluation module is used to determine the evaluation parameters of each model based on the selected options and the same options among all options related to the model. Among all models, there are multiple first models, and at least two first models each include different next-level options of the same first option among all options related to the model. Each evaluation parameter corresponds to the level of each of the options related to the model.

[0019] The priority determination module is used to determine the priority of each model based on its evaluation parameters.

[0020] The loading module is used to load at least one of the highest priority models.

[0021] According to a third aspect of the present invention, an ultrasonic device is also provided, comprising: a processor and a memory, wherein the memory stores computer program instructions, which, when executed by the processor, are used to perform the above-described loading method of the ultrasonic device.

[0022] According to a fourth aspect of the present invention, a storage medium is also provided, on which program instructions are stored, which, when executed, are used to perform the above-described loading method for an ultrasonic device.

[0023] According to a fifth aspect of the present invention, a computer program product is also provided, comprising computer program instructions that, when executed, perform the above-described loading method for an ultrasonic device.

[0024] In the above technical solution, in response to the user's first selection operation, the selected option is determined. All options are organized in a hierarchical structure, with at least one first option and each first option having more than one sub-option. Then, for each model, evaluation parameters are determined based on the selected option and the common options among all options related to that model. Multiple first models exist, and at least two first models each have different sub-options of the same first option among their respective related options. Each evaluation parameter corresponds to the hierarchy of all options related to that model. Next, for each model, its priority is determined based on its evaluation parameters. Finally, at least one model with the highest priority is loaded. By analyzing the model performance based on the user's selected option and selecting the model that best suits the selected option for loading, the operating efficiency of the ultrasound equipment can be improved, and the user experience optimized.

[0025] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0026] The above and other objects, features, and advantages of the present invention will become more apparent from the more detailed description of the embodiments of the invention in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same parts or steps.

[0027] Figure 1 A schematic flowchart of a loading method for an ultrasonic device according to an embodiment of the present invention is shown;

[0028] Figure 2 A schematic diagram of the hierarchical architecture of options according to an embodiment of the present invention is shown;

[0029] Figure 3 A schematic flowchart illustrating the loading of at least one model with the highest priority according to an embodiment of the present invention is shown;

[0030] Figure 4 A schematic flowchart illustrating the updating of a loaded model according to an updated option is shown according to an embodiment of the present invention;

[0031] Figure 5 A schematic flowchart illustrating the determination of weighting coefficients according to an embodiment of the present invention is shown;

[0032] Figure 6 A schematic diagram of a deep learning model according to an embodiment of the present invention is shown;

[0033] Figure 7 A schematic block diagram of a loading device for an ultrasonic device according to an embodiment of the present invention is shown;

[0034] Figure 8 A schematic block diagram of an ultrasonic device according to an embodiment of the present invention is shown. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of the present invention more apparent, exemplary embodiments according to the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely a part of the embodiments of the present invention, and not all of the embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein. Based on the embodiments of the present invention described herein, all other embodiments obtained by those skilled in the art without inventive effort should fall within the protection scope of the present invention.

[0036] To at least partially address the aforementioned problems, a loading method for ultrasound equipment is proposed. This method responds to a user's first selection operation by determining the selected option, wherein all options are organized in a hierarchical structure, with at least one first option and each first option having more than one sub-option. Then, for each model, evaluation parameters are determined based on the selected option and common options among all options related to that model. Multiple first models exist, and at least two first models each have different sub-options of the same first option among their respective related options. Each evaluation parameter corresponds to the hierarchy of all options related to that model. Next, for each model, its priority is determined based on its evaluation parameters. Finally, at least one model with the highest priority is loaded. By analyzing the model performance based on the user's selected option and selecting the model that best suits the selected option for loading, the efficiency of the ultrasound equipment can be improved, and the user experience optimized.

[0037] Figure 1 A schematic flowchart illustrating a loading method for an ultrasonic device according to an embodiment of the present invention is shown. Figure 1 As shown, the loading method of the ultrasound device may include steps S110 to S140.

[0038] In step S110, the selected options are determined, wherein all options are organized in a hierarchical structure, and there is at least one first option among all options, and the number of next-level options of each first option is greater than 1.

[0039] The selected options can be the default options. For example, when performing an ultrasound examination for the same condition, the options that the doctor needs to select are likely to remain the same. Therefore, some options can be pre-selected so that the doctor can use them directly when performing the ultrasound examination.

[0040] The selected options can also be those chosen by the user through a selection operation. This selection operation can be an operation on the ultrasound device, such as clicking on an option, sliding an option, or entering a command corresponding to the option. Each function of the ultrasound device can correspond to one or more options, and the user controls the ultrasound device to perform the corresponding function by selecting the desired option.

[0041] The options for ultrasound equipment are organized hierarchically, meaning some options are subordinate to others. The options can be determined based on different ultrasound probes, departments, scanned tissue sites, image processing methods, etc., and these options are categorized into different levels. For example, the same ultrasound probe may be used for different functions in different departments; therefore, the option corresponding to the department can be a subordinate option to the option corresponding to the ultrasound probe. Similarly, within the same department, different ultrasound probes may be used for different functions; therefore, the option corresponding to the ultrasound probe can be a subordinate option to the option corresponding to the department. Likewise, the options corresponding to ultrasound probes, departments, scanned tissue sites, and image processing methods can also have a similar hierarchical structure.

[0042] Figure 2 A schematic diagram of the hierarchical architecture of options according to an embodiment of the present invention is shown.

[0043] like Figure 2 As shown, taking the option corresponding to the department as a sub-option of the option corresponding to the ultrasound probe as an example, the first option can include options representing the probe, such as Ultrasound Probe 1, Ultrasound Probe 2, etc. The sub-option of Ultrasound Probe 1 can include at least two options representing the department, such as Diagnosis Item 1, Diagnosis Item 2, Diagnosis Item 3, etc. For example, Diagnosis Item 1 could be Gynecology, Diagnosis Item 2 could be Cardiology, and Diagnosis Item 3 could be Hepatobiliary. Similarly, the first option can also include various diagnostic items. For example, the sub-option of Diagnosis Item 1 can include tabs such as Tab 1, Tab 2, Tab 3, etc., representing the tissue location being scanned. For example, Tab 1 could represent the ovary, Tab 2 could represent the uterus, and Tab 3 could represent the fallopian tube. Similarly, the first option can also include various tabs. For example, the sub-option of Tab 1 can include Measurement Item 1, Measurement Item 2, Measurement Item 3, etc., representing different image processing methods. For example, Tab 1 could represent image segmentation, Tab 2 could represent color rendering, and Tab 3 could represent annotating tissue information.

[0044] At each level, at least one option can be selected, or at least one option can be selected from each level within a subset of levels. Understandably, when multiple options are selected simultaneously at the same level, the functions corresponding to the options may conflict, and the higher the level, the greater the potential conflict. Therefore, at higher levels, only one option can be selected simultaneously, while at lower levels, one or more options can be selected simultaneously. For example... Figure 2 Taking the options shown as an example, in the probe, diagnostic items, and tab levels, users can only select one option at a time. In the measurement item level, users can select multiple options simultaneously, thus enabling the use of various image processing methods for ultrasound imaging. Similarly, the number of options that can be selected simultaneously at each level can be set according to actual needs and is not limited here.

[0045] For example, one or more levels of options can be displayed on the user interface for the user to choose from. When an option at a certain level is selected, the option at the level above that level is also automatically selected. Here, we will continue with... Figure 2 For example, when the user selects measurement item 1, tab 1 will also be automatically selected. Once tab 1 is selected, diagnostic item 1 will also be automatically selected, and similarly, probe 1 will also be automatically selected.

[0046] For example, you can start with the highest level of options and guide the user to select the desired option. For example, still using... Figure 2 For example, you could allow users to select diagnostic items only after they have selected the option corresponding to the probe. Alternatively, you could allow users to select tabs only after they have selected diagnostic items. Or, you could allow users to select measurement items only after they have selected a tab.

[0047] In step S120, for each model, the evaluation parameters of the model are determined based on the selected options and the same options among all options related to the model. Among all models, there are multiple first models, and at least two first models each include different next-level options of the same first option among all options related to the model. Each evaluation parameter corresponds to the level of each of the options related to the model.

[0048] Each model can be a deep learning model suitable for different functions, and each model can correspond to different combinations of options based on its applicable functions. The structure of the deep learning models is not limited here. For example... Figure 2 As shown, options related to Model 1 may include Probe 1, Diagnostic Item 1, Tab 1, and Measurement Item 1. Options related to Model 2 may include Probe 1, Diagnostic Item 1, Tab 1, and Measurement Item 2. For each model, at least one option may be associated with each level. For example, as... Figure 2As shown, options related to model a may include probe 1, diagnostic item 1, tab 1, measurement item 3, and measurement item 4. In this case, model a can be associated with multiple measurement items at the measurement item level. For example, the first model may include model 1, model 2, model 3, and model a. For Figure 2 The other options for the model are similar in principle to those for Model 1, and will not be detailed here. Understandably, Figure 2 The hierarchical architecture, options at each level, and models shown are for illustrative purposes only; the specific settings can be customized according to actual needs.

[0049] For each model, the more identical options there are among the selected options and all options related to that model, the more suitable the model is for performing the function corresponding to the selected option.

[0050] For example, for each model and each level, we can first identify the selected options at that level, and then identify the identical options among the options at that level related to the model. Then, based on the relationship between the identical options and the selected options, we determine the evaluation parameters for the model.

[0051] Optionally, the evaluation parameters of the model can be determined based on the number of identical options and / or the proportion of identical options among the selected options at that level. Alternatively, the evaluation parameter corresponding to that level can be set to a non-zero preset value, such as 1, as long as the number of identical options is not zero. For example, in... Figure 2 In the measurement item hierarchy shown, if the user selects measurement item 1, the evaluation parameter of Model 1 corresponding to the measurement item hierarchy can be 1 or a preset value, or it can be an evaluation parameter further determined based on 1 and other correlation coefficients. If the user selects both measurement item 1 and measurement item 2, the evaluation parameter of Model 1 corresponding to the measurement item hierarchy can be 0.5 or a preset value, or it can be an evaluation parameter further determined based on 0.5 and other correlation coefficients.

[0052] For example, based on the selected options and the same options in the first level related to the model, the evaluation parameters of the model corresponding to each first level are determined, wherein the first level includes the level in which each of the selected options is located. For each level in the first level, when the options in that level related to the model do not include any selected option, the evaluation parameter of the model corresponding to that level is determined to be 0.

[0053] The user's first choice may be at a lower level than the options related to that model. Figure 2Taking Model 1 as an example, users can select only Probe 1 and Diagnostic Item 1, without selecting Tab and Measurement Item. The options related to Model 1 are Probe 1, Diagnostic Item 1, Tab 1, and Measurement Item 1. In this case, the selected options are at fewer levels than the other options related to Model 1. Since Tab and Measurement Item are not selected, the evaluation parameters determined based on Tab and Measurement Item have little impact on whether the model is suitable for executing the functions corresponding to the selected options. Therefore, only the identity of the Probe and Diagnostic Item needs to be used to determine the two evaluation parameters of the model, instead of determining four evaluation parameters. The first level includes the levels of the selected options. For each level in the first level, if the options related to the model at that level do not include any selected option, it means that the model is not suitable for the functions of the selected options at that level. Therefore, when calculating the evaluation parameters, the evaluation parameters corresponding to that level can be ignored, i.e., considered as 0. Determining the first level before calculating the evaluation parameters can speed up the calculation of evaluation parameters, thus allowing for faster model loading based on the evaluation parameters.

[0054] For example, the determination of the evaluation parameters of the model is also based on the unselected option among all the options related to the model.

[0055] Compared to the level to which the selected option belongs, although options at other levels are not selected, users may still need the functionality corresponding to options at other levels but have not made a selection. Therefore, for each model, based on the selected options and related options at other levels, the evaluation parameters of the model can be further determined based on the unselected options. This allows us to consider the user's possible next actions to evaluate each model, and the determined evaluation parameters can better represent the model's performance. For example, option 1 is the first option, and options 2 and 3 are the next level options of option 1. The options related to model x1 are options 1 and 2, and the options related to model x2 are options 1 and 3. Based on the selected option 1, the evaluation parameter for both model x1 and model x2 can be determined as a1. Based on option 2, the evaluation parameter for model x1 can be determined as a2. Based on option 3, the evaluation parameter for model x2 can be determined as a3. At this point, the evaluation parameters for model x1 include a1 and a2, and the evaluation parameters for model x2 are a1 and a3. The evaluation parameters for model x1 and model x2 are highly likely to be different.

[0056] For example, the above loading method may also include step S100: determining the weight coefficient of each level, wherein the weight coefficient of the previous level is greater than or equal to the weight coefficient of the next level, and determining the evaluation parameters of the model based on the weight coefficient.

[0057] The weight coefficients for each level can be preset, meaning each level's weight coefficient is a fixed value. Alternatively, the weight coefficients for each level can be predicted and updated in real time based on the user's selected options. For example, a trained deep learning model can be used to predict the weight coefficients for each level.

[0058] By ensuring that the weight coefficient of the higher level is greater than or equal to the weight coefficient of the lower level, we can avoid the excessive influence of the evaluation parameters of the lower level on the priority of the model and determine a more reasonable priority.

[0059] For example, the evaluation parameters of the model are also determined based on the frequency of selection of the model-related options within a preset time period.

[0060] The preset time period can be a preset duration before the current time, a time period from when the ultrasound device is started to the current time, a time period from when the user first made the selection operation to the current time, or a time period ending at the current start time of the ultrasound device.

[0061] When determining the evaluation parameters for each first level of the model, for each first level, the frequency of selection of the model-related options at that first level within a preset time period can be used as a coefficient to calculate the corresponding evaluation parameters, thereby determining the evaluation parameters for that first level of the model. The calculation methods may include addition, multiplication, etc.

[0062] For options at levels other than the first level, since the user did not select options at those levels, the selection frequency of these model-related options at those levels within the preset time period is not very suitable for evaluating the model's performance. Therefore, the selection frequency of these level options can be ignored when determining evaluation parameters. Alternatively, the selection frequency of model-related options at other levels can be combined to calculate the evaluation parameters for the corresponding other levels of the model. Calculation methods can include addition, multiplication, etc.

[0063] Specifically, for each level, the higher the frequency of selection of the model-related options within a preset time period for that level, the higher the evaluation parameter of the model corresponding to that level.

[0064] Understandably, the frequency with which the model's related options are selected within a preset time period reflects the user's selection habits. Such a determined evaluation parameter can more accurately reflect whether the model's performance can meet the user's usage habits, and further, it can identify and load models that meet the user's expectations.

[0065] For example, the selected frequency can be updated in real time as the selected option changes, or when the object being examined changes, or when the ultrasound equipment is turned off.

[0066] For example, the selected option and the frequency of selection can be updated in response to the user's first selection action.

[0067] The user's first action can be to select options related to the ultrasound device, such as clicking, swiping, or inputting commands corresponding to those options. This first action can include selecting a new option and deselecting an existing one. Since the updated options differ from the previous ones, the selection frequency needs to be updated synchronously. Based on the updated options and selection frequency, more accurate evaluation parameters can be determined in real time, further enabling the identification and loading of a model that meets the user's expectations.

[0068] In step S130, for each model, the priority of the model is determined based on the evaluation parameters of the model.

[0069] All evaluation parameters can be calculated using appropriate data methods, such as multiplication or addition, to determine the model's priority. Once the aforementioned weighting coefficients and / or selection frequencies are determined, the model's evaluation parameters can be updated based on these coefficients and selection frequencies, and the model's priority can be determined according to the updated evaluation parameters.

[0070] by Figure 2 For example, for each model, the priority of the model can be determined according to the following formula 1:

[0071] Lc k =α*TF i *x1+β*ZF j *x2+γ*YF m *x3+σ*CF n *x4 Formula 1

[0072] Among them, Lc k Indicates priority. α indicates whether the option corresponding to the probe associated with this model is selected. When selected, α can be 1 or any other non-zero value; when not selected, α is 0. If no option at the same level as the probe's corresponding option is selected, α can be 1 or any other non-zero value. TF iThis indicates the frequency with which the option corresponding to the model and the probe is selected within a preset time period. x1 represents the weight coefficient of the level of the option corresponding to the probe. β indicates whether the diagnostic item related to the model is selected; β can be 1 or other non-zero values ​​when selected, and β is 0 when not selected. If no option is selected at the level of the diagnostic item, β can be 1 or other non-zero values. ZF j This indicates the frequency with which the diagnostic item corresponding to this model is selected within a preset time period. x2 represents the weight coefficient of the level to which the diagnostic item belongs. γ indicates whether the tab related to this model is selected. When selected, γ can be 1 or other non-zero values; when not selected, γ is 0. If no options are selected under the level of the tab, γ can be 1 or other non-zero values. YF m This indicates the frequency with which the tab corresponding to this model is selected within a preset time period. x3 represents the weight coefficient of the level at which the tab is located. σ indicates whether the measurement item related to this model is selected. When selected, σ can be 1 or other non-zero values; when not selected, σ is 0. If no option is selected under the level of the measurement item, σ can be 1 or other non-zero values. CF n This indicates the frequency with which the corresponding measurement item in the model is selected within a preset time period. x4 represents the weight coefficient of the level to which the measurement item belongs.

[0073] For example, the range of values ​​for each selected frequency can be normalized to 0-1.

[0074] For example, in Formula 1 above, considering that the weight coefficient of the upper level is greater than or equal to the weight coefficient of the lower level, this avoids the excessive influence of the evaluation parameters of the corresponding lower level on the priority of the model, and a more reasonable priority can be determined. Therefore, x1≥x2≥x3≥x4 can be set, for example, x1=1000, x2=100, x3=10, x4=1. x1, x2, x3, and x4 can also be other values.

[0075] In step S140, at least one model with the highest priority is loaded.

[0076] All models can be sorted from highest to lowest priority, and at least one model with the highest priority can be loaded directly. The number of at least one model can be a preset value or a value entered by the user.

[0077] For example, at least one model with the highest priority can be loaded based on relevant parameters of the ultrasound equipment's operating status. These parameters may include the ultrasound equipment's RAM, video memory, and number of cores.

[0078] In the above technical solution, in response to the user's first selection operation, the selected option is determined. All options are organized in a hierarchical structure, with at least one first option and each first option having more than one sub-option. Then, for each model, evaluation parameters are determined based on the selected option and the common options among all options related to that model. Multiple first models exist, and at least two first models each have different sub-options of the same first option among their respective related options. Each evaluation parameter corresponds to the hierarchy of all options related to that model. Next, for each model, its priority is determined based on its evaluation parameters, and finally, at least one model with the highest priority is loaded. By analyzing model performance based on the user's selected options and selecting models more favorable to the selected options for loading, the operating efficiency of the ultrasound equipment can be improved, and the user experience optimized.

[0079] Figure 3 A schematic flowchart illustrating the loading of at least one model with the highest priority according to an embodiment of the present invention is shown. Figure 3 As shown, step S140 may include steps S141 to S143.

[0080] In step S141, the cache space required by the model is accumulated according to the priority from high to low until the accumulated sum is greater than or equal to the first space threshold.

[0081] The cache space required for the model can be the ultrasound device's video memory, RAM, etc. The first space threshold can be a value less than or equal to the cache space of the ultrasound device that can be used to load the model.

[0082] In step S142, when the cumulative sum is greater than the first space threshold, load all models except the last model from the models whose cache space has been accumulated.

[0083] In step S143, when the cumulative sum equals the first space threshold, all models with accumulated cache space are loaded.

[0084] When the cumulative sum exceeds the first spatial threshold, it indicates that the available cache space of the current ultrasound device is insufficient to load all models in the cumulative cache space simultaneously. Therefore, only models up to the last one can be loaded. This avoids insufficient cache space required for the models.

[0085] When the accumulated sum equals the first space threshold, there is still not enough space to load the relevant models. Therefore, at this point, only all models with accumulated cache space can be loaded, and the accumulation of cache space can be stopped.

[0086] In the above technical solution, the cache space required by the model is accumulated according to its priority from high to low until the total accumulated space is greater than or equal to a first space threshold. When the total accumulated space exceeds the first space threshold, all models except the last one in the accumulated cache space are loaded. When the total accumulated space equals the first space threshold, all models in the accumulated cache space are loaded. In this way, models can be loaded reasonably according to the cache space of the ultrasound equipment and the cache space required by each model, avoiding cache space overflow and loading models that are more suitable for the selected options.

[0087] For example, the loading method described above may also include step S150: in response to the user's first selection operation, updating the selected option and updating the loaded model according to the updated option.

[0088] The user's first action can be to select options for the ultrasound device, such as clicking an option, swiping an option, or entering a command corresponding to the option. The first action can include selecting a new option and deselecting an already selected option.

[0089] After the selected option is updated, the currently loaded model may not be the optimal model suitable for the currently selected option. Therefore, based on the updated option and the above steps S120 to S140, at least one updated model that needs to be loaded can be determined and loaded.

[0090] For example, after the selected option is updated, all loaded models can be unloaded, and then at least one updated model can be loaded.

[0091] For example, after the selected option is updated, it is also possible to unload only the models that do not currently need to be loaded, based on the updated at least one model that needs to be loaded.

[0092] In the above technical solution, in response to the user's first selection, the selected option is updated, and the loaded model is updated according to the updated option. This allows for real-time updating and loading of the most suitable model based on the user's selection.

[0093] Figure 4 A schematic flowchart illustrating the updating of a loaded model according to an updated option is shown, based on an embodiment of the present invention. Figure 4 As shown, step S150 may include steps S151 to S153.

[0094] In step S151, based on the updated options, at least one model with the highest current priority is determined among all models.

[0095] After determining the updated options, the priority of each model can be determined according to steps S120 and S130 above, and then at least one model with the highest current priority can be determined.

[0096] In step S152, when the determined model is different from the loaded model, the model that is different from the determined model in the loaded model is unloaded.

[0097] When the determined model differs from the loaded models, it indicates that some of the loaded models are not the optimal models for executing the functions corresponding to the updated options. Therefore, in this case, the models that differ from the determined model can be unloaded, while the models identical to the determined model can be retained.

[0098] In step S153, based on the models that have not been uninstalled from the loaded models, load models other than the models that have not been uninstalled from the determined models.

[0099] After unloading some models, the models that have not been unloaded from the loaded models can remain loaded, and models other than the unloaded models in the determined models can be loaded.

[0100] In the above technical solution, based on the updated options, at least one model with the highest current priority is determined from all models. When the determined model differs from the already loaded models, the loaded models that differ from the determined model are unloaded. Based on the already loaded models that have not been unloaded, models other than the determined models that have not been unloaded are loaded. This reduces the number of models reloaded when the selected options are updated, improving the operating efficiency of the ultrasound equipment.

[0101] Figure 5 A schematic flowchart illustrating the determination of weighting coefficients according to an embodiment of the present invention is shown. Figure 5 As shown, the above step S100 may include steps S101 to S103.

[0102] Step S101: Acquire multimodal data, which includes the subject's physiological data, selected options, the operating status of the ultrasound equipment, and / or user identity information.

[0103] The examinee's physiological data may include age, height, weight, medical records, etc. The operating status of the ultrasound equipment may include its RAM, video memory, storage space, number of cores, and operating time (e.g., 9:00-18:00 could be working hours, 22:00-6:00 could be nighttime). User identification information may include the operator's department, position, experience level (junior, intermediate, senior), and user permissions.

[0104] Step S102: Input multimodal data into the deep learning model to output the probability distribution of the weight coefficients of each layer.

[0105] Multimodal data can be encoded, and the encoded results can be fed into a deep learning model to predict the probability distribution of the weight coefficients at each level. For example, for a subject's medical records, a fixed-length vector can be created, where each number represents whether a certain common disease exists (1 for yes, 0 for no), thus obtaining a vector identifying the subject's medical records. Other multimodal data can be encoded in a similar way.

[0106] Step S103: Determine the weight coefficient for each level based on the probability distribution.

[0107] For each level, while ensuring that the weight coefficient of the previous level is greater than or equal to the weight coefficient of the next level, the weight coefficient of the level with the highest probability can be selected according to the probability distribution.

[0108] Figure 6 A schematic diagram of a deep learning model according to an embodiment of the present invention is shown.

[0109] like Figure 6 As shown, this deep learning model can be a deep learning model incorporating a Long Short-Term Memory (LSTM) artificial neural network. This deep learning model can include a multimodal input layer, an LSM layer, an attention layer, a fully connected layer, and an output layer. The LSM layer can contain multiple LSM layers, such as 2-3 layers, with 128-256 neurons per layer. The multimodal input layer receives the encoded multimodal data. The attention layer utilizes an attention mechanism to help the deep learning model focus on the most relevant historical operations. The fully connected layer maps the output of the LSM layer to the final prediction result. The output layer can use the Sofmax activation function to output the probability distribution of the weight coefficients of each layer.

[0110] Multimodal data can be pre-collected to obtain a training dataset. This training dataset is then input into a deep learning model to obtain predictions of the probability distribution of the weight coefficients. Based on the predictions and the pre-defined probability distribution, a cross-entropy loss function is calculated to obtain the loss value. The deep learning model is then optimized based on this loss value.

[0111] For example, the deep learning model can also be optimized using the Adam optimizer, with an initial learning rate set to 0.001.

[0112] For example, the deep learning model can also be optimized using a learning rate decay strategy, where the learning rate is halved every 10 epochs.

[0113] For example, dropout (ratio 0.5) and L2 regularization can also be applied to prevent overfitting, thereby optimizing the deep learning model.

[0114] For example, an early stopping strategy can also be used, where training of the deep learning model is stopped and optimized if there is no improvement after 5 consecutive epochs on the validation set.

[0115] In the above technical solution, multimodal data is acquired, including the subject's physiological data, selected options, the operating status of the ultrasound equipment, and / or user identity information. This multimodal data is then input into a deep learning model to output the probability distribution of the weight coefficients for each level. Based on the probability distribution, the weight coefficients for each level are determined. This allows for the determination of more suitable weight coefficients based on more information, making the loaded model more aligned with the user's needs.

[0116] To more clearly describe the technical solution of this application, the following embodiment is also provided, assuming that the current video memory (the cache space required by the accumulation model) has a capacity of 1000M. Figure 2 Taking models 1, 2, 3, 4 and 5 as examples, model 1 requires 300M of video memory, model 2 requires 200M of video memory, model 3 requires 500M of video memory, model 4 requires 300M of video memory, and model 5 requires 400M of video memory.

[0117] The selected frequency for probe 1 is TF1 = 0.6, and the selected frequency for probe 2 is TF2 = 0.1.

[0118] The frequency of selection for diagnostic item 1 is ZF1 = 0.2, the frequency of selection for diagnostic item 2 is ZF2 = 0.7, the frequency of selection for diagnostic item 3 is ZF3 = 0.1, and the frequency of selection for diagnostic item 4 is ZF4 = 0.4.

[0119] Tab 1 was selected with a frequency of YF1 = 0.1, Tab 2 with a frequency of YF2 = 0.8, Tab 4 with a frequency of YF4 = 0.5, and Tab 10 with a frequency of YF... 10 =0.9.

[0120] The frequency of selection for measurement item 1 is CF1 = 0.8, the frequency of selection for measurement item 2 is CF2 = 0.1, the frequency of selection for measurement item 3 is CF3 = 0.3, the frequency of selection for measurement item 9 is CF9 = 0.7, and the frequency of selection for measurement item 10 is CF... 10 =0.8.

[0121] The priority of each model can be calculated using Formula 1 above, where x1 = 1000, x2 = 100, x3 = 10, and x4 = 1. When no option is selected, the priorities of the initialized models are as follows:

[0122] The priority of Model 1 is LC1 = 1*TF1*1000 + 1*ZF1*100 + 1*YF1*10 + 1*CF1*1 = 621.8; the priority of Model 2 is LC2 = 1*TF1*1000 + 1*ZF1*100 + 1*YF1*10 + 1*CF2*1 = 621.1; the priority of Model 3 is LC3 = 1*TF1*1000 + 1*ZF1*100 + 1*YF1*10 + 1*CF3*1 = 621.3; the priority of Model 4 is LC4 = 1*TF1*1000 + 1*ZF2*100 + 1*YF4*10 + 1*CF9*1 = 675.7; and the priority of Model 5 is LC5 = 1*TF2*1000 + 1*ZF4*100 + 1*YF1*100 + 1*CF9*1 = 675.7. 10 *10+1*CF 10 *1 = 149.8.

[0123] Therefore, the order is LC4>LC1>LC3>LC2>LC5, meaning Model 4 and Model 1 can be loaded. Since loading Model 3 would require more than 1000MB of video memory, Model 3 is not loaded.

[0124] When the user selects probe 1, the priority of model 5 will change to:

[0125] LC5=0*TF2*1000+1*ZF4*100+1*YF 10 *10+1*ZF 10 *1 = 49.8.

[0126] The priority of other models remains unchanged, the loaded models do not need to be changed, and users do not need to wait for models to load.

[0127] When the user continues to select diagnostic option 1

[0128] The priority of Model 1 is LC1 = 1*TF1*1000 + 1*ZF1*100 + 1*YF1*10 + 1*CF1*1 = 621.8; the priority of Model 2 is LC2 = 1*TF1*1000 + 1*ZF1*100 + 1*YF1*10 + 1*CF2*1 = 621.1; the priority of Model 3 is LC3 = 1*TF1*1000 + 1*ZF1*100 + 1*YF1*10 + 1*CF3*1 = 621.3; the priority of Model 4 is LC4 = 1*TF1*1000 + 0*ZF2*100 + 1*YF4*10 + 1*CF9*1 = 604.7; and the priority of Model 5 is LC5 = 0*TF2*1000 + 0*ZF4*100 + 1*YF1*100 + 1*YF1*1*2*1 = 621.1; 10 *10+1*CF 10 *1 = 9.8.

[0129] At this point, LC1 > LC3 > LC2 > LC4 > LC5. Comparing the loaded Model 1 and Model 4, Model 4 needs to be unloaded, and Model 3 and Model 2 need to be loaded (the total video memory required for the loaded models is exactly 1000MB). At this point, Model 1, Model 2, and Model 3 have been loaded.

[0130] When the user clicks on tab 1 and the next level measurement item, all the required models 1, 2, and 3 have already been loaded, and there is no need to wait.

[0131] When the user returns to select probe B...

[0132] The priority of Model 1 is LC1 = 0*TF1*1000 + 1*ZF1*100 + 1*YF1*10 + 1*CF1*1 = 21.8; the priority of Model 2 is LC2 = 0*TF1*1000 + 1*ZF1*100 + 1*YF1*10 + 1*CF2*1 = 21.1; the priority of Model 3 is LC3 = 0*TF1*1000 + 1*ZF1*100 + 1*YF1*10 + 1*CF3*1 = 21.3; the priority of Model 4 is LC4 = 0*TF1*1000 + 1*ZF2*100 + 1*YF4*10 + 1*CF9*1 = 75.7; and the priority of Model 5 is LC5 = 1*TF2*1000 + 1*ZF4*100 + 1*YF1*10 + 1*CF9*1 = 21.8. 10 *10+1*CF 10 *1 = 149.8.

[0133] At this point, LC5 > LC4 > LC1 > LC2 > LC3. Comparing the loaded models 1, 2, and 3, we need to unload models 2 and 3 and load model 4 (the total video memory required for the loaded models is exactly 1000M).

[0134] In summary, each time a user selects an option, the priority of each model is dynamically calculated, and the corresponding model is loaded based on the priority. The required models can be loaded according to the available video memory, so by the time the user selects a specific measurement item, the required model has already been pre-loaded, eliminating the need to wait for model loading.

[0135] Figure 7 A schematic block diagram of the loading device of an ultrasonic apparatus according to an embodiment of the present invention is shown. Figure 7 As shown, the loading device of the ultrasound equipment includes an option determination module 710, an evaluation module 720, a priority determination module 730, and a loading module 740.

[0136] Option determination module 710 is used to determine the selected option, wherein all options are organized in a hierarchical structure, there is at least one first option among all options, and the number of next-level options of each first option is greater than 1.

[0137] The evaluation module 720 is used to determine the evaluation parameters of each model based on the selected options and the same options among all options related to the model. Among all models, there are multiple first models, and at least two first models each include different next-level options of the same first option among all options related to the model. Each evaluation parameter corresponds to the level of each of the options related to the model.

[0138] The priority determination module 730 is used to determine the priority of each model based on its evaluation parameters.

[0139] Loading module 740 is used to load at least one model with the highest priority.

[0140] For example, the loading module 740 includes an accumulation submodule, a first loading submodule, and a second loading submodule. The accumulation submodule is used to accumulate the cache space required by the models according to the order of priority from high to low, until the accumulated sum is greater than or equal to a first space threshold. The first loading submodule is used to load all models except the last one from the models whose cache space has been accumulated when the accumulated sum is greater than the first space threshold. The second loading submodule is used to load all models whose cache space has been accumulated when the accumulated sum is equal to the first space threshold.

[0141] For example, the evaluation module 720 includes a first evaluation submodule. The first evaluation submodule is used to determine the evaluation parameters of the model corresponding to each of the first levels based on the selected options and the same options related to the model in the first level. The first level includes the level in which each of the selected options is located. For each level in the first level, when the options related to the model in that level do not include any selected option, the evaluation parameter of the model corresponding to that level is determined to be 0.

[0142] For example, the determination of the evaluation parameters of the model is also based on the unselected option among all the options related to the model.

[0143] For example, the loading device of the ultrasound equipment also includes an update module. The update module is used to update the selected option in response to the user's first selection operation, and update the loaded model according to the updated option.

[0144] For example, the update module includes a determination submodule, an unload submodule, and a reload submodule. The determination submodule is used to determine at least one model with the highest current priority among all models based on the updated options. The unload submodule is used to unload the models from the loaded models that are different from the determined model when the determined model is different from the loaded models. The reload submodule is used to load models from the determined models that are not unloaded, based on the unloaded models from the loaded models.

[0145] For example, the loading device of the ultrasound equipment further includes a weight coefficient determination module. The weight coefficient determination module is used to determine the weight coefficient of each level, wherein the weight coefficient of the upper level is greater than or equal to the weight coefficient of the lower level, and the determination of the evaluation parameters of the model is also based on the weight coefficients.

[0146] For example, the weight coefficient determination module includes an acquisition submodule, a prediction submodule, and a calculation submodule. The acquisition submodule acquires multimodal data, including the subject's physiological data, the selected options, the operating status of the ultrasound equipment, and / or user identity information. The prediction submodule inputs the multimodal data into a deep learning model to output a probability distribution of the weight coefficients for each level. The calculation submodule determines the weight coefficients for each level based on the probability distribution.

[0147] For example, the determination of the evaluation parameters of the model is also based on the frequency of selection of the model-related options within a preset time period.

[0148] For example, the loading device of the ultrasound equipment further includes a frequency update module. The frequency update module is used to update the selected frequency in response to a user's first selection operation.

[0149] According to another aspect of the present invention, an ultrasonic device is also provided. Figure 8 A schematic block diagram of an ultrasonic device according to an embodiment of the present invention is shown. Figure 8 As shown, the ultrasound device includes a processor and a memory, wherein the memory stores computer program instructions, which are executed by the processor to perform the loading method of the ultrasound device as described above.

[0150] Furthermore, according to another aspect of the present invention, a storage medium is provided, on which program instructions are stored. When the program instructions are executed by a computer or processor, the computer or processor performs corresponding steps of the loading method of the ultrasound device described in the embodiments of the present invention, and is used to implement corresponding modules in the loading device of the ultrasound device described in the embodiments of the present invention. The storage medium may, for example, include a storage component of a tablet computer, a hard disk of a personal computer, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a portable compact disc read-only memory (CD-ROM), a USB memory, or any combination of the above storage media. The computer-readable storage medium may be any combination of one or more computer-readable storage media.

[0151] According to another aspect of the present invention, a computer program product is also provided, including computer program instructions, which, when executed, are used to perform the above-described loading method for an ultrasonic device.

[0152] Those skilled in the art can understand the specific implementation and beneficial effects of the above-described loading method for ultrasonic equipment by reading the detailed description of the loading device, ultrasonic equipment, storage medium, and computer program product. For the sake of brevity, these details will not be repeated here.

[0153] Although exemplary embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above exemplary embodiments are merely illustrative and are not intended to limit the scope of the invention. Various changes and modifications can be made therein by those skilled in the art without departing from the scope and spirit of the invention. All such changes and modifications are intended to be included within the scope of the invention as claimed in the appended claims.

[0154] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0155] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed.

[0156] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0157] Similarly, it should be understood that, in order to streamline the invention and aid in understanding one or more of the various aspects of the invention, features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof in the description of exemplary embodiments of the invention. However, this approach should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as reflected in the corresponding claims, its inventive point lies in solving the corresponding technical problem with fewer features than all of those in a single disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of the invention.

[0158] Those skilled in the art will understand that, apart from the mutual exclusion of features, all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or apparatus so disclosed can be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0159] Furthermore, those skilled in the art will understand that although some embodiments described herein include certain features but not others included in other embodiments, combinations of features from different embodiments are intended to be within the scope of the invention and form different embodiments. For example, in the claims, any of the claimed embodiments can be used in any combination.

[0160] The various component embodiments of the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some modules in the loading device for an ultrasound device according to embodiments of the present invention. The present invention can also be implemented as an apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such programs implementing the present invention can be stored on a computer-readable medium or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0161] It should be noted that the above embodiments are illustrative of the invention and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0162] The above description is merely a specific embodiment of the present invention or an explanation of that embodiment. The scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A loading method for an ultrasonic device, characterized in that, The method includes: Determine the selected options, wherein all options are organized in a hierarchical structure, and there is at least one first option among all options, and the number of next-level options of each first option is greater than 1; For each model, the evaluation parameters of the model are determined based on the selected options and the same options among all options related to the model. Among all models, there are multiple first models, and at least two first models each include different next-level options of the same first option among all options related to the model. Each evaluation parameter corresponds to the level of each of the options related to the model. For each model, its priority is determined based on its evaluation parameters; Load at least one model with the highest priority.

2. The method according to claim 1, characterized in that, Loading at least one model with the highest priority includes: Based on the priority from high to low, the cache space required by the model is accumulated until the accumulated sum is greater than or equal to the first space threshold. When the cumulative sum exceeds the first space threshold, load all models except the last one from the models whose cache space has been accumulated. When the cumulative sum equals the first space threshold, load all models that have accumulated cache space.

3. The method according to claim 1, characterized in that, The step of determining the evaluation parameters of the model based on the selected options and the same options among all options related to the model includes: Based on the selected options and the same options in the first level related to the model, the evaluation parameters of the model corresponding to each first level are determined. The first level includes the level in which each selected option is located. For each level in the first level, when the options related to the model in that level do not include any selected option, the evaluation parameter of the model corresponding to that level is determined to be 0.

4. The method according to claim 1, characterized in that, The determination of the evaluation parameters for the model is also based on the unselected options among all the options related to the model.

5. The method according to claim 1, characterized in that, The method further includes: In response to the user's first choice, update the selected option and update the loaded model according to the updated option.

6. The method according to claim 5, characterized in that, The step of updating the loaded model according to the updated options includes: Based on the updated options, determine at least one model with the highest current priority among all models; When the determined model is different from the loaded model, the model that is different from the determined model in the loaded model will be unloaded. Based on the models that have not been uninstalled from the already loaded models, load models other than the models that have not been uninstalled from the determined models.

7. The method according to claim 8, characterized in that, The method further includes: The weight coefficients for each level are determined, wherein the weight coefficient of the previous level is greater than or equal to the weight coefficient of the next level, and the evaluation parameters of the model are also determined based on the weight coefficients.

8. The method according to claim 7, characterized in that, The determination of the weight coefficients for each level includes: Acquire multimodal data, which includes the subject's physiological data, the selected options, the operating status of the ultrasound equipment, and / or user identity information; The multimodal data is input into a deep learning model to output the probability distribution of the weight coefficients for each layer; Based on the probability distribution, the weight coefficients for each level are determined.

9. The method according to any one of claims 1-8, characterized in that, The evaluation parameters for the model are also determined based on the frequency of selection of the model-related options within a preset time period.

10. The method according to claim 9, characterized in that, The method further includes: In response to the user's first selection action, the selected frequency is updated.

11. An ultrasonic device, comprising a processor and a memory, characterized in that, The memory stores computer program instructions, which, when executed by the processor, are used to perform the loading method of the ultrasound device as described in any one of claims 1 to 10.

12. A storage medium on which program instructions are stored, characterized in that, The program instructions, when executed, are used to perform the loading method of the ultrasound device as described in any one of claims 1 to 10.

13. A computer program product comprising computer program instructions, characterized in that, The computer program instructions, when executed, are used to perform the loading method of the ultrasonic device as described in any one of claims 1 to 10.

Images