Data processing method and apparatus, health assessment method and apparatus, electronic device and readable storage medium

Abstract

The present disclosure provides a data processing method and apparatus, a health assessment method and apparatus, an electronic device and a readable storage medium. The data processing method comprises the following steps: acquiring vital signs data of a target subject, wherein the vital signs data comprises monitoring data obtained by monitoring vital signs of the target subject and completion data for completing the vital signs data on the basis of the monitoring data; collecting multimodal data of the target subject, wherein the multimodal data comprises at least one of image data of the target subject and research data for a preset symptom; and on the basis of the vital signs data and the multimodal data, generating physical state data of the target subject.

Description

Data processing method and device, health assessment method and device, electronic device and readable storage medium

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This disclosure claims priority to Chinese patent application No. 202310752854.4 filed in China on June 25, 2023, the entire contents of which are incorporated herein by reference. Technical Field

[0003] The embodiments of the present disclosure relate to the technical field of disease diagnosis, and in particular to a data processing method and device, a health assessment method and device, an electronic device, and a readable storage medium. Background Art

[0004] With the development of computer technology, related technologies can collect user's physical data and analyze and predict the user's disease status or disease probability based on the collected results.

[0005] Summary of the Invention

[0006] Embodiments of the present disclosure provide a data processing method and device, a health assessment method and device, an electronic device, and a readable storage medium.

[0007] In a first aspect, an embodiment of the present disclosure provides a data processing method, comprising the following steps:

[0008] Acquiring vital sign data of a target object, wherein the vital sign data includes monitoring data obtained by monitoring the vital signs of the target object and supplemented data obtained by supplementing the vital sign data based on the monitoring data;

[0009] Collecting multimodal data of the target object, wherein the multimodal data includes at least one of image data of the target object and survey data for preset symptoms;

[0010] The body state data of the target object is generated according to the vital sign data and the multimodal data.

[0011] In some embodiments, obtaining the target object's vital sign data includes:

[0012] Acquiring monitoring data obtained by monitoring the vital signs of the target object;

[0013] Dividing the monitoring data into boxes;

[0014] Grouping the monitoring data after binning according to a preset monitoring period;

[0015] Generating supplementary data to supplement the missing monitoring data in each monitoring cycle;

[0016] The monitoring data and the supplemented data are used as the vital sign data of the target object.

[0017] In some embodiments, binning the monitoring data includes:

[0018] The monitoring data is binned according to the degree of impact of the monitoring data on the target disease.

[0019] In some embodiments, binning the monitoring data includes:

[0020] The monitoring data is binned using the minimum entropy binning method.

[0021] In some embodiments, after grouping the binned monitoring data according to a preset monitoring period, the method further includes:

[0022] detecting the number of first cycles in the monitoring cycles, wherein the first cycles are monitoring cycles in which monitoring data at the target time is missing;

[0023] When the number of the first cycles is greater than the preset number threshold, the monitoring data at the target moment of the first cycle is determined based on the monitoring data at the target moment of the second cycle, wherein the second cycle is a monitoring period in which the monitoring data at the target moment is not missing.

[0024] In some embodiments, generating the supplementary data to supplement the missing monitoring data in each monitoring period includes:

[0025] Complementary data for the monitoring data is generated by cubic spline interpolation.

[0026] In some embodiments, collecting multimodal data of the target object includes:

[0027] Pushing a questionnaire targeting the preset symptoms to the target subject;

[0028] receiving survey data input by the target subject in response to the questionnaire, wherein the survey data includes a selection input of at least one option from a plurality of options for each question set for the preset symptoms;

[0029] The survey data are standardized according to preset rules to form multimodal data, wherein the standardized survey data are used as external variables of a health assessment model.

[0030] In a second aspect, the present disclosure also provides a health assessment method, comprising the following steps:

[0031] Acquiring physical status data of the target object, wherein the physical status data is obtained by the data processing method described in any one of the first aspects;

[0032] The physical condition data is input into a health assessment model to obtain a health assessment result of the target object suffering from the target disease, wherein the health assessment model is a pre-trained model that takes the physical condition data as input and the probability of suffering from the target disease as output.

[0033] In some embodiments, the health assessment model includes an integrated first model, a second model, and a third model, wherein the first model is an ARIMA model, the second model is an Informer model, and the third model is an N-BeatXs model.

[0034] In some embodiments, the input data of the first model and the second model include the vital sign data;

[0035] The input data of the third model includes the vital sign data and the multimodal data.

[0036] In some embodiments, the health assessment model further includes a fourth model, which is a model that takes the resampled vital sign data as input and takes the probability of suffering from the target disease as output.

[0037] In some embodiments, the fourth model is an Informer model.

[0038] In some embodiments, obtaining a health assessment result of the target subject suffering from the target disease using the health assessment model includes:

[0039] Resampling the vital sign data into high-frequency data and low-frequency data, wherein the sampling frequency of the high-frequency data is greater than the sampling frequency of the low-frequency data, and the sampling frequency of the low-frequency data is no less than twice in each monitoring cycle;

[0040] Performing difference processing on the low-frequency data according to the sampling frequency of the high-frequency data;

[0041] The high-frequency data and the low-frequency data subjected to difference processing are respectively input into the fourth model to obtain a prediction result of the periodic trend of the vital sign data.

[0042] In some embodiments, inputting the physical status data into a health assessment model to obtain a health assessment result of the target subject suffering from the target disease includes:

[0043] The output results of the first model, the second model, the third model and the third model are spliced ​​together according to time, and then input into the fifth model for integrated training to obtain the health assessment model.

[0044] In some embodiments, the fifth model is a LightGBM model.

[0045] In some embodiments, the target disease is chronic obstructive pulmonary disease.

[0046] In a third aspect, an embodiment of the present disclosure further provides a data processing device, including:

[0047] a vital sign data acquisition module, configured to acquire vital sign data of a target object, wherein the vital sign data includes monitoring data obtained by monitoring the vital signs of the target object and supplemented data obtained by supplementing the vital sign data based on the monitoring data;

[0048] a multimodal data acquisition module, configured to acquire multimodal data of the target object, wherein the multimodal data includes at least one of image data of the target object and survey data on preset symptoms;

[0049] A body state data generating module is used to generate the body state data of the target object according to the vital sign data and the multimodal data.

[0050] In a fourth aspect, an embodiment of the present disclosure further provides a health assessment device, comprising:

[0051] a body state data acquisition module, configured to acquire body state data of a target subject, wherein the body state data is obtained by the data processing method described in any one of the first aspects;

[0052] A health assessment module is used to input the physical condition data into a health assessment model to obtain a health assessment result of the target subject suffering from the target disease, wherein the health assessment model is a pre-trained model that takes the physical condition data as input and the probability of suffering from the target disease as output.

[0053] In a fifth aspect, an embodiment of the present disclosure further provides an electronic device, comprising: a memory, a processor, and a program stored in the memory and executable on the processor; the processor is configured to read the program in the memory to implement the steps of any one of the methods described in the first aspect;

[0054] Or implement the steps in any one of the methods as described in the second aspect.

[0055] In a sixth aspect, an embodiment of the present disclosure further provides a readable storage medium for storing a program, wherein when the program is executed by a processor, the steps in the method according to any one of the first aspects are implemented;

[0056] Or implement the steps in any one of the methods as described in the second aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following briefly introduces the drawings required for use in the description of the embodiments of the present disclosure. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0058] FIG1 is a flow chart of a data processing method according to an embodiment of the present disclosure;

[0059] FIG2A is a schematic diagram of a questionnaire provided by an embodiment of the present disclosure;

[0060] FIG2B is a schematic diagram of another questionnaire provided by an embodiment of the present disclosure;

[0061] FIG2C is a schematic diagram of another questionnaire provided by an embodiment of the present disclosure;

[0062] FIG3 is a schematic diagram of a health assessment process according to an embodiment of the present disclosure;

[0063] FIG4 is a schematic diagram of a model architecture provided by an embodiment of the present disclosure;

[0064] FIG5 is a schematic diagram of another model architecture provided by an embodiment of the present disclosure;

[0065] FIG6 is a schematic diagram of another model architecture provided by an embodiment of the present disclosure;

[0066] FIG7 is a schematic structural diagram of a data processing device provided by an embodiment of the present disclosure;

[0067] FIG8 is a schematic structural diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0068] The following will clearly and completely describe the technical solutions in the embodiments of the present disclosure in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present disclosure without making any creative efforts shall fall within the scope of protection of the present disclosure.

[0069] The terms "first", "second" etc. in the disclosed embodiments are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. In addition, the terms "include" and "have" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or equipment comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or that are inherent to these processes, methods, products or equipment. In addition, "and / or" is used in this application to represent at least one of the connected objects, for example A and / or B and / or C, indicating the presence of 7 situations including single A, single B, single C, and both A and B, both B and C, both A and C, and both A, B and C.

[0070] An embodiment of the present disclosure provides a data processing method.

[0071] In some embodiments, the data processing method is used to collect and process the collected physical status data of the user to obtain more comprehensive and complete data.

[0072] Taking patients with chronic obstructive pulmonary disease (COPD) as an example, theoretically, if you need to obtain the patient's complete physical condition data, you need to monitor the patient's physical condition for several consecutive days for 24 hours. The data that needs to be monitored include dynamic electrocardiogram, continuous blood sugar, blood oxygen saturation and other different data. However, this kind of continuous monitoring for several consecutive days is difficult to achieve. Therefore, in actual operation, physical signs are usually monitored during specific time periods.

[0073] For example, blood oxygen saturation might be measured for ten minutes at 8 a.m., half an hour at 3 p.m., and continuously monitored for four to five hours at night. If data is sampled at one-minute intervals, 10 data points can be obtained at 8 a.m., 30 data points at 3 p.m., and 240 to 300 data points at night. However, when a patient is feeling unwell, they may need to monitor changes in their physical condition at any time, and these irregular, intermittent data cannot be used directly.

[0074] As shown in FIG1 , in one embodiment of the present disclosure, the data processing method includes the following steps:

[0075] Step 101: Obtain vital sign data of a target object.

[0076] In the technical solution of this embodiment, the vital sign data includes two parts: monitoring data and supplementary data, wherein the monitoring data refers to the monitoring of the vital sign data of the target object (for example, a patient) according to certain measurement conditions. For example, the monitoring time can be set to monitor the patient's physical condition at a specified time or time period every day. For example, the blood oxygen saturation can be measured for 10 minutes at 8 o'clock in the morning, the blood oxygen saturation can be measured for 30 minutes at 3 o'clock in the afternoon, and the blood oxygen saturation can be continuously monitored for 4 to 5 hours at night. The specific items of monitoring data may include blood oxygen saturation, heart rate, dynamic electrocardiogram, continuous blood glucose, etc. During implementation, the specific items to be detected can be determined by professional doctors and other professionals based on the type of disease to be detected or predicted.

[0077] During implementation, when the target object is undergoing quality and normal respiratory sleep monitoring, monitoring data of the target object can be collected using detection instruments such as a ventilator, a pulmonary function meter, a blood oximeter, a respiratory detector, etc., and then the data is collected and processed to form the monitoring data of the target object.

[0078] After obtaining the monitoring data obtained by monitoring the target object, the obtained monitoring data is supplemented to obtain supplemented data to improve the completeness of the obtained vital sign data. It should be understood that continuous monitoring of the vital signs of the target object is difficult to implement in practice. In this embodiment, by supplementing the monitoring data to obtain supplemented data, and using the monitoring data and the monitoring data as the vital sign data of the target object, it is possible to obtain more accurate and comprehensive physical data of the target object through limited monitoring means and monitoring time, thereby improving the data collection effect.

[0079] In some embodiments, the step of completing the data further includes:

[0080] Acquiring monitoring data obtained by monitoring the vital signs of the target object;

[0081] Dividing the monitoring data into boxes;

[0082] Grouping the monitoring data after binning according to a preset monitoring period;

[0083] Generating supplementary data to supplement the missing monitoring data in each monitoring cycle;

[0084] The monitoring data and the supplemented data are used as the vital sign data of the target object.

[0085] In this embodiment, by binning the monitoring data and then grouping them according to the monitoring period, and further completing the monitoring data in each monitoring period after binning, it can be ensured that the data in each bin has a high correlation when the data is completed, thereby improving the accuracy of the completed data.

[0086] In this embodiment, the monitoring data is first binned. In some embodiments, the steps of binning the monitoring data specifically include:

[0087] The monitoring data is binned according to the degree of impact of the monitoring data on the target disease.

[0088] In the technical solution of this embodiment, the monitoring data is obtained by monitoring the target subject's vital signs. Different monitoring results may reflect different likelihoods of the target subject suffering from the target disease. For example, taking COPD as an example, under the same other conditions, the lower the target subject's blood oxygen concentration and the higher the heart rate, the higher the likelihood of the disease and the higher the symptom level.

[0089] In order to distinguish such influences, in this embodiment, the monitoring data is binned according to the interval size of the influence of different monitoring data on the target disease. It can be understood that, under the same other conditions, if the difference in the heart rates of the two target subjects is 1, it may be a normal fluctuation and has no medical or clinical reference significance. If the difference reaches 5 or 10, it may have certain medical or clinical reference significance. Therefore, in order to distinguish different monitoring data, in this embodiment, the monitoring data can be binned according to the medical or clinical reference significance of the difference in the monitoring results for the influence of the target disease. It can be understood that if a certain interval corresponding to the heart rate has the same level for the target disease, these monitoring data will be divided into data within one bin.

[0090] In this way, by binning the monitoring data according to the degree of its impact on the target disease, the monitoring data can be accurately divided, and the subsequent completion of the binned data can improve the accuracy of the obtained data and improve the diagnostic effect of the target disease.

[0091] Furthermore, in one embodiment, the monitoring data may be binned using a minimum entropy binning method.

[0092] In this embodiment, the binning results of the minimum entropy binning method need to meet the following requirements:

[0093] Among them, the dependent variable is a categorical variable and can take values ​​1, 2, 3...J, p ij is the proportion of observations in the i-th bin whose dependent variable takes value j, where i = 1, 2, 3, ... K and j = 1, 2, 3, ... J.

[0094] During implementation, the monitoring data set is first sorted and counted. Next, the monitoring data is split into two, and the entropies of the two subsets are calculated and summed. Next, the split point with the smallest sum of conditional entropies is selected. For each subset obtained, the sorting, splitting, and conditional entropy summing steps are repeated until the terminating condition is met.

[0095] In this embodiment, the condition for stopping the iteration can be determined based on the advice of a doctor or other professional. For example, if the doctor indicates that a heart rate difference of less than 5 times is clinically significant, the condition for stopping the iteration can be set to that the interval length of the obtained subset is less than or equal to 5. In this way, this embodiment uses the minimum entropy binning method to bin the monitoring data, which can more accurately extract the differences between the monitoring data, thereby obtaining a more accurate binning result for the monitoring data, which helps to improve the accuracy of subsequent data completion.

[0096] After the data is binned, it is divided into multiple groups according to a certain monitoring period. Generally speaking, the selection of the monitoring period needs to have certain practical significance. Therefore, in this embodiment, each day is used as a monitoring period. In this way, under normal circumstances, within each monitoring period, the vital signs of the target object change from morning to night according to a certain pattern.

[0097] In some embodiments, after grouping the binned monitoring data according to a preset monitoring period, the method further includes:

[0098] detecting the number of first cycles in the monitoring cycles, wherein the first cycles are monitoring cycles in which monitoring data at the target time is missing;

[0099] When the number of the first cycles is greater than the preset number threshold, the monitoring data at the target moment of the first cycle is determined based on the monitoring data at the target moment of the second cycle, wherein the second cycle is a monitoring period in which the monitoring data at the target moment is not missing.

[0100] In this embodiment, the number of monitoring cycles with missing data is further detected. For example, if, among the multiple monitoring cycles obtained, most of the monitoring cycles include heart rate data at 8 o'clock, but the heart rate data at 8 o'clock is missing in several of the detection cycles, the monitoring cycle with missing heart rate data at 8 o'clock is defined as the first cycle, and the monitoring cycle with no missing heart rate data at 8 o'clock is defined as the second cycle.

[0101] Next, determine the number of first cycles. If the number of first cycles is less than the preset number threshold, for example, there is only one first cycle, then the absence of heart rate data at 8 o'clock in the first cycle is considered to be accidental. If the number of first cycles is large, then it is considered that there is an abnormality in the number of first cycles. At this time, it is necessary to compensate for the monitoring data of the target moment in the first cycle. During implementation, the monitoring data of the target moment in the first cycle can be compensated by weighted averaging based on the monitoring data of the target moment in the second cycle.

[0102] In some embodiments, generating the supplementary data to supplement the missing monitoring data in each monitoring period includes:

[0103] Complementary data for the monitoring data is generated by cubic spline interpolation.

[0104] In this embodiment, when filling in the missing data, it should be understood that the changes in human vital signs data are smooth. Therefore, under normal circumstances, the human vital signs data will not undergo sudden changes, that is, when reflected on the curve, the corresponding human vital signs data curve will not have a sharp inflection point. Therefore, linear interpolation is not used for data completion in this embodiment.

[0105] As the order of polynomial interpolation increases, the amount of calculation increases greatly, and oscillation (Runge phenomenon) may occur near the endpoints. Therefore, this embodiment adopts a piecewise low-order interpolation method. Specifically, cubic spline interpolation is used to construct multiple cubic functions from several segments in the original monitoring data sequence. This makes the connection between each segment have the property of continuous second-order derivatives, which can be smoothly connected, balancing the amount of calculation and improving the accuracy of the completed data.

[0106] It should be understood that, before data completion, the acquired monitoring data may be data accumulated over a long period of time. For example, it may be multiple segments of data spanning several months or years, with each segment separated by the same time period (e.g., 1 minute). Therefore, in this embodiment, after completion, the data between two adjacent segments will also be separated by this time period, forming a continuous curve.

[0107] During implementation, the monitoring data of each target object can be used as a sample, and the value of the blood oxygen saturation and other monitoring indicators can be used as labels to form a data set of vital sign data corresponding to each target object.

[0108] Step 102: Collect multimodal data of the target object.

[0109] Next, in this embodiment, multimodal data of the target object is extracted. In one embodiment, the multimodal data includes at least one of image data of the target object and survey data for preset symptoms.

[0110] In this embodiment, the image data refers to analysis results of detection images obtained by various imaging detection methods such as CT (Computed Tomography), X-ray, and magnetic resonance imaging.

[0111] Taking the analysis results of CT images as an example, clinically, COPD is generally divided into four levels. Therefore, in this embodiment, the analysis results of CT images are set to five types, corresponding to negative (no disease) and four different levels of disease. In an exemplary embodiment, the relatively mature ResNet151 model can be used to identify and process the obtained CT images. Its output range is defined as 0 to 4, where 0 corresponds to negative and 1 to 4 correspond to four different levels of disease. In this way, the image data of the target object can be obtained.

[0112] Survey data can be developed using questionnaires tailored to different target diseases. For example, as shown in Figures 2A to 2C, in this embodiment, a COPD screening questionnaire (as shown in Figure 2A), a STOP-Bang questionnaire for determining risk factors for obstructive sleep apnea, and an OSAS symptom collection form for determining risk factors for obstructive sleep apnea syndrome (OSAS) are provided for COPD. During implementation, professionals such as doctors can develop questionnaires tailored to different pre-defined symptoms to generate survey data tailored to these pre-defined symptoms.

[0113] As shown in Figures 2A to 2C, in this embodiment, a questionnaire can be pushed to the target subject's account. The target subject then logs in to their account on a mobile device such as a mobile phone or personal computer, or a diagnosis and treatment terminal at a hospital or healthcare facility, and then completes the questionnaire. In other embodiments, corresponding equipment can also be configured to collect relevant data, such as using a sound sensor to monitor snoring volume. Furthermore, a nurse or other monitoring device can monitor the target subject's sleep status while they are asleep and fill in the corresponding data.

[0114] The questionnaire results of the target object are standardized according to certain rules and sent to the host side as multimodal data corresponding to the target object.

[0115] During implementation, the quantitative indicators of the survey results based on the questionnaires filled out by the target subjects are used as research data.

[0116] Step 103: Generate physical status data of the target object according to the vital sign data and the multimodal data.

[0117] After the vital sign data and the multimodal data are determined, the obtained vital sign data and the multimodal data are used as the body state data of the target object.

[0118] The disclosed embodiment also provides a health assessment method.

[0119] As shown in FIG3 , in one embodiment, the health assessment method includes the following steps:

[0120] Step 301: Obtain the target object's physical condition data.

[0121] In this embodiment, the physical state data of the target object is first obtained, wherein the physical state data is obtained by the above-mentioned data processing method, which will not be described in detail here.

[0122] Step 302: Input the physical condition data into a health assessment model to obtain a health assessment result of the target subject suffering from the target disease, wherein the health assessment model is a pre-trained model that takes the physical condition data as input and the probability of suffering from the target disease as output.

[0123] In this embodiment, by setting up a pre-trained health assessment model, it is possible to quickly and easily obtain a prediction result of the target object's illness based on the input physical data, reducing manual participation and saving human resources. It can accurately and quickly generate a preliminary prediction result of the target object's illness, and can provide important reference information for the target object to seek medical treatment or doctor's diagnosis.

[0124] As shown in Figure 4, in some embodiments, the health assessment model includes an integrated first model, a second model, and a third model. In some embodiments, the first model is an ARIMA model, the second model is an Informer model, and the third model is an N-BeatXs model. In this embodiment, the input data of the first and second models includes vital sign data. As shown in Figure 5, the input data of the third model includes vital sign data and multimodal data. Furthermore, the multimodal data includes imaging data and survey data.

[0125] It should be understood that the ARIMA model is a time series model in traditional machine learning. It is effective for predicting short-term weak stationary series. The Informer model is better for long-term prediction results. It is based on the characteristics of time series and adds self-attention distillation mechanism and generative decoder optimization schemes on the basis of the Transformer model, which greatly improves the complexity and efficiency of time series prediction. In addition, in this embodiment, the data of quantitative indicators is input into the N-BeatsX model. The N-BeatsX model can input time series data and external variables at the same time. Because from a medical perspective, the impact of external characteristics on disease is very important, the N-BeatXs model can capture the influence of other factors.

[0126] In the technical solution of this embodiment, multiple integrated models are set up to process the physical status data of the target object, and the different characteristics of different models are used to obtain more accurate health assessment results for the target disease.

[0127] As shown in Figure 4, in some embodiments, the health assessment model further includes a fourth model, which takes the resampled vital sign data as input and outputs the probability of the target disease. In some embodiments, the fourth model is an Informer model.

[0128] In some embodiments, obtaining a health assessment result of the target subject suffering from the target disease using the health assessment model includes:

[0129] Resampling the vital sign data into high-frequency data and low-frequency data, wherein the sampling frequency of the high-frequency data is greater than the sampling frequency of the low-frequency data, and the sampling frequency of the low-frequency data is no less than twice in each monitoring cycle;

[0130] Performing difference processing on the low-frequency data according to the sampling frequency of the high-frequency data;

[0131] The high-frequency data and the low-frequency data subjected to difference processing are respectively input into the fourth model to obtain a prediction result of the periodic trend of the vital sign data.

[0132] In this embodiment, the monitoring data is resampled into low-frequency data and high-frequency data, and the resampled monitoring data are respectively input into the fourth model.

[0133] As shown in FIG6 , in this embodiment, the fourth model specifically includes a fourth model A for processing low-frequency data and a fourth model B for processing high-frequency data.

[0134] To enable the low-frequency data model to analyze periodic changes, according to Shannon's theorem, low-frequency data should be sampled at a frequency of at least twice a day to capture periodic changes. After downsampling to obtain the low-frequency data, the low-frequency data is interpolated using the same frequency as the high-frequency data to be added to the high-frequency data. In this embodiment, piecewise linear interpolation is specifically selected.

[0135] In this embodiment, the outputs of the fourth model are all classification probabilities of five categories, corresponding to 0 to 4 mentioned above. The low-frequency data and the high-frequency data are added through the model respectively and then calculated. The loss is calculated using the cross-entropy loss function. During optimization, the loss is simultaneously passed back to the fourth model A corresponding to the low-frequency data and the fourth model B corresponding to the high-frequency data.

[0136] In some embodiments, inputting the physical status data into a health assessment model to obtain a health assessment result of the target subject suffering from the target disease includes:

[0137] The output results of the first model, the second model, the third model and the fourth model are spliced ​​together according to time, and then input into the fifth model for integrated training to obtain the health assessment model.

[0138] In some embodiments, the fifth model is a LightGBM model. In this embodiment, the output results of the first model, the second model, the third model and the fourth model are finally input into the fifth model for integrated training, and the model obtained by training is used as a health assessment model as a whole.

[0139] In some embodiments, during the integrated training process, the parameters of the first model, the second model, the third model, and the fourth model are frozen so that they no longer change. Then, the original data are input in batches, and the output data of these models are spliced ​​and given to the LightGBM model. The output is the probability of the five categories from 0 to 4 mentioned above. Cross entropy is used as the loss function, and SGD stochastic gradient descent is used as the optimization method for integrated training to obtain a health assessment model.

[0140] In the technical solution of this embodiment, the output results of the generated health assessment model can be five levels of health assessment results from 0 to 4, where 0 represents negative, meaning no disease, and 1 to 4 represent different levels of disease severity.

[0141] During implementation, the above-mentioned disease level can be pushed to users in the form of examination results, diagnosis sheets, etc. In order to facilitate users to understand their disease level, a detailed introduction or description of the disease level can be added to the push results, and corresponding medical advice can be pushed at the same time, such as adding various additional information such as eating habits, living habits, and smoking cessation suggestions.

[0142] In other embodiments, the prediction results may also be sent to professionals, such as doctors, who will refer to the prediction results to form medical advice or treatment results, and push them to the target object together with additional information such as corresponding treatment methods.

[0143] The embodiment of the present disclosure also provides a data processing device.

[0144] As shown in FIG7 , in one embodiment, the data processing device 700 includes:

[0145] A vital sign data acquisition module 701 is configured to acquire vital sign data of a target object, wherein the vital sign data includes monitoring data obtained by monitoring the vital signs of the target object and supplemented data obtained by supplementing the vital sign data based on the monitoring data;

[0146] A multimodal data acquisition module 702 is configured to acquire multimodal data of the target object, wherein the multimodal data includes at least one of image data of the target object and survey data on preset symptoms;

[0147] The body state data generating module 703 is configured to generate the body state data of the target object according to the vital sign data and the multimodal data.

[0148] In some embodiments, the vital sign data acquisition module 701 includes:

[0149] An acquisition submodule, configured to acquire monitoring data obtained by monitoring the vital signs of the target object;

[0150] A binning module, used for binning the monitoring data;

[0151] A grouping submodule, configured to group the monitoring data after binning according to a preset monitoring period;

[0152] A completion submodule, configured to generate completion data to complete the missing monitoring data in each monitoring cycle;

[0153] The vital sign data confirmation submodule is used to use the monitoring data and the supplemented data as the vital sign data of the target object.

[0154] In some embodiments, the binning module is specifically used to bin the monitoring data according to the degree of impact of the monitoring data on the target disease.

[0155] In some embodiments, the binning module is specifically used to bin the monitoring data using a minimum entropy binning method.

[0156] In some embodiments, further comprising:

[0157] a monitoring period detection module, configured to detect the number of first periods in the monitoring period, wherein the first period is a monitoring period in which monitoring data at a target time is missing;

[0158] A monitoring data determination module is used to determine the monitoring data of the target moment of the first cycle based on the monitoring data of the target moment of the second cycle when the number of the first cycles is greater than a preset number threshold, wherein the second cycle is a monitoring period in which the monitoring data at the target moment is not missing.

[0159] In some embodiments, the completion submodule is specifically configured to generate completion data for the monitoring data by cubic spline interpolation.

[0160] In some embodiments, the multimodal data acquisition module 702 includes:

[0161] A questionnaire push submodule is used to push a questionnaire targeting the preset symptoms to the target subject;

[0162] a survey data receiving submodule, configured to receive survey data input by the target subject in response to the questionnaire, wherein the survey data includes a selection input of at least one option from among a plurality of options for each question set for the preset symptoms;

[0163] The multimodal data generation submodule is used to standardize the survey data according to preset rules to form multimodal data, wherein the standardized survey data is used as an external variable of the health assessment model.

[0164] The data processing device 700 of this embodiment can implement each step of the above-mentioned data processing method embodiment and can achieve basically the same technical effects, which will not be described in detail here.

[0165] The present disclosure also provides a health assessment device, including:

[0166] a body state data acquisition module, configured to acquire body state data of a target subject, wherein the body state data is obtained by the data processing method described in any one of the first aspects;

[0167] A health assessment module is used to input the physical condition data into a health assessment model to obtain a health assessment result of the target subject suffering from the target disease, wherein the health assessment model is a pre-trained model that takes the physical condition data as input and the probability of suffering from the target disease as output.

[0168] In some embodiments, the health assessment model includes an integrated first model, a second model, and a third model, wherein the first model is an ARIMA model, the second model is an Informer model, and the third model is an N-BeatXs model.

[0169] In some embodiments, the input data of the first model and the second model include the vital sign data;

[0170] The input data of the third model includes the vital sign data and the multimodal data.

[0171] In some embodiments, the health assessment model further includes a fourth model, which is a model that takes the resampled vital sign data as input and takes the probability of suffering from the target disease as output.

[0172] In some embodiments, the fourth model is an Informer model.

[0173] In some embodiments, the health assessment module includes:

[0174] a resampling submodule, configured to resample the vital sign data into high-frequency data and low-frequency data, wherein the sampling frequency of the high-frequency data is greater than the sampling frequency of the low-frequency data, and the sampling frequency of the low-frequency data is no less than twice in each monitoring cycle;

[0175] A difference processing submodule, configured to perform difference processing on the low-frequency data according to the sampling frequency of the high-frequency data;

[0176] The input submodule is used to input the high-frequency data and the low-frequency data after difference processing into the fourth model respectively to obtain the prediction result of the periodic trend of the vital sign data.

[0177] In some embodiments, the health assessment module is specifically used to splice the output results of the first model, the second model, the third model and the third model according to time, and input them into the fifth model for integrated training to obtain the health assessment model.

[0178] In some embodiments, the fifth model is a LightGBM model.

[0179] In some embodiments, the target disease is chronic obstructive pulmonary disease.

[0180] The health assessment device of this embodiment can implement each step of the above-mentioned health assessment method embodiment and can achieve basically the same technical effects, which will not be repeated here.

[0181] The present disclosure also provides an electronic device. Referring to FIG8 , the electronic device may include a processor 801 , a memory 802 , and a program 8021 stored in the memory 802 and executable on the processor 801 .

[0182] When program 8021 is executed by processor 801, it can implement any steps in the above method embodiment and achieve the same beneficial effects, which will not be repeated here.

[0183] Those skilled in the art will appreciate that all or part of the steps of implementing the above-described embodiment method may be accomplished through hardware associated with program instructions, and the program may be stored in a readable medium.

[0184] The embodiments of the present disclosure also provide a readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, any step in the above method embodiment can be implemented and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0185] The storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0186] It should be noted that it should be understood that the division of the above modules is merely a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. Moreover, these modules can all be implemented in the form of software called by a processing element; or they can all be implemented in the form of hardware; or some modules can be implemented in the form of software called by a processing element, and some modules can be implemented in the form of hardware. For example, the determination module can be a separately established processing element, or it can be integrated into a chip of the above-mentioned device. In addition, it can also be stored in the memory of the above-mentioned device in the form of program code, and called by a processing element of the above-mentioned device to perform the functions of the above-mentioned determination module. The implementation of other modules is similar. In addition, these modules can all or partly be integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each of the above modules can be completed by the hardware integrated logic circuit in the processor element or by instructions in the form of software.

[0187] For example, each module, unit, sub-unit or sub-module may be one or more integrated circuits configured to implement the above method, such as one or more application-specific integrated circuits (ASICs), one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs). For another example, when a module is implemented by scheduling program code through a processing element, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0188] The above is a preferred implementation of the embodiment of the present disclosure. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles described in the present disclosure. These improvements and modifications should also be regarded as the scope of protection of the present disclosure.

Claims

1. A data processing method comprising the following steps: Acquiring vital sign data of a target object, wherein the vital sign data includes monitoring data obtained by monitoring the vital signs of the target object and supplemented data obtained by supplementing the vital sign data according to the monitoring data; Collecting multimodal data of the target object, wherein the multimodal data includes at least one of image data of the target object and survey data for preset symptoms; The body state data of the target object is generated according to the vital sign data and the multimodal data.

2. The method of claim 1, wherein: The step of obtaining the vital sign data of the target object includes: Acquiring monitoring data obtained by monitoring the vital signs of the target object; Binning the monitoring data; Grouping the monitoring data after binning according to a preset monitoring period; Generate supplementary data to supplement the missing monitoring data in each monitoring cycle; The monitoring data and the supplemented data are used as vital sign data of the target object.

3. The method of claim 2, wherein: The binning of the monitoring data comprises: The monitoring data is binned according to the degree of impact of the monitoring data on the target disease.

4. The method of claim 3, wherein: The binning of the monitoring data comprises: The monitoring data is binned using the minimum entropy binning method.

5. The method according to any one of claims 2 to 4, wherein: After grouping the monitoring data after the binning according to the preset monitoring period, the method further includes: Detecting the number of first cycles in the monitoring cycle, wherein the first cycle is a monitoring cycle in which monitoring data at a target time is missing; When the number of the first cycles is greater than a preset number threshold, the monitoring data at the target moment of the first cycle is determined according to the monitoring data at the target moment of the second cycle, wherein the second cycle is a monitoring period in which the monitoring data at the target moment is not missing.

6. The method of claim 2, wherein: The generating of the supplementary data to supplement the missing monitoring data in each monitoring cycle includes: The complementary data of the monitoring data are generated by cubic spline interpolation.

7. The method of claim 1, wherein: The collecting of multimodal data of the target object includes: Pushing a questionnaire targeting the preset symptoms to the target object; Receiving survey data input by the target subject in response to the questionnaire, wherein the survey data includes a selection input of at least one option among a plurality of options for each question set for the preset symptom; The survey data are standardized according to preset rules to form multimodal data, wherein the standardized survey data are used as external variables of a health assessment model.

8. A health assessment method comprising the following steps: Acquire physical state data of the target object, wherein the physical state data is obtained by the data processing method according to any one of claims 1 to 7; The physical condition data is input into a health assessment model to obtain a health assessment result of the target object suffering from a target disease, wherein the health assessment model is a pre-trained model that takes the physical condition data as input and the probability of suffering from the target disease as output.

9. The method of claim 8, wherein: The health assessment model includes an integrated first model, a second model and a third model, wherein the first model is an ARIMA model, the second model is an Informer model, and the third model is an N-BeatXs model.

10. The method of claim 9, wherein: The input data of the first model and the second model include the vital sign data; The input data of the third model includes the vital sign data and the multimodal data.

11. The method of claim 9, wherein: The health assessment model also includes a fourth model, which is a model that takes the resampled vital sign data as input and takes the probability of suffering from the target disease as output.

12. The method of claim 11, wherein: The fourth model is the Informer model.

13. The method according to claim 11 or 12, wherein: The step of obtaining a health assessment result of the target object suffering from the target disease by using the health assessment model includes: The vital sign data is resampled into high-frequency data and low-frequency data, wherein the sampling frequency of the high-frequency data is greater than the sampling frequency of the low-frequency data, and the sampling frequency of the low-frequency data is No less than twice within the test cycle; Performing difference processing on the low-frequency data according to the sampling frequency of the high-frequency data; The high-frequency data and the low-frequency data after difference processing are respectively input into the fourth model to obtain the prediction result of the periodic trend of the vital sign data.

14. The method according to claim 11 or 12, wherein: Inputting the physical status data into a health assessment model to obtain a health assessment result of the target subject suffering from a target disease includes: The output results of the first model, the second model, the third model and the third model are spliced ​​in time, and then input into the fifth model for integrated training to obtain the health assessment model.

15. The method of claim 14, wherein: The fifth model is the LightGBM model.

16. The method according to any one of claims 8 to 12, wherein: The target disease is chronic obstructive pulmonary disease.

17. A data processing device, comprising: A vital sign data acquisition module, used to acquire vital sign data of a target object, wherein the vital sign data includes monitoring data obtained by monitoring the vital signs of the target object and supplemented data supplemented by the vital sign data according to the monitoring data; A multimodal data acquisition module, used to acquire multimodal data of the target object, wherein the multimodal data includes at least one of image data of the target object and survey data for preset symptoms; A body state data generating module is used to generate the body state data of the target object according to the vital sign data and the multimodal data.

18. A health assessment device, comprising: A body state data acquisition module, used to acquire body state data of a target object, wherein the body state data is obtained by the data processing method according to any one of claims 1 to 7; A health assessment module is used to input the physical condition data into a health assessment model to obtain a health assessment result of the target object suffering from a target disease, wherein the health assessment model is a pre-trained model that takes the physical condition data as input and the probability of suffering from the target disease as output.

19. An electronic device comprising: A memory, a processor, and a program stored in the memory and executable on the processor; The processor is configured to read a program in a memory to implement the steps of the method according to any one of claims 1 to 7; Or implement the steps in the method as claimed in any one of claims 8 to 16.

20. A readable storage medium for storing a program, wherein when the program is executed by a processor, the steps in the method according to any one of claims 1 to 7 are implemented; Or implement the steps in the method as claimed in any one of claims 8 to 16.

Images