Intervention program recommendation method based on multidimensional physical sign assessment and training feedback

Abstract

The present invention belongs to the technical field of intelligent health management, and discloses an intervention plan recommendation method based on multidimensional physical sign evaluation and training feedback; the method comprises: collecting multidimensional physical sign data of patients in real time; collecting historical physiological physical sign data of patients; analyzing multidimensional physical sign data and historical physiological sign data to obtain training feedback data; evaluating the rehabilitation progress of patients according to the training feedback data; combining multidimensional physical sign data, training feedback data and rehabilitation progress, and adopting a nature-inspired optimization algorithm to recommend an intervention plan to patients; the present invention makes full use of multidimensional feature data and historical physiological sign data, and captures complex data relationships through deep learning technology, so as to fully and accurately understand the health status of patients, provide patients with customized rehabilitation plans, and thus effectively improve the rehabilitation quality and efficiency of patients.

Description

Technical Field

[0001] The present invention relates to the field of intelligent health management technology, and more specifically, to an intervention program recommendation method based on multi-dimensional physical sign evaluation and training feedback. Background Art

[0002] With the advancement of medical technology, more and more patients and rehabilitation populations need personalized rehabilitation training. For example, patients with chronic obstructive pulmonary disease usually need breathing training and aerobic exercise, and patients with neurological diseases (such as stroke and Parkinson's disease) usually need muscle strength training, motor coordination training, etc. Traditional rehabilitation training methods mainly rely on the experience and subjective diagnosis of doctors and therapists, which are less targeted and difficult to meet the individual needs of different patients. With the rapid development of wearable devices and Internet of Things technology, various physiological indicators and behavioral data of patients, such as heart rate, blood pressure, activity level, oxygen saturation, etc., can be monitored in real time. At the same time, through human-computer interaction feedback, the subjective feelings and feedback of patients during rehabilitation training can also be collected. Using these multi-dimensional physical sign data and training feedback, the health status and rehabilitation progress of patients can be evaluated more comprehensively and accurately, and more personalized intervention plans can be formulated to provide customized rehabilitation intervention measures such as exercise training, nutritional guidance, psychological counseling, etc. according to the specific conditions of patients, so as to improve the rehabilitation effect of patients and shorten the rehabilitation cycle of patients.

[0003] The Chinese patent with publication number CN118230977A discloses a rehabilitation assistance management system in smart medical care; it includes: a user end and a server end; the server end includes a patient information management component, a symptom status analysis component, a rehabilitation plan recommendation component, a patient status monitoring component and a rehabilitation therapist training component; the patient information management component is used to receive and store basic patient information transmitted by the user end; the symptom status analysis component is used to receive patient vital sign status information transmitted by the user end to parse the patient's symptom status; the rehabilitation plan recommendation component is used to match the rehabilitation recommendation plan according to the patient's symptom status; the patient status monitoring component is used to monitor the abnormal physical signs of the patient's vital sign status information; the rehabilitation therapist training component is used for rehabilitation therapist training and rehabilitation therapist training assessment; it solves the technical problem that it is difficult to perform high-level semantic analysis of patient information in the prior art, resulting in low reference value of patient information and poor rehabilitation assistance management effect.

[0004] However, when recommending a plan, the above-mentioned technology generates an initial rehabilitation plan for further screening and confirmation by clustering analysis and screening of patient data. It relies on user feedback and is difficult to directly recommend the optimal rehabilitation plan. In addition, the above-mentioned technology mainly relies on linear or piecewise linear distance analysis and clustering methods. When faced with large amounts of data and high-dimensional features, it requires a lot of computing resources and has low computing efficiency. At the same time, it performs poorly when faced with nonlinear and multimodal patient vital sign data and has difficulty capturing complex data relationships.

[0005] In view of this, the present invention proposes an intervention plan recommendation method based on multi-dimensional physical sign assessment and training feedback to solve the above problems. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: an intervention plan recommendation method based on multi-dimensional physical sign assessment and training feedback, comprising:

[0007] S1: Collect multi-dimensional vital sign data of patients in real time;

[0008] S2: Collect historical physiological data of patients;

[0009] S3: Analyze multidimensional physical sign data and historical physiological sign data to obtain training feedback data;

[0010] S4: Evaluate the patient’s rehabilitation progress based on training feedback data;

[0011] S5: Combining multidimensional physical sign data, training feedback data and rehabilitation progress, a nature-inspired optimization algorithm is used to recommend intervention plans for patients.

[0012] Furthermore, the intervention program is a rehabilitation training program for the patient, and the step of recommending the intervention program to the patient includes:

[0013] Step 1: Preset a plan set, set different digital labels for different intervention plans in the plan set, and mark them as plan labels;

[0014] Step 2: Preset the population size Z and the number threshold T;

[0015] Step 3: Initialize the population. The positions of the remoras in the initial population are defined in a one-dimensional search space. The positions of the remoras correspond to the solution labels one by one.

[0016] Step 4: Determine the fitness function;

[0017] Step 5: Select the host that each remora parasites and update the position of each remora;

[0018] Step 6: Calculate the fitness of each remora after the position is updated, and determine whether to switch hosts;

[0019] Step 7: Update the position of the remora that has switched hosts again;

[0020] Step 8: Each remora enters the foraging phase, and the position of each remora is updated;

[0021] Step 9: Calculate the dynamic boundary, generate the reverse solution corresponding to each remora and add it to the population; use the reverse solution of the remora as the remora, screen the remora in the population, and retain the Z remora with the largest fitness;

[0022] Step 10: If the number of iterations t is less than the number threshold T, return to step 5 and set the number of iterations ; If the number of iterations t is greater than or equal to the number threshold T, proceed to step 11;

[0023] Step 11: Calculate the fitness corresponding to the position of each remora, sort all fitnesses from large to small, obtain the position of the remora corresponding to the fitness ranked first, and obtain the intervention plan corresponding to the corresponding plan label according to the obtained remora position.

[0024] Furthermore, in step 3, the population is initialized , including Z remoras;

[0025] The range of solution labels and one-dimensional search space are , W is the number of intervention plans in the plan set; the expression for the position of each remora is: ; In the formula, is the initial position of the ith reef fish, for A random number between , the number of iterations t corresponding to the initialization population is 0;

[0026] In step 4, the fitness function is expressed as: KF; where f is fitness and KF is the increase in rehabilitation progress;

[0027] The method for obtaining the increase in rehabilitation progress is as follows: using multidimensional physical sign data, training feedback data and rehabilitation progress as test data; using the program label and test data corresponding to the position of the remora as research data, inputting the research data into the trained gain prediction model, and predicting the corresponding increase in rehabilitation progress.

[0028] Furthermore, in step 5, the host includes a swordfish and a whale; a random number corresponding to each remora is obtained. And rounding operation is performed. If the random number after rounding is 0, then the host of the remora is a whale. If the rounded random number If it is 1, then the host of the remora is the swordfish;

[0029] Calculate the fitness of each remora position and mark the remora position with the largest fitness as ;

[0030] If the parasitic host is a whale, the updated calculation method of the remora position includes:

[0031] ;

[0032] ;

[0033] In the formula, is the updated position of the i-th reef fish, is the position of the i-th reef fish before updating, is the search range adjustment factor, for A random number between

[0034] If the parasitic host is a sailfish, the updated calculation method of the position of the remora includes:

[0035] ;

[0036] In the formula, The position of a random remora before the update.

[0037] Furthermore, in step 6, when the remora parasitizes on the host, it will tentatively move according to the final position of the remora in the previous iteration and the current position of the remora; calculate the fitness corresponding to the position of the remora after the tentative movement, and mark it as the tentative fitness; mark the fitness corresponding to the current position of the remora as the current fitness; compare the tentative fitness with the current fitness; if the tentative fitness is greater than or equal to the current fitness, do not switch the host; if the tentative fitness is less than the current fitness, switch the host; the expression of the position of the remora after the tentative movement is: ; In the formula, The remora fish tentatively moves to its rear position. is the final position of the reef fish in the last iteration, for Normally distributed random numbers;

[0038] In step 7, the method for updating the position of the remora that switches hosts is consistent with the method for updating the position of each remora in step 5.

[0039] Furthermore, in step 8, the method for updating the position of the remora when it enters the foraging stage includes:

[0040] ;

[0041] ;

[0042] ;

[0043] ;

[0044] In the formula, is the updated position of the ith reef fish after the foraging phase, is the updated position of the i-th reef fish after switching hosts, is the moving distance of the ith remora in the foraging stage, is the volume of the host parasitized by the ith remora, is the remora factor, , is the volume of the ith reef fish, for A random number between

[0045] In step 9, the dynamic boundary is , is the label of the solution with the smallest value in the population, The label of the solution with the largest value in the population;

[0046] The methods for generating the inverse solution include:

[0047] ;

[0048] In the formula, is the reverse solution of the ith reef fish, for Random numbers in ;

[0049] If the generated inverse solution is greater than or less than , it is marked as a transcendental solution, a random number function is used to randomly generate a value within the dynamic boundary, and the randomly generated value is assigned to the transcendental solution.

[0050] Furthermore, the multidimensional physical sign data includes basic physical sign data, clinical physical sign data and physiological physical sign data; the basic physical sign data is an indicator describing the individual characteristics of the patient; the clinical physical sign data is the condition of the patient; and the physiological physical sign data is various indicators reflecting the patient's physiological state;

[0051] The historical physiological sign data include historical pre-training sign data and historical post-training sign data; the historical pre-training sign data are the physiological sign data of the patient before each rehabilitation training begins, and the historical post-training sign data are the physiological sign data of the patient after each rehabilitation training ends.

[0052] Furthermore, the method for obtaining training feedback data includes:

[0053] Training feedback data includes short-term feedback data and long-term feedback data;

[0054] The short-term feedback data is the amount of change in the patient's physiological sign data after each rehabilitation training. The method for obtaining the short-term feedback data is as follows: the historical pre-training sign data and the historical post-training sign data corresponding to the same rehabilitation training in the historical physiological sign data are taken as a first set; each data in the historical post-training sign data in each first set is subtracted from the corresponding data in the historical pre-training sign data in the same first set, and the amount of change in the physiological sign data corresponding to each first set is obtained as the short-term feedback data.

[0055] Long-term feedback data refers to the change in the patient's physiological sign data before the last rehabilitation training and before the current rehabilitation training. The method for obtaining long-term feedback data is as follows: the historical pre-training sign data and the real-time collected physiological sign data are sorted according to the order of the corresponding rehabilitation training time, that is, the physiological sign data of each group are sorted from small to large according to the corresponding collection time, and a sorting table is generated; the collection time is represented by a timestamp;

[0056] Each group of physiological sign data in the sorting table is sequentially sorted with the next group of physiological sign data as a second set; the collection time corresponding to the two groups of physiological sign data in each second set is compared, the physiological sign data with a shorter collection time is marked as the first data, and the physiological sign data with a longer collection time is marked as the second data; each data in the second data in each second set is subtracted from the corresponding data in the first data in the same second set, and the change in the physiological sign data corresponding to each second set is obtained, and used as the long-term feedback data.

[0057] Furthermore, the method for evaluating the patient's recovery progress includes:

[0058] Different digital labels are set for different clinical signs data, and marked as disease labels; the changes in the same data in the short-term feedback data and the long-term feedback data are respectively taken as a third set; the disease label, the data in a third set and the basic signs data are taken as a group of analysis data, and each group of analysis data is input into the trained weight distribution model to predict the corresponding weight coefficient, and the weight coefficient corresponds to the data in the physiological signs data one by one;

[0059] Preset the normal range of physical signs, which includes the normal range of each data in the physiological sign data; obtain the physiological sign data before the first rehabilitation training in the historical physiological sign data, and mark it as the first physical sign data; compare each data in the first physical sign data with the corresponding normal range of physical signs, and mark the data with a value greater than the maximum value of the normal range of the corresponding physical sign or less than the minimum value of the corresponding physical sign range as abnormal data; data with a value less than or equal to the maximum value of the normal range of the corresponding physical sign and greater than or equal to the minimum value of the normal range of the corresponding physical sign will not be marked;

[0060] Subtract the maximum and minimum values ​​of the normal range of the corresponding sign from each abnormal data in the first vital sign data, and take the absolute values ​​to obtain two differences; compare the two differences corresponding to each abnormal data in the first vital sign data, and take the difference with the smaller value as the abnormal amount of the corresponding abnormal data; mark the data corresponding to the abnormal data in the physiological sign data collected in real time as real-time abnormal data, subtract the corresponding abnormal data from each real-time abnormal data and take the absolute value to obtain the rehabilitation amount of each real-time abnormal data; divide the rehabilitation amount of each real-time abnormal data by the abnormal amount of the corresponding abnormal data to obtain the rehabilitation rate of each real-time abnormal data; obtain the weight coefficient corresponding to each real-time abnormal data, multiply the rehabilitation rate corresponding to each real-time abnormal data by the corresponding weight coefficient, and add them up in sequence to obtain the patient's rehabilitation progress.

[0061] Furthermore, the training process of the weight distribution model includes:

[0062] Collect a group of analysis data in advance, set corresponding weight coefficients for the group of analysis data, where a is an integer greater than 1, and convert the analysis data and the corresponding weight coefficients into a corresponding set of feature vectors;

[0063] Each group of feature vectors is used as the input of the weight allocation model. The weight allocation model uses a group of predicted weight coefficients corresponding to each group of analysis data as output, and uses the actual weight coefficients corresponding to each group of analysis data as the prediction target, where the actual weight coefficients are the pre-collected weight coefficients corresponding to the analysis data; minimizing the sum of the prediction errors of all analysis data is used as the training target; the weight allocation model is trained until the sum of the prediction errors reaches convergence and the training is stopped; the weight allocation model is a deep neural network model;

[0064] The training process of the gain prediction model is consistent with the training process of the weight allocation model, and both are deep neural network models.

[0065] Technical effects and advantages of the intervention program recommendation method based on multi-dimensional physical sign assessment and training feedback of the present invention:

[0066] By collecting the patient's multidimensional vital sign data in real time and combining it with historical physiological sign data, the training feedback data of the patient after each rehabilitation training is analyzed to accurately evaluate the patient's rehabilitation progress; the improved fish optimization algorithm is used to recommend the best personalized intervention plan for the patient, based on the parallel computing processing mechanism to improve computing efficiency, and the effect of the plan is evaluated through deep learning technology, thereby improving the pertinence and effectiveness of rehabilitation training; making full use of multidimensional feature data and historical physiological sign data, and capturing complex data relationships through deep learning technology, it can fully and accurately understand the patient's health status, provide patients with customized rehabilitation plans, thereby effectively improving the patient's rehabilitation quality and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 This is a flow chart of a method for recommending an intervention plan based on multi-dimensional physical sign assessment and training feedback according to Embodiment 1 of the present invention;

[0068] Figure 2 This is a recommended flow chart of the intervention scheme of Example 1 of the present invention;

[0069] Figure 3 This is a schematic diagram of an electronic device according to Embodiment 2 of the present invention;

[0070] Figure 4 This is a schematic diagram of a storage medium according to Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0071] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0072] Example 1

[0073] See also Figure 1 As shown, the intervention plan recommendation method based on multi-dimensional physical sign assessment and training feedback described in this embodiment includes:

[0074] S1: Collect multi-dimensional vital sign data of patients in real time.

[0075] Multidimensional vital sign data include basic vital sign data, clinical vital sign data and physiological vital sign data; basic vital sign data are indicators describing individual characteristics of patients, such as height, weight, age, etc.; clinical vital sign data are the symptoms of patients, such as stroke, Parkinson's disease, chronic obstructive pulmonary disease, etc.; basic vital sign data and clinical vital sign data are obtained from the hospital's medical records. If the patient is visiting the hospital for the first time, the patient will fill in the data based on his or her own physical signs; physiological vital sign data are various indicators reflecting the patient's physiological state, such as heart rate, blood pressure, muscle strength, oxygen saturation, activity level, etc., and physiological vital sign data are obtained after the patient wears wearable devices; wearable devices include smart clothing, smart protective gear, smart bracelets, etc., and wearable devices are integrated with heart rate sensors, blood pressure sensors, electromyography sensors, acceleration sensors, etc.; patients wear wearable devices before rehabilitation training to accurately collect the patient's physiological vital sign data.

[0076] S2: Collect historical physiological sign data of the patient.

[0077] The historical physiological signs data include historical pre-training signs data and historical post-training signs data; the historical pre-training signs data are the physiological signs data of the patient before the start of each rehabilitation training, and the historical post-training signs data are the physiological signs data of the patient after the end of each rehabilitation training; the historical physiological signs data are also obtained after the patient wears the wearable device.

[0078] It should be noted that the purpose of collecting patients' multidimensional vital signs data and historical physiological signs data is to fully understand the patients' individual characteristics, physiological function status and specific symptoms, and by comparing the changes in the patients' physiological signs data before and after rehabilitation training, we can objectively evaluate the actual effect of each rehabilitation training, so as to more accurately evaluate the patients' health status and rehabilitation progress, and provide data support for the subsequent formulation of personalized rehabilitation training plans.

[0079] S3: Analyze multi-dimensional vital sign data and historical physiological sign data to obtain training feedback data.

[0080] Methods for obtaining training feedback data include:

[0081] Training feedback data includes short-term feedback data and long-term feedback data;

[0082] The short-term feedback data is the change in the patient's physiological sign data after each rehabilitation training is completed; the method for obtaining the short-term feedback data is: the historical pre-training sign data and the historical post-training sign data corresponding to the same rehabilitation training in the historical physiological sign data are taken as a first set; the number of the first set is half of the corresponding number of physiological sign data in the historical physiological sign data; each data in the historical post-training sign data in each first set is subtracted from the corresponding data in the historical pre-training sign data in the same first set, and the change in the physiological sign data corresponding to each first set is obtained, and used as the short-term feedback data.

[0083] Long-term feedback data is the change in the patient's physiological sign data before the last rehabilitation training and before the current rehabilitation training. The method for obtaining long-term feedback data is as follows: sort the historical pre-training sign data and the real-time collected physiological sign data according to the order of the corresponding rehabilitation training time, that is, sort the data from small to large according to the corresponding collection time of each group of physiological sign data, and generate a sorting table; the collection time is represented by a timestamp, which is a digital representation of time. The larger the timestamp, the later the collection time, and vice versa. The smaller the timestamp, the earlier the collection time; the collection time is synchronously obtained when the wearable device collects the physiological sign data;

[0084] Each group of physiological sign data in the sorting table is combined with the next group of physiological sign data in positive order as a second set; the number of second sets is the corresponding number of physiological sign data in the historical pre-training physiological sign data plus one; the collection time corresponding to the two groups of physiological sign data in each second set is compared, the physiological sign data with a smaller collection time is marked as the first data, and the physiological sign data with a larger collection time is marked as the second data, and each data in the second data in each second set is subtracted from the corresponding data in the first data in the same second set to obtain the change in the physiological sign data corresponding to each second set, and use it as the long-term feedback data.

[0085] For example, before the first rehabilitation training, the patient's heart rate was 70 and blood pressure was 80. At the end of the first rehabilitation training, the heart rate was 90 and the blood pressure was 100. Before the second rehabilitation training, the heart rate was 75 and the blood pressure was 85. Then in the short-term feedback data, the change in heart rate is 20, and the change in blood pressure is 20. In the long-term feedback data, the change in heart rate is 5, and the change in blood pressure is 5.

[0086] It should be understood that the reason for obtaining short-term feedback data and long-term feedback data is that short-term feedback data mainly reflects the immediate impact and improvement of each rehabilitation training on the patient's physiological signs data, and long-term feedback data mainly reflects the cumulative effect of training on patients over a long period of time and the long-term changes in the influence of non-training factors, which are used to understand the changing trends of patients' physiological signs data; the combination of the two types of feedback data can comprehensively evaluate the patient's rehabilitation progress, understand the patient's overall rehabilitation effect, and continuously optimize personalized rehabilitation training plans based on short-term feedback data and long-term feedback data, thereby improving rehabilitation effects.

[0087] S4: Evaluate the patient’s rehabilitation progress based on training feedback data.

[0088] Methods of assessing a patient's recovery progress include:

[0089] Different digital labels are set for different clinical signs data and marked as disease labels; the changes in the same data in the short-term feedback data and the long-term feedback data are respectively taken as a third set; the disease labels, the data in a third set and the basic signs data are taken as a group of analysis data, and each group of analysis data is input into the trained weight allocation model to predict the corresponding weight coefficients, which correspond one-to-one to the data in the physiological signs data.

[0090] The training process of the weight distribution model includes:

[0091] Collect a group of analysis data in advance, set corresponding weight coefficients for the group of analysis data, a is an integer greater than 1, and convert the analysis data and the corresponding weight coefficients into a corresponding set of feature vectors; the weight coefficients corresponding to the analysis data are collected by technicians in the historical patient rehabilitation progress assessment process, and under the conditions of each group of analysis data, comprehensively analyze the change trends of different data in the physiological sign data, and set corresponding weight coefficients for the degree of influence on different patients in combination with actual experience. The weight coefficients reflect the importance of different data in the physiological sign data to the patient's disease recovery under different change trends; set corresponding weight coefficients for the group of analysis data in turn;

[0092] Each set of feature vectors is used as the input of the weight distribution model. The weight distribution model uses a set of prediction weight coefficients corresponding to each set of analysis data as output, and uses the actual weight coefficients corresponding to each set of analysis data as the prediction target. The actual weight coefficients are the pre-collected weight coefficients corresponding to the analysis data. The training goal is to minimize the sum of the prediction errors of all analysis data. The calculation formula of the prediction error is: ,in is the prediction error, k is the group number of the eigenvector corresponding to the analysis data, is the prediction weight coefficient corresponding to the kth group of analysis data, is the actual weight coefficient corresponding to the kth group of analysis data; the weight allocation model is trained until the sum of the prediction errors reaches convergence and the training is stopped.

[0093] The above-mentioned weight distribution model is specifically a deep neural network model; it includes an input layer, a hidden layer and an output layer; each hidden layer includes multiple neurons, each neuron is connected to the neurons in the next layer, and the connection contains weights, which determine the importance and influence of data transmission in the neural network; an activation function is applied to each neuron between the hidden layer and the output layer, and the activation function reflects nonlinearity, allowing the network to learn more complex patterns and features.

[0094] It should be noted that when allocating weights, it is necessary to consider both the importance and variation defects of the physiological sign data itself. The reason is that the importance of the same physiological sign data may vary for patients with different diseases. For example, for patients with lung disease, the weight of lung function indicators should be higher than that of heart rate indicators, while for patients with neurological diseases, the weights of muscle strength and coordination indicators are more important. In addition, even if two patients have the same indicator values ​​at the same time point, their rehabilitation progress should be different if the change trends are different. Among them, physiological sign data that continue to improve should have a larger weight, while the weight corresponding to physiological sign data that fluctuates or worsens should be relatively smaller. Therefore, the combination of the two can more accurately reflect the patient's actual recovery during the entire training process.

[0095] The normal range of physical signs is preset, and the normal range of physical signs includes the normal range of each data in the physiological sign data. The normal range of physical signs is preset by technical personnel in this field based on actual experience and reference to the normal value standards of physiological indicators formulated by authoritative institutions such as clinical medicine.

[0096] Obtain the physiological sign data before the first rehabilitation training in the historical physiological sign data and mark them as the first sign data; compare each data in the first sign data with the corresponding normal range of the sign, and mark the data with a value greater than the maximum value of the normal range of the sign or less than the minimum value of the normal range of the sign as abnormal data; and do not mark the data with a value less than or equal to the maximum value of the normal range of the sign and greater than or equal to the minimum value of the normal range of the sign;

[0097] Subtract the maximum and minimum values ​​of the normal range of the corresponding sign from each abnormal data in the first vital sign data, and take the absolute values ​​to obtain two differences; compare the two differences corresponding to each abnormal data in the first vital sign data, and take the difference with the smaller value as the abnormal amount of the corresponding abnormal data; mark the data corresponding to the abnormal data in the physiological sign data collected in real time as real-time abnormal data, subtract the corresponding abnormal data from each real-time abnormal data and take the absolute value to obtain the rehabilitation amount of each real-time abnormal data; divide the rehabilitation amount of each real-time abnormal data by the abnormal amount of the corresponding abnormal data to obtain the rehabilitation rate of each real-time abnormal data; obtain the weight coefficient corresponding to each real-time abnormal data, multiply the rehabilitation rate corresponding to each real-time abnormal data by the corresponding weight coefficient, and add them up in sequence to obtain the patient's rehabilitation progress.

[0098] S5: Combining multidimensional physical sign data, training feedback data and rehabilitation progress, a nature-inspired optimization algorithm is used to recommend intervention plans for patients.

[0099] The intervention program is a rehabilitation training program for the patient; for example, the rehabilitation training program includes respiratory muscle strength training, lung capacity training, walking training, etc. for chronic obstructive pulmonary disease, limb muscle strength training, balance training, etc. for Parkinson's disease, and includes the training frequency and intensity corresponding to each training.

[0100] Nature-inspired optimization algorithms include genetic algorithms, particle swarm optimization algorithms, simulated annealing algorithms, clonal selection algorithms, and remora optimization algorithms. This embodiment takes the improved remora optimization algorithm as an example to introduce the intervention plan recommendation method.

[0101] like Figure 2 As shown, the steps for recommending intervention options for patients include:

[0102] Step 1: Preset a plan set, set different digital labels for different intervention plans in the plan set, and mark them as plan labels;

[0103] Step 2: Preset the population size Z and the number threshold T;

[0104] Step 3: Initialize the population. The positions of the remoras in the initial population are defined in a one-dimensional search space. The positions of the remoras correspond to the solution labels one by one.

[0105] Step 4: Determine the fitness function;

[0106] Step 5: Select the host that each remora parasites and update the position of each remora;

[0107] Step 6: Calculate the fitness of each remora after the position is updated, and determine whether to switch hosts;

[0108] Step 7: Update the position of the remora that has switched hosts again;

[0109] Step 8: Each remora enters the foraging phase, and the position of each remora is updated;

[0110] Step 9: Calculate the dynamic boundary, generate the reverse solution corresponding to each remora and add it to the population; use the reverse solution of the remora as the remora, screen the remora in the population, and retain the Z remora with the largest fitness;

[0111] Step 10: If the number of iterations t is less than the number threshold T, return to step 5 and set the number of iterations , that is, the value of the number of iterations is increased by one and then assigned to the number of iterations; if the number of iterations t is greater than or equal to the number threshold T, then go to step 11;

[0112] Step 11: Calculate the fitness corresponding to the position of each remora, sort all fitnesses from large to small, obtain the position of the remora corresponding to the fitness ranked first, and obtain the intervention plan corresponding to the corresponding plan label according to the obtained remora position.

[0113] In the above step 1, the program set is obtained by technicians in this field by collecting historically successfully executed rehabilitation training programs and modifying the collected rehabilitation training programs in combination with diagnosis and treatment guidelines or treatment specifications issued by authoritative medical institutions.

[0114] In the above step 2, the population size Z is determined by technical personnel in this field. In the process of recommending historical intervention plans, under multiple groups of different test data conditions, multiple different population sizes are set for the same test data, and the fish optimization algorithm is optimized multiple times. The corresponding plan label is obtained after the same number of iterations. The population size with the same plan label and the actual plan label is used as the population size corresponding to the group of test data. The test data includes multidimensional sign data, training feedback data and rehabilitation progress, wherein the clinical sign data in the multidimensional sign data is replaced with the corresponding symptom label; the actual plan label is the digital label of the intervention plan that best matches the group of test data, and the actual plan label is obtained by technical personnel in this field based on actual experience; and the population size corresponding to each group of test data is obtained by analogy, and the average of the multiple population sizes is used as the preset population size Z.

[0115] The number threshold T is determined by technicians in this field. In the process of recommending historical intervention plans, under multiple groups of different test data conditions, the same test data is subjected to multiple fish optimization algorithms to obtain multiple plan labels, wherein the number of iterations of the fish optimization algorithm is different each time, and the population size is the same and is Z; the number of iterations that is closest to the plan label and the actual plan label is used as the number of iterations corresponding to the group of test data; and the number of iterations corresponding to each group of test data is obtained by analogy, and the average of multiple iteration numbers is used as the number threshold T.

[0116] It should be understood that the population size Z determines the breadth of the search. A larger population size can explore more solution spaces, while the number threshold T determines the termination condition of the algorithm and can control the speed at which the algorithm converges.

[0117] In step 3 above, the population is initialized , including Z remoras.

[0118] The range of solution labels and one-dimensional search space are , W is the number of intervention plans in the plan set; the expression for the position of each remora is: ; In the formula, is the initial position of the ith reef fish, for A random number between , the number of iterations t corresponding to the initialization population is 0.

[0119] In step 4 above, the fitness function is expressed as: KF; where f is fitness and KF is the increase in rehabilitation progress.

[0120] The method for obtaining the increase in rehabilitation progress is: taking the scheme label and test data corresponding to the position of the reel as research data, inputting the research data into the trained gain prediction model, and predicting the corresponding increase in rehabilitation progress; the training process of the gain prediction model is consistent with the training process of the weight allocation model, and both are deep neural network models.

[0121] In step 5 above, the hosts include sailfish and whales; get the random number corresponding to each remora And rounding operation is performed. If the random number after rounding is 0, then the host of the remora is a whale. If the rounded random number If it is 1, then the host of the remora is the swordfish; the remora effectively guides the search direction by parasitizing the host.

[0122] Calculate the fitness of each remora position and mark the remora position with the largest fitness as .

[0123] If the parasitic host is a whale, the updated calculation method of the remora position includes:

[0124] ;

[0125] ;

[0126] In the formula, is the updated position of the i-th reef fish, is the position of the i-th reef fish before updating, is the search range adjustment factor, for A random number between .

[0127] If the parasitic host is a sailfish, the updated calculation method of the position of the remora includes:

[0128] ;

[0129] In the formula, The position of a random remora before the update.

[0130] In the above step 6, when the remora parasitizes on the host, it will tentatively move according to the final position of the remora in the previous iteration and the current position of the remora; calculate the fitness corresponding to the position of the remora after the tentative movement and mark it as the tentative fitness; mark the fitness corresponding to the current position of the remora as the current fitness; compare the tentative fitness with the current fitness; if the tentative fitness is greater than or equal to the current fitness, do not switch the host; if the tentative fitness is less than the current fitness, switch the host; the expression of the position of the remora after the tentative movement is: ; In the formula, The remora fish tentatively moves to its rear position. is the final position of the reef fish in the last iteration, for Normally distributed random numbers.

[0131] It should be noted that the purpose of switching hosts is that when the trial fitness is less than the current fitness, it means that after the remora has moved in a small range around the host, the fitness of the new position is lower, that is, the current host position may have been close to the local optimum, and continuing to explore around the host may not bring significant improvements. Therefore, by switching hosts, the remora can jump out of the current local area and explore a wider solution space, increasing the chance of finding the global optimal solution, and avoiding all individuals from gathering in the same area, increasing the diversity of the population, and making the algorithm more adaptable and robust.

[0132] In the above step 7, the method for updating the position of the remora that switches hosts is the same as the method for updating the position of each remora in step 5.

[0133] In the above step 8, the method for updating the position of the remora when it enters the foraging stage includes:

[0134] ;

[0135] ;

[0136] ;

[0137] ;

[0138] In the formula, is the updated position of the ith reef fish after the foraging phase, is the updated position of the i-th reef fish after switching hosts, is the moving distance of the ith remora in the foraging stage, is the volume of the host parasitized by the ith remora, is the fish factor, which is used to limit the location of the fish. , is the volume of the ith reef fish, for A random number between .

[0139] In step 9 above, the dynamic boundary is , is the label of the solution with the smallest value in the population, is the label of the solution with the largest value in the population.

[0140] The methods for generating the inverse solution include:

[0141] ;

[0142] In the formula, is the reverse solution of the ith reef fish, for The random number in .

[0143] If the generated inverse solution is greater than or less than , it is marked as a transcendental solution, a random number function is used to randomly generate a value within the dynamic boundary, and the randomly generated value is assigned to the transcendental solution.

[0144] It should be noted that the reason for using the optimized remora optimization algorithm to recommend intervention plans for patients is that the remora optimization algorithm is flexible and adaptable, and can define fitness functions according to specific performance requirements, and flexibly respond to the problem of recommending intervention plans for patients with different symptoms; and it has excellent global search capabilities. By continuously updating the position of the remora parasitic host, it can explore in a wide search space and increase the probability of finding the global optimal solution; at the same time, mechanisms such as host switching and foraging are also adopted, which can effectively overcome the problem of local optimality and accelerate the convergence of the algorithm; furthermore, the optimized remora optimization algorithm has low requirements for computing resources, the results are highly interpretable, and the scope of application is wide, and it can perform parallel computing processing to improve computing efficiency; in addition, integrating the trained neural network model into the fitness calculation of the optimized remora optimization algorithm can give full play to the nonlinear and multimodal patient vital sign data modeling capabilities of the neural network and effectively capture complex data relationships.

[0145] This embodiment collects the patient's multidimensional vital sign data in real time, and combines it with historical physiological sign data to analyze the patient's training feedback data after each rehabilitation training, so as to accurately evaluate the patient's rehabilitation progress; uses the improved fish optimization algorithm to recommend the best personalized intervention plan for the patient, improves computing efficiency based on the parallel computing processing mechanism, and evaluates the effect of the plan through deep learning technology, thereby improving the pertinence and effectiveness of rehabilitation training; makes full use of multidimensional feature data and historical physiological sign data, and captures complex data relationships through deep learning technology, so as to fully and accurately understand the patient's health status, provide patients with customized rehabilitation plans, and effectively improve the patient's rehabilitation quality and efficiency.

[0146] Example 2

[0147] See also Figure 3 As shown, the present application also provides an electronic device 500. The electronic device 500 may include one or more processors and one or more memories. The memories store computer readable codes, and when the computer readable codes are executed by one or more processors, the intervention scheme recommendation method based on multi-dimensional physical sign assessment and training feedback as described above may be executed.

[0148] The method or system according to the embodiment of the present application can also be used by Figure 3 The electronic device architecture shown in FIG. Figure 3As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, ROM 503, RAM 504, a communication port 505 connected to a network, an input / output 506, a hard disk 507, etc. The storage device in the electronic device 500, such as ROM 503 or hard disk 507, may store the intervention plan recommendation method based on multi-dimensional physical sign assessment and training feedback provided by the present application. Further, the electronic device 500 may also include a user interface 508. Of course, Figure 3 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 3 One or more components of an electronic device are shown.

[0149] Example 3

[0150] See also Figure 4 As shown, one embodiment of the present application discloses a computer-readable storage medium 600. Computer-readable instructions are stored on the computer-readable storage medium 600. When the computer-readable instructions are executed by the processor, the intervention plan recommendation method based on multidimensional sign assessment and training feedback according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium 600 includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0151] In addition, according to the embodiments of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided in the present application, for example: an intervention program recommendation method based on multidimensional physical sign assessment and training feedback. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0152] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0153] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. Recommendation method for intervention plan based on multi-dimensional physical sign assessment and training feedback, It is characterized in that include: S1: Collect multi-dimensional vital sign data of patients in real time; The multidimensional physical sign data includes basic physical sign data, clinical physical sign data and physiological physical sign data; S2: Collecting historical physiological sign data of the patient; the historical physiological sign data includes historical pre-training sign data and historical post-training sign data; S3: Analyze multidimensional physical sign data and historical physiological sign data to obtain training feedback data; The method for obtaining training feedback data includes: Training feedback data includes short-term feedback data and long-term feedback data; The short-term feedback data is the amount of change in the patient's physiological sign data after each rehabilitation training. The method for obtaining the short-term feedback data is as follows: the historical pre-training sign data and the historical post-training sign data corresponding to the same rehabilitation training in the historical physiological sign data are taken as a first set; each data in the historical post-training sign data in each first set is subtracted from the corresponding data in the historical pre-training sign data in the same first set, and the amount of change in the physiological sign data corresponding to each first set is obtained as the short-term feedback data. Long-term feedback data refers to the change in the patient's physiological sign data before the last rehabilitation training and before the current rehabilitation training. The method for obtaining long-term feedback data is as follows: the historical pre-training sign data and the real-time collected physiological sign data are sorted according to the order of the corresponding rehabilitation training time, that is, the physiological sign data of each group are sorted from small to large according to the corresponding collection time, and a sorting table is generated; the collection time is represented by a timestamp; Each set of physiological sign data in the sorting table is sequentially combined with the next set of physiological sign data in positive order as a second set; the collection time corresponding to the two sets of physiological sign data in each second set is compared, the physiological sign data with a shorter collection time is marked as the first data, and the physiological sign data with a longer collection time is marked as the second data; each data in the second data in each second set is subtracted from the corresponding data in the first data in the same second set, and the change amount of the physiological sign data corresponding to each second set is obtained as the long-term feedback data; S4: Evaluate the patient’s rehabilitation progress based on training feedback data; S5: Combining the multi-dimensional physical sign data, training feedback data and rehabilitation progress, a nature-inspired optimization algorithm is used to recommend an intervention plan for the patient; the intervention plan is a rehabilitation training plan for the patient, and the steps of recommending an intervention plan for the patient include: Step 1: Preset a plan set, set different digital labels for different intervention plans in the plan set, and mark them as plan labels; Step 2: Preset the population size Z and the number threshold T; Step 3: Initialize the population. The positions of the remoras in the initial population are defined in a one-dimensional search space. The positions of the remoras correspond to the solution labels one by one. Step 4: Determine the fitness function; Step 5: Select the host that each remora parasites and update the position of each remora; Step 6: Calculate the fitness of each remora after the position is updated, and determine whether to switch hosts; Step 7: Update the position of the remora that has switched hosts again; Step 8: Each remora enters the foraging phase, and the position of each remora is updated; Step 9: Calculate the dynamic boundary, generate the reverse solution corresponding to each remora and add it to the population; use the reverse solution of the remora as the remora, screen the remora in the population, and retain the Z remora with the largest fitness; Step 10: If the number of iterations t is less than the number threshold T, return to step 5 and set the number of iterations ; If the number of iterations t is greater than or equal to the number threshold T, proceed to step 11; Step 11: Calculate the fitness corresponding to the position of each remora, sort all fitnesses from large to small, obtain the position of the remora corresponding to the fitness ranked first, and obtain the intervention plan corresponding to the corresponding plan label according to the obtained remora position.

2. The intervention program recommendation method based on multidimensional physical sign assessment and training feedback according to claim 1, It is characterized in that In step 3, the population is initialized , including Z remoras; The range of solution labels and one-dimensional search space are , W is the number of intervention plans in the plan set; The expression for the position of each remora is: ; In the formula, is the initial position of the ith reef fish, for A random number between , the number of iterations t corresponding to the initialization population is 0; In step 4, the fitness function is expressed as: KF; where f is fitness and KF is the increase in rehabilitation progress; The method for obtaining the increase in rehabilitation progress is as follows: using multidimensional physical sign data, training feedback data and rehabilitation progress as test data; using the program label and test data corresponding to the position of the remora as research data, inputting the research data into the trained gain prediction model, and predicting the corresponding increase in rehabilitation progress.

3. The intervention program recommendation method based on multidimensional physical sign assessment and training feedback according to claim 2, It is characterized in that In step 5, the hosts include sailfish and whales; a random number corresponding to each remora is obtained. And rounding operation is performed. If the random number after rounding is 0, then the host of the remora is a whale. If the rounded random number If it is 1, then the host of the remora is the swordfish; Calculate the fitness of each remora position and mark the remora position with the largest fitness as ; If the parasitic host is a whale, the updated calculation method of the remora position includes: ； ； In the formula, is the updated position of the i-th reef fish, is the position of the i-th reef fish before updating, is the search range adjustment factor, for A random number between If the parasitic host is a sailfish, the updated calculation method of the position of the remora includes: ； In the formula, The position of a random remora before the update.

4. The intervention program recommendation method based on multidimensional physical sign assessment and training feedback according to claim 3, It is characterized in that In step 6, when the remora parasitizes on the host, it will tentatively move according to the final position of the remora in the previous iteration and the current position of the remora; calculate the fitness corresponding to the position of the remora after the tentative movement, and mark it as the tentative fitness; mark the fitness corresponding to the current position of the remora as the current fitness; and compare the tentative fitness with the current fitness; If the tentative fitness is greater than or equal to the current fitness, the host will not be switched; if the tentative fitness is less than the current fitness, the host will be switched; the expression of the position of the remora after the tentative movement is: ; In the formula, The remora fish tentatively moves to its rear position. is the final position of the reef fish in the last iteration, for Normally distributed random numbers; In step 7, the method for updating the position of the remora that switches hosts is consistent with the method for updating the position of each remora in step 5.

5. The intervention program recommendation method based on multidimensional physical sign assessment and training feedback according to claim 4, It is characterized in that In step 8, the method for updating the position of the remora when it enters the foraging stage includes: ； ； ； ； In the formula, is the updated position of the ith reef fish after the foraging phase, is the updated position of the i-th reef fish after switching hosts, is the moving distance of the ith remora in the foraging stage, is the volume of the host parasitized by the ith remora, is the remora factor, , is the volume of the ith reef fish, for A random number between In step 9, the dynamic boundary is , is the label of the solution with the smallest value in the population, The label of the solution with the largest value in the population; The methods for generating the inverse solution include: ； In the formula, is the reverse solution of the ith reef fish, for Random numbers in ; If the generated inverse solution is greater than or less than , it is marked as a transcendental solution, a random number function is used to randomly generate a value within the dynamic boundary, and the randomly generated value is assigned to the transcendental solution.

6. The intervention program recommendation method based on multidimensional physical sign assessment and training feedback according to claim 5, It is characterized in that The basic physical sign data are indicators describing the individual characteristics of the patient; the clinical physical sign data are the symptoms of the patient; and the physiological physical sign data are various indicators reflecting the physiological state of the patient. The historical pre-training physical sign data are the physiological sign data of the patient before each rehabilitation training begins, and the historical post-training physical sign data are the physiological sign data of the patient after each rehabilitation training ends.

7. The method for recommending an intervention plan based on multidimensional physical sign assessment and training feedback according to claim 6, It is characterized in that The method for evaluating the patient's recovery progress includes: Different digital labels are set for different clinical signs data, and marked as disease labels; the changes in the same data in the short-term feedback data and the long-term feedback data are respectively taken as a third set; the disease label, the data in a third set and the basic signs data are taken as a group of analysis data, and each group of analysis data is input into the trained weight distribution model to predict the corresponding weight coefficient, and the weight coefficient corresponds to the data in the physiological signs data one by one; Preset the normal range of physical signs, which includes the normal range of each data in the physiological sign data; obtain the physiological sign data before the first rehabilitation training in the historical physiological sign data, and mark it as the first physical sign data; compare each data in the first physical sign data with the corresponding normal range of physical signs, and mark the data with a value greater than the maximum value of the normal range of the corresponding physical sign or less than the minimum value of the corresponding physical sign range as abnormal data; data with a value less than or equal to the maximum value of the normal range of the corresponding physical sign and greater than or equal to the minimum value of the normal range of the corresponding physical sign will not be marked; Subtract the maximum and minimum values ​​of the normal range of the corresponding sign from each abnormal data in the first vital sign data, and take the absolute values ​​to obtain two differences; compare the two differences corresponding to each abnormal data in the first vital sign data, and take the difference with the smaller value as the abnormal amount of the corresponding abnormal data; mark the data corresponding to the abnormal data in the physiological sign data collected in real time as real-time abnormal data, subtract the corresponding abnormal data from each real-time abnormal data and take the absolute value to obtain the rehabilitation amount of each real-time abnormal data; divide the rehabilitation amount of each real-time abnormal data by the abnormal amount of the corresponding abnormal data to obtain the rehabilitation rate of each real-time abnormal data; obtain the weight coefficient corresponding to each real-time abnormal data, multiply the rehabilitation rate corresponding to each real-time abnormal data by the corresponding weight coefficient, and add them up in sequence to obtain the patient's rehabilitation progress.

8. The method for recommending intervention plans based on multidimensional physical sign assessment and training feedback according to claim 7, It is characterized in that The training process of the weight distribution model includes: Collect a group of analysis data in advance, set corresponding weight coefficients for the group of analysis data, where a is an integer greater than 1, and convert the analysis data and the corresponding weight coefficients into a corresponding set of feature vectors; Each group of feature vectors is used as the input of the weight allocation model. The weight allocation model uses a group of predicted weight coefficients corresponding to each group of analysis data as output, and uses the actual weight coefficients corresponding to each group of analysis data as the prediction target, where the actual weight coefficients are the pre-collected weight coefficients corresponding to the analysis data; minimizing the sum of the prediction errors of all analysis data is used as the training target; the weight allocation model is trained until the sum of the prediction errors reaches convergence and the training is stopped; the weight allocation model is a deep neural network model; The training process of the gain prediction model is consistent with the training process of the weight allocation model, and both are deep neural network models.

Images