Postoperative rehabilitation management method and system

Abstract

The invention provides a postoperative rehabilitation management method and system, and the method comprises the following steps: 1, inputting parameters to a postoperative rehabilitation management model, and outputting the input parameters of a patient and the rehabilitation management suggestion of the patient; 2, the input parameters are processed and analyzed through a deep neural network, and rehabilitation evaluation signals are obtained; and step 3, analyzing the rehabilitation evaluation signal based on regression analysis of fuzzy integral of choquet, and evaluating the rehabilitation effect. The rehabilitation management model has the advantages of being simple, objective, accurate, capable of achieving real-time evaluation and the like, and postoperative rehabilitation of a patient can be accelerated.

Description

Technical Field

[0001] The present invention relates to the field of health information technology, and in particular to a postoperative rehabilitation management method and system. Background Art

[0002] After an operation, "postoperative recovery" is a very important period for patients, especially after a larger operation, when the body needs to recuperate and rest. During the postoperative recovery stage, long-term bed rest and inappropriate diet may cause various complications and sequelae, so postoperative recovery management is very important. As long as the exercise, diet, and sleep methods are reasonable, the risk of complications can be greatly reduced, and the postoperative recovery of patients can be effectively accelerated, allowing them to return to normal life as soon as possible.

[0003] At present, postoperative rehabilitation management is mainly determined by medical staff based on clinical experience and personal opinions, and there is a lack of a reliable evaluation standard to test whether rehabilitation management is safe and effective. The patient's rehabilitation is closely related to the experience of the clinician. In addition, postoperative rehabilitation not only includes rehabilitation management during hospitalization, but also rehabilitation management at home. Due to the differences in the rehabilitation progress of different patients, even if the medical staff has formulated a corresponding rehabilitation plan, the rehabilitation plan is still not adaptable due to the different rehabilitation progress of the patients. Therefore, how to achieve objective and real-time postoperative rehabilitation management is the main difficulty at present. Summary of the invention

[0004] In view of the difficulties in current postoperative rehabilitation management, the present invention proposes a postoperative rehabilitation management method based on artificial intelligence, which has the advantages of being simple to use, objective and accurate, and capable of real-time evaluation. It is conducive to accelerating the patient's postoperative rehabilitation and improving the patient's quality of life. It is achieved through the following technical solution.

[0005] A postoperative rehabilitation management method comprises the following steps: Step 1: Input parameters to the postoperative rehabilitation management model, and output the patient's input parameters and the patient's rehabilitation management suggestions; Step 2: Processing and analyzing the input parameters through a deep neural network to obtain a rehabilitation evaluation signal; Step 3: Analyze the rehabilitation evaluation signal based on the regression analysis of Choquet's fuzzy integral and evaluate the rehabilitation effect.

[0006] The beneficial effects of the present invention are as follows: the rehabilitation management model has the advantages of being simple, objective, accurate, and capable of real-time evaluation, and is conducive to accelerating the postoperative rehabilitation of patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 It is a flow chart of the postoperative rehabilitation management method of the present invention.

[0008] Figure 2 It is a management model diagram of postoperative rehabilitation management of the present invention. DETAILED DESCRIPTION

[0009] The embodiments of the present invention are described in detail below with reference to the accompanying drawings, but the present invention can be implemented in many different ways as defined and covered by the claims.

[0010] In the embodiment of the present invention, the device for collecting physiological signs is a smart wearable device, such as a smart wearable watch, a ring, a smart ECG patch or an EEG patch, etc. These smart wearable devices have certain data processing functions.

[0011] The model algorithm is configured on a mobile phone, computer or server, or on a smart wearable device with a display and input function.

[0012] like Figure 1 As shown, the embodiment of the present invention is mainly aimed at people after surgery, and provides a reminder for rehabilitation intervention after surgery, including the following steps: S1. Input parameters to the postoperative rehabilitation management model and output the patient's input parameters and suggestions for the patient's rehabilitation management; S2. Processing and analyzing the input parameters through a deep neural network to obtain a rehabilitation evaluation signal; S3. Regression analysis based on Choquet's fuzzy integral is used to analyze the rehabilitation evaluation signal and conduct a real-time and objective evaluation of the rehabilitation effect.

[0013] The method also includes: step 4, automatically adjusting the weight factors of the neuron connections between different hidden layers in the signal processing module according to the patient's postoperative rehabilitation effect.

[0014] like Figure 1-2 As shown, the method of postoperative rehabilitation management is explained through the embodiments of the present invention.

[0015] S1. Input parameters to the postoperative rehabilitation management model and output the patient's input parameters and recommendations for the patient's rehabilitation management.

[0016] The input module of the postoperative rehabilitation management model mainly sets the input parameters of the postoperative rehabilitation management model based on artificial intelligence. The input parameters include: patient's age, gender, body mass index (BMI), blood pressure, blood sugar, surgical site, surgical type, and surgical time.

[0017] On this basis, further subdivision is carried out as follows: the patient's age is divided into six subsets: <10 years old, 10-18 years old, 19-35 years old, 36-60 years old, 60-80 years old, and over 80 years old; gender is divided into two subsets: male and female; BMI is divided into five subsets: <18kg / m2, 18-24kg / m2, 25-28kg / m2, 29-35kg / m2, and greater than 35kg / m2; blood pressure is divided into three subsets: normal blood pressure, hypertension, and hypotension; blood sugar is divided into three subsets: normal blood sugar, hypoglycemia, and hyperglycemia; surgical sites are divided into eight subsets: musculoskeletal system, nervous system, endocrine system, circulatory system, respiratory system, digestive system, urinary system, and reproductive system; surgical types are divided into three subsets: non-invasive surgery, minimally invasive surgery, and invasive surgery; surgical time is divided into six subsets: <3 days after surgery, 3-7 days after surgery, 8-14 days after surgery, 15-21 days after surgery, 22-30 days after surgery, and more than 30 days after surgery.

[0018] Therefore, the input parameter is a matrix of size 36 × 1. The above matrix can also be adjusted or further subdivided according to actual conditions.

[0019] In other embodiments of the present invention, age, surgical grade, surgical incision size, and hospital stay may be further subdivided into more parameter matrices.

[0020] For example, surgical levels are divided into level one, level two, level three and level four based on the degree of ease to difficulty.

[0021] The output module of the postoperative rehabilitation management model mainly outputs the patient's input parameters and the patient's rehabilitation management suggestions. The output parameters include: the type of exercise on that day, the duration of exercise, the intensity of exercise, the nutritional combination of each meal, the interval between each meal, the total sleep time, the start time of sleep, and the number of sleeps on that day. Therefore, the output parameters are a matrix of size 8×1.

[0022] S2. Processing and analyzing the input parameters through a deep neural network to obtain a rehabilitation evaluation signal is achieved through a signal processing module.

[0023] The signal processing module mainly processes and analyzes the input parameters of the postoperative rehabilitation management model, which is mainly achieved through deep neural networks.

[0024] The deep neural network consists of 6 hidden layers. Each neuron in a hidden layer is connected to all neurons in the previous hidden layer to integrate the features extracted previously.

[0025] The number of neurons in each hidden layer is determined as follows: the number of neurons in the first hidden layer is 4 times the length of the input parameter matrix. When the input parameter matrix is ​​36×1, the number of neurons is 144; the number of neurons in the second, third, fourth, and fifth hidden layers is half of the number of neurons in the previous hidden layer, that is, 72, 36, 18, and 9 respectively. The number of neurons in the sixth hidden layer is consistent with the size of the output parameter matrix. When the output parameter is an 8×1 matrix, the number of neurons in the sixth hidden layer is 8. In order to reduce overfitting, a batch normalization layer and a dropout layer are added between each hidden layer, and the dropout probability is set to 0.2.

[0026] The input parameters obtained in the above steps are processed by a deep neural network to output a rehabilitation evaluation signal.

[0027] S3. The rehabilitation evaluation signal is analyzed based on the regression analysis of Choquet's fuzzy integral, and the rehabilitation effect is evaluated in real time and objectively, which is achieved through the postoperative rehabilitation evaluation module.

[0028] The main function of the postoperative rehabilitation evaluation module is to conduct real-time and objective evaluation of the patient's rehabilitation effect, thereby providing support for the patient to adjust the rehabilitation plan.

[0029] The input signal of the rehabilitation evaluation signal of this module is a set of {αt1, βt2, ..., γtn-1,δtn}, where tn represents the rehabilitation evaluation signal obtained at the current moment, which is a 1×12 matrix. The values ​​of the 12 elements in the matrix are: the patient's body fat mass, visceral fat mass, body water mass, muscle mass, basal metabolic rate, SDNN value (that is, the standard deviation of all NN intervals of heart rate, NN interval refers to the difference between two adjacent heartbeat times), RMSSD value (that is, the root mean square value of the difference between all adjacent NN intervals), triangle index (that is, the total number of NN intervals divided by the height of the NN interval histogram), total sleep time at night, deep sleep time at night, light sleep time at night, and wakefulness time at night.

[0030] Similarly, tn-1 represents the rehabilitation evaluation signal obtained at the previous moment, which is also a 1×12 matrix. α, β, γ, δ, etc. represent the weight values ​​of the rehabilitation evaluation signals at different moments. The setting rule of the weight value is as follows: the closer the rehabilitation evaluation signal is to the current moment, the greater its weight, and the distribution of the weight value changes exponentially.

[0031] In the postoperative rehabilitation evaluation module, the above rehabilitation evaluation signals can be obtained through wearable devices such as smart watches, smart bracelets, etc. After the data is acquired, the rehabilitation evaluation signals are analyzed by regression analysis based on choquet's fuzzy integral. The choquet integral of the fuzzy value fuzzy measure measures the uncertainty level of the fuzzy set by summarizing the possibility of each element in the above rehabilitation evaluation signal matrix, and finally outputs the patient's rehabilitation effect, which is a value between 0 and 100, where 0 represents no progress in the patient's rehabilitation and 100 represents the patient's optimal rehabilitation.

[0032] S4. According to the patient's postoperative rehabilitation effect, the weight factor of the neuron connection between different hidden layers in the signal processing module is automatically adjusted.

[0033] In deep learning or machine learning, each signal is called a "neuron". For different problems, the weight and bias (b) of each neuron are different. In a single-layer neural network, the input signal is first weighted and summed, and then the bias term is added, and then nonlinearly transformed by the activation function to finally get the output.

[0034] According to the patient's condition, the weights can be adjusted in time to obtain more accurate results.

[0035] The main function of the weight adaptive adjustment module is to automatically adjust the weight factors of the neuron connections between different hidden layers in the signal processing module according to the patient's postoperative rehabilitation effect, further optimize the rehabilitation plan, accelerate the patient's postoperative rehabilitation, and improve the patient's quality of life.

[0036] The input of this module consists of three parts: the first part is the output result of the postoperative rehabilitation evaluation module, that is, a value between 0 and 100; the second part is the output result of the postoperative rehabilitation management model (that is, the postoperative rehabilitation management recommendations output by the model), including the type of exercise on that day, exercise duration, exercise intensity, nutritional combination of each meal, interval time between each meal, total sleep time, sleep start time, and number of sleep times on that day; the third part is the patient's actual execution, such as the actual type of exercise on that day, exercise duration, etc.

[0037] In this module, the fish school algorithm is used to process the input data to obtain the weight factors when connecting neurons in different hidden layers. The steps are as follows: Step 1. Determine the dimension of the artificial fish: The dimension of the artificial fish is the input signal of the weight adaptive adjustment module, which is a 17×1 matrix, and the elements of the matrix are the above values; Step 2. Initialize the parameters of the fish school algorithm: The parameters that need to be initialized for the fish school algorithm include: population size fish_size, the perception distance of the artificial fish fish_distange, the maximum step length of the artificial fish movement fish_step, and the crowding factor , maximum number of iterations max_gen and target value G; Step 3. Set the initial number of iterations: When the calculation starts, the initial number of iterations is set to k=0, and fish_size artificial fish individuals are randomly generated in the feasible domain to form an initial fish school, and each component is a random number in the interval (−1,1); Step 4. Calculate the current position Y of each artificial fish in the initial fish school, compare the sizes, and retain the maximum value to enter the bulletin board; Step 5. Simulate the foraging behavior, grouping behavior and tail-chasing behavior of each artificial fish, and select the behavior with the larger Y value after the action to actually execute; Step 6. Update the bulletin board every iteration. If the distance of the artificial fish is greater than the Y previously set on the bulletin board, write the larger value to the bulletin board; Step 7. Termination condition judgment: judge whether the number of iterations has reached the maximum number of iterations or has met the accuracy error of the solution. If so, output the weight factors when connecting neurons in different hidden layers.

[0038] After adjusting the weight factors, the parameters are adjusted appropriately according to the patient's recovery status to obtain an objective and accurate postoperative evaluation. Real-time evaluation can also be obtained, which is beneficial for postoperative patients to adjust intervention measures in a timely manner, adjust diet, sleep, exercise and other factors, and speed up the recovery process.

[0039] like Figure 2 As shown in the figure, it is a postoperative rehabilitation management model based on artificial intelligence, which mainly includes five modules, namely input module, signal processing module, output module, postoperative rehabilitation evaluation module, and weight adaptive adjustment module. The main functions and implementation schemes of each module are as follows: An input module is used to input the input parameters of the postoperative rehabilitation management model, including: patient's age, gender, body mass index (BMI), blood pressure, blood sugar, surgical site, surgical type, and surgical time; The signal processing module processes and analyzes the input parameters of the postoperative rehabilitation management model through a deep neural network; Output module, outputs the patient's input parameters and suggestions for patient rehabilitation management; Postoperative rehabilitation evaluation module, output parameters include: the day's exercise type, exercise duration, exercise intensity, nutritional combination of each meal, interval between each meal, total sleep duration, sleep start time, number of sleep times on the day; The postoperative rehabilitation evaluation module provides real-time and objective evaluation of the patient's rehabilitation effect, thus providing support for the patient to adjust the rehabilitation plan; The weight adaptive adjustment module automatically adjusts the weight factors of the neuron connections between different hidden layers in the signal processing module according to the patient's postoperative rehabilitation effect.

[0040] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A postoperative rehabilitation management method, characterized in that: The steps include: Step 1: Input parameters to the postoperative rehabilitation management model, and output the patient's input parameters and the patient's rehabilitation management suggestions; Step 2: Processing and analyzing the input parameters through a deep neural network to obtain a rehabilitation evaluation signal; Step 3: Analyze the rehabilitation evaluation signal based on the regression analysis of Choquet's fuzzy integral and evaluate the rehabilitation effect.

2. The postoperative rehabilitation management method according to claim 1, characterized in that: The input parameters include: patient's age, gender, body mass index, blood pressure, blood sugar, surgical site, surgical type and surgical time.

3. The postoperative rehabilitation management method according to claim 1, characterized in that: The output parameters include: the type of exercise for the day, the duration of exercise, the intensity of exercise, the nutritional combination of each meal, the interval between each meal, the total sleep time, the start time of sleep and the number of sleep times for the day.

4. The postoperative rehabilitation management method according to claim 1, characterized in that: The input signal of the rehabilitation evaluation signal is {αt1, βt2, ..., γt n-1 ,δt n }, where t n Represents the rehabilitation evaluation signal obtained at the current moment, where: The values ​​of the 12 elements in the matrix are: the patient's body fat mass, visceral fat mass, body water mass, muscle mass, basal metabolic rate, SDNN value, RMSSD value, triangle index, total sleep time at night, deep sleep time at night, light sleep time at night, and wakefulness time at night.

5. The postoperative rehabilitation management method according to claim 4, characterized in that: The set of rehabilitation evaluation signals α, β, γ, δ represents the weight values ​​of rehabilitation evaluation signals at different times. The setting rule of the weight values ​​is as follows: the closer the rehabilitation evaluation signal is to the current moment, the greater its weight is, and the distribution of weight values ​​changes exponentially.

6. The postoperative rehabilitation management method according to claim 1, characterized in that: In step 3, after the data is acquired, the rehabilitation evaluation signal is analyzed by regression analysis based on Choquet's fuzzy integral. The Choquet integral of the fuzzy value fuzzy measure measures the uncertainty level of the fuzzy set by summarizing the possibility of each element in the above rehabilitation evaluation signal matrix, and finally outputs the patient's rehabilitation effect.

7. The postoperative rehabilitation management method according to claim 1, characterized in that: The method further includes step 4: automatically adjusting the weight factors of the neuron connections between different hidden layers in the signal processing module according to the patient's postoperative rehabilitation effect.

8. The postoperative rehabilitation management method according to claim 7, characterized in that: The method for processing the weight factor in step 4 is to process the input data by means of the fish school algorithm, so as to obtain the weight factor when the neurons of different hidden layers are connected, and the steps are as follows: Step 1. Determine the dimension of the artificial fish: The dimension of the artificial fish is the input signal of the weight adaptive adjustment module, which is a 17×1 matrix, and the elements of the matrix are the above values; Step 2. Initialize the parameters of the fish school algorithm: The parameters that need to be initialized for the fish school algorithm include: population size fish_size, the perception distance of the artificial fish fish_distange, the maximum step length of the artificial fish movement fish_step, and the crowding factor , maximum number of iterations max_gen and target value G; Step 3. Set the initial number of iterations: When the calculation starts, the initial number of iterations is set to k=0, and fish_size artificial fish individuals are randomly generated in the feasible domain to form an initial fish school, and each component is a random number in the interval (−1,1); Step 4. Calculate the current position Y of each artificial fish in the initial fish school, compare the sizes, and retain the maximum value to enter the bulletin board; Step 5. Simulate the foraging behavior, grouping behavior and tail-chasing behavior of each artificial fish, and select the behavior with the larger Y value after the action to actually execute; Step 6. Update the bulletin board every iteration. If the distance of the artificial fish is greater than the Y previously set on the bulletin board, write the larger value to the bulletin board; Step 7. Termination condition judgment: judge whether the number of iterations has reached the maximum number of iterations or has met the accuracy error of the solution. If so, output the weight factors when connecting neurons in different hidden layers.

9. A postoperative rehabilitation management system, characterized in that: The postoperative rehabilitation management method according to any one of claims 1 to 8 is implemented by a postoperative rehabilitation management model, wherein the postoperative rehabilitation management model comprises: An input module is used to input the input parameters of the postoperative rehabilitation management model, including: patient's age, gender, body mass index (BMI), blood pressure, blood sugar, surgical site, surgical type, and surgical time; The signal processing module processes and analyzes the input parameters of the postoperative rehabilitation management model through a deep neural network; Output module, outputs the patient's input parameters and suggestions for patient rehabilitation management; Postoperative rehabilitation evaluation module, output parameters include: the day's exercise type, exercise duration, exercise intensity, nutritional combination of each meal, interval between each meal, total sleep duration, sleep start time, number of sleep times on the day; The postoperative rehabilitation evaluation module provides real-time and objective evaluation of the patient's rehabilitation effect, thus providing support for the patient to adjust the rehabilitation plan; The weight adaptive adjustment module automatically adjusts the weight factors of the neuron connections between different hidden layers in the signal processing module according to the patient's postoperative rehabilitation effect.

Images