AI-based bone rehabilitation decision-making support system

Abstract

The present invention relates to an AI-based bone rehabilitation decision-making support system, and relates to the technical field of medical data processing, comprising: a sample acquisition module, used to acquire multidimensional information, rehabilitation programs, rehabilitation status information and rehabilitation progress information of multiple sample patients; a sample analysis module, used to determine multiple program-related factors, rehabilitation monitoring factors and rehabilitation evaluation factors; an information acquisition module, used to acquire multidimensional information of the current patient; a sample determination module, used to determine similar sample patients; a program generation module, used to generate a rehabilitation program and a rehabilitation monitoring program; an information collection module, used to collect rehabilitation status information; a progress determination module, used to determine rehabilitation progress information; a program adjustment module, used to adjust the rehabilitation monitoring program of the current patient through a program adjustment model, and has the advantages of realizing flexible adjustment of the bone rehabilitation program, real-time monitoring of the patient's rehabilitation status, and timely adjustment of the rehabilitation plan.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to an AI-based bone rehabilitation decision-making assistance system. Background Art

[0002] The rehabilitation process after orthopedic surgery is of great significance to the patient's physical function and mental health. With people's increasing attention to health management and demand for personalized medicine, personalized rehabilitation plans have received widespread attention and application in the field of orthopedic rehabilitation. Therefore, a personalized orthopedic rehabilitation plan recommendation system came into being.

[0003] In the existing technology, rehabilitation plans are often a universal standard program that cannot fully take into account the individual differences of each patient and the changes in the rehabilitation process. Obviously, this recommendation system has at least the following problems: 1. Failure to take into account the individual differences and special needs of each patient leads to unsatisfactory rehabilitation results. Patients often need a long rehabilitation process, and the rehabilitation results are not satisfactory. 2. Lack of data monitoring and analysis of patients during the rehabilitation process will lead to the inability to accurately determine the rehabilitation effect, and problems in the rehabilitation process cannot be dealt with in a timely manner. There is a lack of real-time monitoring and feedback mechanisms. Patients may face health risks during the rehabilitation process. The most common problem for patients during the rehabilitation process is the uncertainty of the rehabilitation cycle. Rehabilitation plans that lack data support make it difficult to determine the patient's rehabilitation stage and the next rehabilitation plan.

[0004] Therefore, it is necessary to provide an AI-based orthopedic rehabilitation decision-making support system to flexibly adjust the orthopedic rehabilitation plan, monitor the patient's rehabilitation status in real time, and adjust the rehabilitation plan in a timely manner. Summary of the invention

[0005] The present invention provides an AI-based bone rehabilitation decision-making support system, comprising: a sample acquisition module, used to acquire multidimensional information, rehabilitation schemes, rehabilitation status information and rehabilitation progress information of multiple sample patients; a sample analysis module, used to determine multiple scheme-related factors, rehabilitation monitoring factors and rehabilitation evaluation factors according to the multidimensional information, rehabilitation schemes, rehabilitation status information and rehabilitation progress information of multiple sample patients; the sample analysis module is also used to group multiple sample patients based on the rehabilitation schemes of the multiple sample patients to determine multiple patient groups; the sample analysis module is also used to determine, for each patient group, the target scheme-related factors, target rehabilitation monitoring factors and target rehabilitation evaluation factors corresponding to each patient group according to the multidimensional information, rehabilitation schemes, rehabilitation status information and rehabilitation progress information of each sample patient included in the patient group; an information acquisition module, used to acquire the multidimensional information of the current patient; a sample determination module, used to determine the target patient monitoring factor, target rehabilitation monitoring factor and target rehabilitation evaluation factor corresponding to the current patient based on the multidimensional information of the current patient and the target scheme-related factors corresponding to each patient group. a rehabilitation program for the current patient according to the multidimensional information of the current patient and the multidimensional information and rehabilitation programs of similar sample patients through a first program generation model; the program generation module is also used to generate a rehabilitation monitoring program for the current patient according to the target rehabilitation monitoring factors corresponding to the target patient group through a second program generation model; an information collection module is used to collect the rehabilitation status information of the current patient according to the rehabilitation monitoring program of the current patient; a progress determination module is used to determine the rehabilitation progress information of the current patient according to the rehabilitation status information of the current patient and the target rehabilitation evaluation factors corresponding to the target patient group; a program adjustment module is used to determine whether to adjust the program according to the rehabilitation progress information of the current patient and the rehabilitation progress information of similar sample patients; the program adjustment module is also used to adjust the rehabilitation monitoring program of the current patient according to the rehabilitation progress information of the current patient and the rehabilitation progress information of similar sample patients through the program adjustment model after determining to adjust the program.

[0006] Furthermore, the sample analysis module groups the multiple sample patients based on the rehabilitation programs of the multiple sample patients to determine the multiple patient groups, including: S11, for each of the sample patients, extracting multi-dimensional rehabilitation program features from the rehabilitation program of the sample patient, encoding the multi-dimensional rehabilitation program features, and generating a rehabilitation program feature matrix corresponding to the sample patient; S12, for any two of the sample patients, calculating the matrix similarity of the rehabilitation program feature matrices corresponding to the two sample patients; S13, for each of the sample patients, calculating the sample patient based on the matrix similarity of the rehabilitation program feature matrix corresponding to the sample patient and the rehabilitation program feature matrix corresponding to any one of the sample patients. S14, based on the central value corresponding to each of the sample patients, determine multiple candidate group central sample patients from the multiple sample patients; S15, for any one of the candidate group central sample patients, based on the matrix similarity between the rehabilitation program feature matrix corresponding to the candidate group central sample patient and the rehabilitation program feature matrix corresponding to any sample patient, generate a similarity sequence corresponding to the candidate group central sample patient; S16, for any two of the candidate group central sample patients, calculate the sequence similarity of the similarity sequences corresponding to the two candidate group central sample patients; S17, based on the sequence similarity of the similarity sequences corresponding to any two of the candidate group central sample patients, generate a similarity sequence corresponding to the candidate group central sample patients from the multiple sample patients. Screen multiple group center sample patients from candidate group center sample patients; S18, extract multiple target group center sample patients from multiple group center sample patients; S19, group multiple sample patients according to the matrix similarity between the rehabilitation program feature matrix corresponding to each sample patient and the rehabilitation program feature matrix corresponding to each target group center sample patient, and generate an intermediate grouping result. Specifically, the sample patients can be grouped into the intermediate patient group where the target group center sample patient with the largest matrix similarity is located; S19, group multiple sample patients according to the matrix similarity between the rehabilitation program feature matrix corresponding to each sample patient and the rehabilitation program feature matrix corresponding to each target group center sample patient, and generate an intermediate grouping result. If the intermediate grouping result satisfies the preset conditions, the sample patients can be grouped into the intermediate patient group where the target group center sample patient with the largest matrix similarity is located; S20, determine whether the intermediate grouping result meets the preset conditions, if so, determine multiple patient groups according to the intermediate grouping result, if not, execute S21, wherein the preset condition can be that the matrix similarity variance corresponding to each intermediate patient group is less than the matrix similarity variance threshold, and the matrix similarity variance corresponding to the intermediate patient group is the variance of the matrix similarity of the rehabilitation program feature matrices corresponding to any two sample patients included in the intermediate patient group; S21, whether the number of intermediate grouping results is greater than the number threshold, if so, execute S22, if not, execute S18; S22, adjust the group center sample patient and execute S19.

[0007] Furthermore, the sample analysis module adjusts the group center sample patients, including: S201, obtaining a plurality of intermediate grouping results; S202, determining a plurality of optimal group center sample patients from the plurality of group center sample patients according to the plurality of intermediate grouping results through a center adjustment model.

[0008] Furthermore, the sample analysis module determines the target scheme-related factors corresponding to each patient group according to the multidimensional information of each sample patient included in the patient group, including: for each patient group, according to the multidimensional information of each sample patient included in the patient group, calculating the correlation coefficient between each scheme-related factor and the patient group, and determining the target scheme-related factors corresponding to the patient group according to the correlation coefficient between each scheme-related factor and the patient group; the sample analysis module determines the target rehabilitation monitoring factors corresponding to the patient group according to the rehabilitation status information of each sample patient included in the patient group, including: for each patient group, according to the rehabilitation status information of each sample patient included in the patient group, determining the factor importance of each rehabilitation detection factor corresponding to the patient group, and determining the target rehabilitation monitoring factors corresponding to the patient group according to the factor importance of each rehabilitation detection factor corresponding to the patient group; the sample analysis module determines the target rehabilitation evaluation factors corresponding to the patient group according to the rehabilitation progress information of each sample patient included in the patient group, including: for each patient group, according to the rehabilitation status information of each sample patient included in the patient group, determining the factor importance of each rehabilitation evaluation factor corresponding to the patient group, and determining the target rehabilitation evaluation factors corresponding to the patient group according to the factor importance of each rehabilitation evaluation factor corresponding to the patient group.

[0009] Furthermore, the sample determination module determines the target patient group of the current patient based on the multidimensional information of the current patient and the target regimen-related factors corresponding to each patient group, including: for any two of the patient groups, calculating the factor overlap of the two patient groups according to the target regimen-related factors corresponding to the two patient groups; determining a plurality of patient units and the screening factors corresponding to each patient unit according to the factor overlap of any two patient groups, wherein each patient unit includes at least two patient groups; for each of the patient units, extracting a first feature vector from the multidimensional information of the current patient based on the screening factors corresponding to the patient unit, and calculating the matching degree between the current patient and the patient unit according to the first feature vector; determining the target patient unit according to the matching degree between the current patient and each of the patient units; for each of the patient groups included in the target patient unit, extracting a second feature vector from the multidimensional information of the current patient based on the target regimen-related factors corresponding to the patient group, and calculating the matching degree between the current patient and the patient groups included in the target patient unit according to the second feature vector; determining the target patient group of the current patient according to the matching degree between the current patient and each of the patient groups included in the target patient unit.

[0010] Furthermore, the sample determination module determines similar sample patients based on the multidimensional information of the current patient and the target regimen related factors corresponding to each patient group, including: extracting a third feature vector from the multidimensional information of the current patient based on the target regimen related factors corresponding to the target patient group of the current patient; extracting a fourth feature vector from the multidimensional information of each sample patient included in the target patient group of the current patient based on the target regimen related factors corresponding to the target patient group of the current patient; for each sample patient included in the target patient group of the current patient, calculating the information similarity between the current patient and the sample patient based on the third feature vector and the fourth feature vector corresponding to the sample patient; determining the similar sample patients based on the information similarity between the current patient and each sample patient included in the target patient group of the current patient.

[0011] Furthermore, the program generation module is also used to: when the rehabilitation program of the current patient includes physical therapy, determine the physical therapy parameters corresponding to the current patient according to the multi-dimensional information of the multiple sample patients through a parameter determination model.

[0012] Furthermore, the information collection module at least includes an activity monitoring unit, the activity monitoring unit includes a plurality of activity monitoring devices, and the current patient's rehabilitation monitoring plan includes the number of activity monitoring devices and the installation position of each activity monitoring device.

[0013] Furthermore, the information collection module includes at least a sleep monitoring unit and an emotion analysis unit, wherein the sleep monitoring unit is used to collect sleep state information of the current patient, and the emotion analysis unit is used to determine the emotional state of the current patient based on the collected sleep state information of the current patient.

[0014] Furthermore, the regimen adjustment module determines whether to adjust the regimen based on the rehabilitation progress information of the current patient and the rehabilitation progress information of similar sample patients, including: calculating a rehabilitation difference value based on the rehabilitation progress information of the current patient and the rehabilitation progress information of similar sample patients; and determining whether to adjust the regimen based on the rehabilitation difference value.

[0015] Compared with the prior art, the AI-based bone rehabilitation decision-making support system provided in this specification has at least the following beneficial effects:

[0016] 1. Through automation and intelligence, it is possible to quickly acquire, analyze and process the multi-dimensional information of patients, thereby generating personalized rehabilitation plans. This greatly reduces the time and energy of manual operation and improves the efficiency of rehabilitation work. The system can generate personalized rehabilitation plans based on the specific conditions of the patient, including age, gender, condition, physical condition, etc. This personalized plan can better meet the actual needs of patients and improve the rehabilitation effect. Through the information acquisition module and the progress determination module, the system can collect the patient's rehabilitation status information in real time, and accurately evaluate the patient's rehabilitation progress according to the target rehabilitation evaluation factor corresponding to the target patient group. This helps doctors to understand the patient's rehabilitation status in a timely manner and provide strong support for subsequent treatment and rehabilitation. The system can intelligently determine whether the rehabilitation plan needs to be adjusted based on the patient's rehabilitation progress information and the rehabilitation progress information of similar sample patients. When adjustment is needed, the system can intelligently adjust the rehabilitation monitoring plan through the plan adjustment model to ensure that the patient gets the most suitable rehabilitation plan for himself. Personalized rehabilitation plans and accurate rehabilitation monitoring and evaluation can enhance the patient's rehabilitation experience and increase the patient's satisfaction and trust in the rehabilitation process. This helps to establish a good doctor-patient relationship and improve the patient's enthusiasm for rehabilitation.

[0017] 2. By extracting the multi-dimensional rehabilitation program characteristics of sample patients and generating a feature matrix, the system can capture the differences and similarities of patient rehabilitation programs more comprehensively and accurately. Using matrix similarity calculation and central value calculation, the system can accurately group patients into patient groups with similar rehabilitation program characteristics, thereby improving the accuracy of grouping. Through grouping, the system can have a deeper understanding of the rehabilitation needs and characteristics of different patient groups. This helps the system provide each patient with a more personalized rehabilitation program to meet their unique rehabilitation needs. Detailed grouping steps and conditional judgments (such as matrix similarity variance thresholds and quantity thresholds) ensure the stability and reliability of the system grouping. This helps to reduce system misjudgments and incorrect groupings, and improve the overall efficiency and accuracy of the system;

[0018] 3. By calculating the overlap of factors between patient groups, the system can more accurately identify patient groups with similar rehabilitation needs. Furthermore, by extracting feature vectors and calculating the matching degree, the system can accurately assign the current patient to the target patient group that best suits his rehabilitation needs. Accurate target patient group matching means that the system can provide a more personalized rehabilitation plan for the current patient. This personalized plan can better meet the specific needs of the patient, thereby improving the rehabilitation effect. Within the target patient group, the system can accurately identify the sample patients who are most similar to the current patient by calculating the information similarity between the current patient and each sample patient. The rehabilitation plans and data of these similar sample patients provide valuable references for generating rehabilitation plans for the current patient. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:

[0020] Figure 1 It is a module diagram of an AI-based bone rehabilitation decision support system shown in an embodiment of the present application;

[0021] Figure 2 FIG. 4 is a flowchart of determining multiple patient groups shown in one embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following briefly introduces the drawings required for describing the embodiments.

[0023] Figure 1 is a module diagram of an AI-based bone rehabilitation decision support system shown in an embodiment of the present application, such as Figure 1As shown, the AI-based bone rehabilitation decision-making support system may include a sample acquisition module, a sample analysis module, an information acquisition module, a sample determination module, a plan generation module, an information collection module, a progress determination module and a plan adjustment module.

[0024] The sample acquisition module can be used to obtain multi-dimensional information, rehabilitation plans, rehabilitation status information and rehabilitation progress information of multiple sample patients.

[0025] The patient's multidimensional information can include:

[0026] Basic information: age, gender, weight, height, occupation, etc.

[0027] Medical history information: past medical history, surgical history, allergy history, etc.

[0028] Fracture information: fracture type (such as distal humeral fracture), fracture extent, surgical method, surgical date, medical images, etc.

[0029] Physical condition: physical condition before surgery, degree of pain after surgery, wound recovery, degree of swelling, etc.

[0030] Psychological state: expectations for recovery, anxiety level, need for psychological support, etc.

[0031] A rehabilitation program may include rehabilitation aids for patients at multiple rehabilitation nodes. For example, a rehabilitation program may include:

[0032] 1. Early rehabilitation (0-4 weeks after surgery):

[0033] Goal: Reduce swelling, relieve pain, and prevent complications.

[0034] Contents: Local immobilization, protection of the affected limb; pressure bandage; ice compress to relieve pain, constrict blood vessels, reduce bleeding and exudation; raise the affected limb to promote blood circulation and lymphatic return. Perform isometric contraction training of the injured limb muscles, and carry out early rehabilitation of non-injured parts to prevent secondary functional disorders.

[0035] 2. Mid-term rehabilitation (5 to 8 weeks after surgery):

[0036] Goal: Gradually restore range of motion, increase strength training, and reestablish neuromuscular control.

[0037] Contents: Raise the affected limb and maintain the correct posture; isometric contraction training; range of motion training of the distal end of the injured part and adjacent joints. At the same time, physical factors such as pulsed electromagnetic therapy, low-intensity pulsed ultrasound, and electrical stimulation therapy are used for treatment.

[0038] 3. Late rehabilitation:

[0039] Goal: Eliminate residual swelling, soften and stretch contracted fibrous tissue, increase joint range of motion and muscle strength, and restore muscle coordination.

[0040] Content: Continue to strengthen joint range of motion and muscle strength training, while performing aerobic endurance training, and encourage patients to carry out daily life activities, work and recreational activities. For patients with stiffness after periarticular fracture surgery, dynamic or static progressive braces may be considered. Continue to perform muscle strength and endurance training to restore physical fitness; perform proprioception and neuromuscular control to strengthen functional recovery.

[0041] Example recovery status information may include:

[0042] Pain level: The patient's pain level was assessed using a pain rating scale.

[0043] Swelling degree: Observe the swelling of the affected limb and record the reduction of swelling.

[0044] Range of motion: Measure the range of motion of the joint of the affected limb and compare it with the healthy side.

[0045] Muscle strength recovery: Muscle strength recovery of the affected limb is assessed through muscle strength testing.

[0046] Psychological state: Pay attention to changes in the patient's psychological state and assess their anxiety, depression and other emotional problems.

[0047] Example recovery progress information may include:

[0048] Pain relief: The patient's pain score gradually decreased and the pain intensity was relieved.

[0049] Swelling subsides: The swelling of the affected limb gradually subsides and the skin color returns to normal.

[0050] Increased range of joint motion: The range of joint motion of the affected limb gradually increases, approaching or reaching the level of the healthy side.

[0051] Muscle strength recovery: The muscle strength of the affected limb gradually recovers and can complete basic movements in daily life.

[0052] Improved mental state: Patients’ emotional problems such as anxiety and depression are alleviated, and their confidence in recovery is enhanced.

[0053] The sample analysis module can be used to determine multiple program-related factors, rehabilitation monitoring factors and rehabilitation evaluation factors based on the multi-dimensional information, rehabilitation programs, rehabilitation status information and rehabilitation progress information of multiple sample patients.

[0054] Specifically, multiple program-related factors, rehabilitation monitoring factors, and rehabilitation evaluation factors can be determined manually (for example, by experts) or by machine learning models based on multi-dimensional information, rehabilitation programs, rehabilitation status information, and rehabilitation progress information of multiple sample patients.

[0055] Example multiple regimen-related factors may include age, gender, physical condition, fracture type, etc.

[0056] Examples of multiple rehabilitation monitoring factors may include activity levels of key parts, sleep conditions, etc.

[0057] Exemplary multiple rehabilitation evaluation factors may include pain level, swelling level, range of joint motion, muscle strength recovery, etc.

[0058] Specifically, the patient's values ​​of multiple program-related factors, multiple rehabilitation monitoring factors and multiple rehabilitation evaluation factors can be determined according to the preset scoring rules corresponding to each program-related factor, rehabilitation monitoring factor and rehabilitation evaluation factor.

[0059] As an example only, different scores are set according to age groups, such as 10 points for 0-18 years old, 20 points for 19-45 years old, 15 points for 46-60 years old, and 10 points for over 60 years old.

[0060] 10 points for males and 20 points for females.

[0061] Based on the patient's overall health status, such as good, fair, and poor, the scores are set as 20 points, 15 points, and 10 points respectively.

[0062] According to the severity and complexity of the fracture, the corresponding score of the fracture type is determined, such as simple fracture, complex fracture, and comminuted fracture, which are set as 15 points, 10 points, and 5 points respectively.

[0063] Use the joint range of motion measurement tool to set scores based on the range of motion, such as completely normal is 20 points, mild limitation is 15 points, moderate limitation is 10 points, and severe limitation is 5 points.

[0064] According to the patient's sleep quality, such as good, average, and poor, it is set as 15 points, 10 points, and 5 points respectively.

[0065] Use a pain assessment scale, such as the Visual Analogue Scale, to determine the corresponding score for pain severity, with 0 for no pain and 10 for severe pain.

[0066] Depending on the severity of the swelling, such as mild, moderate, and severe, it is set as 15 points, 10 points, and 5 points respectively.

[0067] The range of motion of joints is similar to the range of motion of key parts, but it focuses more on the assessment of the overall joint. The score setting can refer to the range of motion of key parts.

[0068] According to the muscle strength assessment scale, such as the Manual Muscle Test (MMT), level 0 is 0 points, level 1 is 5 points, level 2 is 10 points, and so on, level 5 is 25 points.

[0069] The sample analysis module can also be used to group multiple sample patients based on their rehabilitation programs to determine multiple patient groups.

[0070] Figure 2 is a flowchart of determining multiple patient groups shown in an embodiment of the present application, such as Figure 2 As shown, specifically including:

[0071] S11. For each sample patient, extract multi-dimensional rehabilitation program features from the rehabilitation program of the sample patient, encode the multi-dimensional rehabilitation program features, and generate a rehabilitation program feature matrix corresponding to the sample patient;

[0072] S12. For any two sample patients, calculate the matrix similarity of the rehabilitation program feature matrices corresponding to the two sample patients;

[0073] S13. For each sample patient, based on the matrix similarity between the rehabilitation program feature matrix corresponding to the sample patient and the rehabilitation program feature matrix corresponding to any sample patient, calculate the central value corresponding to the sample patient;

[0074] S14, determining a plurality of candidate group central sample patients from a plurality of sample patients based on the central value corresponding to each sample patient;

[0075] S15. For any candidate group central sample patient, based on the matrix similarity between the rehabilitation program feature matrix corresponding to the candidate group central sample patient and the rehabilitation program feature matrix corresponding to any sample patient, generate a similarity sequence corresponding to the candidate group central sample patient, wherein an element of the similarity sequence corresponding to the candidate group central sample patient represents the matrix similarity between a sample patient and the candidate group central sample patient;

[0076] S16. For any two candidate group central sample patients, the sequence similarity of the similarity sequences corresponding to the two candidate group central sample patients is calculated. For example, the cosine similarity of the similarity sequences corresponding to the two candidate group central sample patients can be calculated as the sequence similarity;

[0077] S17, based on the sequence similarity of the similarity sequences corresponding to any two candidate group center sample patients, multiple group center sample patients are screened from multiple candidate group center sample patients. For example, if there are two candidate group center sample patients whose sequence similarity of the similarity sequences corresponding to the two candidate group center sample patients is greater than the sequence similarity threshold, one of the candidate group center sample patients is retained as the group center sample patient;

[0078] S18, extracting multiple target group center sample patients from multiple group center sample patients;

[0079] S19, grouping multiple sample patients according to the matrix similarity between the rehabilitation program feature matrix corresponding to each sample patient and the rehabilitation program feature matrix corresponding to each target group central sample patient, and generating an intermediate grouping result. Specifically, the sample patients may be grouped into the intermediate patient group where the target group central sample patient with the greatest matrix similarity is located;

[0080] S20, judging whether the intermediate grouping result meets the preset condition, if so, determining multiple patient groups according to the intermediate grouping result, if not, executing S21, wherein the preset condition may be that the matrix similarity variance corresponding to each intermediate patient group is less than the matrix similarity variance threshold, and the matrix similarity variance corresponding to the intermediate patient group is the variance of the matrix similarity of the rehabilitation program feature matrices corresponding to any two sample patients included in the intermediate patient group;

[0081] S21: Is the number of intermediate grouping results greater than the number threshold? If so, execute S22; if not, execute S18;

[0082] S22. Adjust the group center sample patients and execute S19.

[0083] Specifically, for the rehabilitation program features of different dimensions, the corresponding coding methods are determined. As an example only, the coding rules may be:

[0084] 1. Type of treatment plan:

[0085] Physical therapy: coded as [1, 0, 0] (assuming there are three types of treatment options, namely physical therapy, drug therapy and psychotherapy, represented by one-hot encoding)

[0086] Medication: coded as [0, 1, 0]

[0087] Psychotherapy: coded as [0, 0, 1]

[0088] 2. Drug use:

[0089] Use drug A: coded as [0, 0, 1]

[0090] Use drug B: coded as [0, 1, 0]

[0091] Types of rehabilitation activities:

[0092] 3. Physical therapy: coded as [1, 0, 0] (assuming there are three types of rehabilitation activities, namely physical therapy, massage therapy and speech therapy, also represented by one-hot encoding)

[0093] Massage therapy: coded as [0, 1, 0]

[0094] Speech therapy: coded as [0, 0, 1]

[0095] A row vector of the rehabilitation program feature matrix is ​​an encoding of a one-dimensional rehabilitation program feature.

[0096] The matrix similarity of the rehabilitation program feature matrices corresponding to two sample patients can be calculated according to the following formula:

[0097]

[0098] in, is the matrix similarity between the rehabilitation program feature matrix corresponding to the i-th sample patient and the rehabilitation program feature matrix corresponding to the j-th sample patient, is the total number of dimensions, is the kth row vector of the rehabilitation program feature matrix corresponding to the i-th sample patient, is the kth row vector of the rehabilitation program feature matrix corresponding to the jth sample patient, It is the cosine similarity between the kth row vector of the rehabilitation program feature matrix corresponding to the i-th sample patient and the kth row vector of the rehabilitation program feature matrix corresponding to the j-th sample patient.

[0099] The central value corresponding to the sample patient can be calculated according to the following formula:

[0100]

[0101] in, is the central value corresponding to the i-th sample patient, is the total number of sample patients.

[0102] Sample patients whose center values ​​are greater than the center value threshold can be used as center sample patients of the candidate group.

[0103] Preferably, the sample analysis module adjusts the group center sample patient, including:

[0104] S201, obtaining multiple intermediate grouping results;

[0105] S202. Determine multiple optimal group center sample patients from multiple group center sample patients according to multiple intermediate grouping results through a center adjustment model.

[0106] The center adjustment model may be a convolutional neural network (CNN) model. The input of the center adjustment model may include multiple intermediate grouping results, wherein the intermediate grouping results may include the extracted central sample patients of each target group and the matrix similarity variance corresponding to each intermediate patient group. The output of the center adjustment model may include multiple optimal group central sample patients. The multiple optimal group central sample patients are used as multiple target group central sample patients to execute S19.

[0107] The sample analysis module can also be used to determine the target program-related factors, target rehabilitation monitoring factors and target rehabilitation evaluation factors corresponding to each patient group based on the multidimensional information, rehabilitation program, rehabilitation status information and rehabilitation progress information of each sample patient included in the patient group.

[0108] Preferably, the sample analysis module determines the target regimen-related factors corresponding to each patient group according to the multi-dimensional information of each sample patient included in the patient group, including:

[0109] For each patient group, the correlation coefficient between each regimen-related factor and the patient group is calculated according to the multidimensional information of each sample patient included in the patient group, and the target regimen-related factor corresponding to the patient group is determined according to the correlation coefficient between each regimen-related factor and the patient group. For example, the regimen-related factor with a correlation coefficient greater than a correlation coefficient threshold can be used as the target regimen-related factor corresponding to the patient group;

[0110] The sample analysis module determines the target rehabilitation monitoring factor corresponding to the patient group according to the rehabilitation status information of each sample patient included in the patient group, including:

[0111] For each patient group, the factor importance of each rehabilitation detection factor corresponding to the patient group is determined according to the rehabilitation status information of each sample patient included in the patient group, and the target rehabilitation monitoring factor corresponding to the patient group is determined according to the factor importance of each rehabilitation detection factor corresponding to the patient group;

[0112] The sample analysis module determines the target rehabilitation evaluation factor corresponding to the patient group according to the rehabilitation progress information of each sample patient included in the patient group, including:

[0113] For each patient group, the factor importance of each rehabilitation evaluation factor corresponding to the patient group is determined based on the rehabilitation status information of each sample patient included in the patient group. Based on the factor importance of each rehabilitation evaluation factor corresponding to the patient group, the target rehabilitation evaluation factor corresponding to the patient group is determined.

[0114] Specifically, the correlation coefficient between the regimen-related factor and the patient group can be calculated according to the following formula:

[0115]

[0116] in, is the correlation coefficient between the g-th regimen-related factor and the h-th patient group, -1 is the total number of sample patients in the hth patient group except the sample patients in the group center, is the value of the g-th scheme-related factor for the i-th sample patient included in the h-th patient group, It is the matrix similarity between the rehabilitation program feature matrix corresponding to the i-th sample patient included in the h-th patient group and the rehabilitation program feature matrix corresponding to the group center sample patient of the h-th patient group.

[0117] The factor importance of each rehabilitation detection factor corresponding to the patient group can be determined according to the rehabilitation status information of each sample patient included in the patient group through manual or machine learning models, and the rehabilitation detection factor whose factor importance is greater than the first factor importance threshold is used as the target rehabilitation monitoring factor corresponding to the patient group.

[0118] The factor importance of each rehabilitation evaluation factor corresponding to the patient group can be determined according to the rehabilitation status information of each sample patient included in the patient group through manual or machine learning models, and the rehabilitation evaluation factor whose factor importance is greater than the second factor importance threshold is used as the target rehabilitation evaluation factor corresponding to the patient group.

[0119] The information acquisition module can be used to obtain multi-dimensional information of the current patient.

[0120] The sample determination module can be used to determine the target patient group and similar sample patients of the current patient based on the multi-dimensional information of the current patient and the target regimen-related factors corresponding to each patient group.

[0121] Preferably, the sample determination module determines the target patient group of the current patient based on the multi-dimensional information of the current patient and the target regimen-related factors corresponding to each patient group, including:

[0122] For any two patient groups, the factor overlap between the two patient groups was calculated based on the target scheme-related factors corresponding to the two patient groups;

[0123] According to the factor overlap between any two patient groups, multiple patient units and the screening factors corresponding to each patient unit are determined, wherein each patient unit includes at least two patient groups, and the intersection of the target scheme related factors corresponding to each patient group included in the patient unit can be used as the screening factor corresponding to the patient unit, and two patient groups whose factor overlap is greater than the factor overlap threshold can be grouped into the same patient unit;

[0124] For each patient unit, based on the screening factor corresponding to the patient unit, a first feature vector is extracted from the multidimensional information of the current patient, and the matching degree between the current patient and the patient unit is calculated according to the first feature vector, wherein the first feature vector may include the value of the screening factor corresponding to the current patient in the patient unit, a sample patient may be sampled from the patient unit, and a first comparison feature vector is extracted from the multidimensional information of the sample patient according to the screening factor corresponding to the patient unit, and the cosine similarity between the first feature vector and the first comparison feature vector is calculated as the matching degree between the current patient and the patient unit;

[0125] Determine a target patient unit according to the matching degree between the current patient and each patient unit. For example, a patient unit whose matching degree is greater than a first matching degree threshold may be used as the target patient unit.

[0126] For each patient group included in the target patient unit, based on the target regimen related factors corresponding to the patient group, a second feature vector is extracted from the multidimensional information of the current patient, and according to the second feature vector, a matching degree between the current patient and the patient group included in the target patient unit is calculated, wherein the second feature vector may include the value of the target regimen related factors corresponding to the current patient in the patient group, a sample patient may be sampled from the patient group, and according to the target regimen related factors corresponding to the patient group, a second comparison feature vector is extracted from the multidimensional information of the sample patient, and the cosine similarity between the second feature vector and the second comparison feature vector is calculated as the matching degree between the current patient and the patient group included in the target patient unit;

[0127] The target patient group of the current patient is determined according to the matching degree between the current patient and each patient group included in the target patient unit. For example, the patient group included in the target patient unit whose matching degree is greater than the second matching degree threshold can be used as the target patient group of the current patient.

[0128] Specifically, the factor overlap between two patient groups can be calculated according to the following formula:

[0129]

[0130] in, is the factor overlap between the i-th patient group and the j-th patient group, is the number of target regimen-related factors included in the intersection of the target regimen-related factors corresponding to the i-th patient group and the target regimen-related factors corresponding to the j-th patient group, is the number of target regimen-related factors corresponding to the i-th patient group, is the number of target regimen-related factors corresponding to the jth patient group.

[0131] Preferably, the sample determination module determines similar sample patients based on the multi-dimensional information of the current patient and the target scheme-related factors corresponding to each patient group, including:

[0132] Extracting a third feature vector from the multidimensional information of the current patient based on the target regimen related factor corresponding to the target patient group of the current patient, wherein the third feature vector may include the value of the target regimen related factor corresponding to the target patient group of the current patient;

[0133] Based on the target regimen related factor corresponding to the target patient group of the current patient, extracting a fourth feature vector from the multidimensional information of each sample patient included in the target patient group of the current patient, wherein the fourth feature vector may include the value of the target regimen related factor corresponding to the sample patient in the target patient group;

[0134] For each sample patient included in the target patient group of the current patient, the information similarity between the current patient and the sample patient is calculated according to the third eigenvector and the fourth eigenvector corresponding to the sample patient. For example, the cosine similarity between the third eigenvector and the fourth eigenvector corresponding to the sample patient can be calculated as the information similarity between the current patient and the sample patient.

[0135] Similar sample patients are determined based on the information similarity between the current patient and each sample patient included in the target patient group of the current patient. For example, sample patients included in the target patient group whose information similarity is greater than an information similarity threshold may be taken as similar sample patients.

[0136] The solution generation module can be used to generate a rehabilitation solution for the current patient according to the multidimensional information of the current patient and the multidimensional information and rehabilitation solutions of similar sample patients through a first solution generation model. The first solution generation model can be a convolutional neural network model.

[0137] Preferably, the solution generation module is also used for:

[0138] When the rehabilitation program of the current patient includes physical therapy, the physical therapy parameters corresponding to the current patient, such as the position and parameters of heat therapy, electrotherapy, and magnetic therapy, are determined by the parameter determination model according to the multidimensional information of multiple sample patients. The parameter determination model may be a convolutional neural network model.

[0139] The solution generation module can also be used to generate a rehabilitation monitoring solution for the current patient according to the target rehabilitation monitoring factor corresponding to the target patient group through the second solution generation model. The second solution generation model can be a convolutional neural network model.

[0140] The information collection module can be used to collect the rehabilitation status information of the current patient according to the rehabilitation monitoring plan of the current patient.

[0141] Preferably, the information collection module includes at least an activity monitoring unit, the activity monitoring unit includes a plurality of activity monitoring devices, and the current patient's rehabilitation monitoring plan includes the number of activity monitoring devices and the installation location of each activity monitoring device.

[0142] Preferably, the information collection module includes at least a sleep monitoring unit and an emotion analysis unit, wherein the sleep monitoring unit is used to collect the current patient's sleep state information (such as breathing rate, blood oxygen saturation, etc.), and the emotion analysis unit is used to determine the current patient's emotional state based on the collected current patient's sleep state information.

[0143] Specifically, the emotion analysis unit is a unit that further analyzes and processes the data collected by the sleep monitoring unit. Its main task is to infer the current emotional state of the patient based on the patient's sleep state information. There is a close connection between sleep quality and emotional state. For example, long-term lack of sleep or poor sleep quality may lead to negative emotions such as anxiety and depression. Certain physiological data (such as heart rate variability, breathing patterns, etc.) may be associated with specific emotional states. By analyzing these data, the emotion analysis unit can infer the patient's emotional state. The emotion analysis unit can use a machine learning algorithm to establish a prediction model by training a large amount of sleep data and emotion labels. The prediction model determines the current patient's emotional state based on the current patient's sleep state information.

[0144] The progress determination module can be used to determine the rehabilitation progress information of the current patient according to the rehabilitation status information of the current patient and the target rehabilitation evaluation factor corresponding to the target patient group.

[0145] Specifically, the rehabilitation progress information of the current patient can be determined by the rehabilitation progress evaluation model according to the rehabilitation status information of the current patient and the target rehabilitation evaluation factor corresponding to the target patient group. The rehabilitation progress evaluation model can be a convolutional neural network model.

[0146] The regimen adjustment module can be used to determine whether to adjust the regimen based on the rehabilitation progress information of the current patient and the rehabilitation progress information of similar sample patients.

[0147] Specifically include:

[0148] Calculate the rehabilitation difference value based on the rehabilitation progress information of the current patient and the rehabilitation progress information of similar sample patients;

[0149] Determine whether to adjust the plan based on the recovery difference value.

[0150] Specifically, the expected rehabilitation progress information of the current patient can be generated by the progress prediction model based on the rehabilitation progress information of similar sample patients. The difference between the rehabilitation progress information of the current patient and the expected rehabilitation progress information is calculated as a rehabilitation difference value. When the rehabilitation difference value is greater than the rehabilitation difference value threshold, it is determined that the plan should be adjusted.

[0151] The program adjustment module can also be used to adjust the rehabilitation monitoring program of the current patient according to the rehabilitation progress information of the current patient and the rehabilitation progress information of similar sample patients through the program adjustment model after determining to adjust the program. The program adjustment model can be a convolutional neural network model.

[0152] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.

Claims

1. The AI-based bone rehabilitation decision-making support system is characterized by: include: A sample acquisition module is used to obtain multi-dimensional information, rehabilitation plans, rehabilitation status information and rehabilitation progress information of multiple sample patients; A sample analysis module is used to determine multiple program-related factors, rehabilitation monitoring factors and rehabilitation evaluation factors based on the multi-dimensional information, rehabilitation programs, rehabilitation status information and rehabilitation progress information of multiple sample patients; The sample analysis module is also used to group the multiple sample patients based on the rehabilitation programs of the multiple sample patients to determine multiple patient groups; The sample analysis module is also used to determine, for each patient group, the target program-related factors, target rehabilitation monitoring factors and target rehabilitation evaluation factors corresponding to each patient group according to the multidimensional information, rehabilitation program, rehabilitation status information and rehabilitation progress information of each sample patient included in the patient group; Information acquisition module, used to obtain multi-dimensional information of the current patient; A sample determination module, used to determine a target patient group and similar sample patients of the current patient based on the multi-dimensional information of the current patient and target regimen-related factors corresponding to each patient group; A plan generation module, used for generating a rehabilitation plan for the current patient according to the multidimensional information of the current patient and the multidimensional information and rehabilitation plans of similar sample patients through a first plan generation model; The scheme generation module is also used to generate a rehabilitation monitoring scheme for the current patient according to the target rehabilitation monitoring factor corresponding to the target patient group through the second scheme generation model; An information collection module, used to collect the rehabilitation status information of the current patient according to the rehabilitation monitoring plan of the current patient; A progress determination module, used to determine the rehabilitation progress information of the current patient according to the rehabilitation status information of the current patient and the target rehabilitation evaluation factor corresponding to the target patient group; A program adjustment module is used to determine whether to adjust the program based on the rehabilitation progress information of the current patient and the rehabilitation progress information of similar sample patients; The scheme adjustment module is also used to adjust the rehabilitation monitoring scheme of the current patient according to the rehabilitation progress information of the current patient and the rehabilitation progress information of similar sample patients through the scheme adjustment model after determining to adjust the scheme; The sample analysis module groups the multiple sample patients based on the rehabilitation programs of the multiple sample patients to determine multiple patient groups, including: S11, for each of the sample patients, extracting multi-dimensional rehabilitation program features from the rehabilitation program of the sample patient, encoding the multi-dimensional rehabilitation program features, and generating a rehabilitation program feature matrix corresponding to the sample patient; S12. For any two of the sample patients, calculate the matrix similarity of the rehabilitation program feature matrices corresponding to the two sample patients; S13, for each of the sample patients, based on the matrix similarity between the rehabilitation program feature matrix corresponding to the sample patient and the rehabilitation program feature matrix corresponding to any sample patient, calculating the central value corresponding to the sample patient; S14, determining a plurality of candidate group central sample patients from the plurality of sample patients based on the central value corresponding to each of the sample patients; S15. For any one of the candidate group central sample patients, based on the matrix similarity between the rehabilitation program feature matrix corresponding to the candidate group central sample patient and the rehabilitation program feature matrix corresponding to any sample patient, generate a similarity sequence corresponding to the candidate group central sample patient; S16. For any two of the candidate group central sample patients, calculate the sequence similarity of the similarity sequences corresponding to the two candidate group central sample patients; S17, based on the sequence similarity of the similarity sequences corresponding to any two of the candidate group central sample patients, screening multiple group central sample patients from the multiple candidate group central sample patients; S18, extracting multiple target group center sample patients from multiple group center sample patients; S19, grouping multiple sample patients according to the matrix similarity between the rehabilitation program feature matrix corresponding to each sample patient and the rehabilitation program feature matrix corresponding to each target group central sample patient, generating an intermediate grouping result, specifically, grouping the sample patients into the intermediate patient group where the target group central sample patient with the greatest matrix similarity is located; S20, judging whether the intermediate grouping result meets the preset conditions, if so, determining multiple patient groups according to the intermediate grouping result, if not, executing S21, wherein the preset condition is that the matrix similarity variance corresponding to each intermediate patient group is less than the matrix similarity variance threshold, and the matrix similarity variance corresponding to the intermediate patient group is the variance of the matrix similarity of the rehabilitation program feature matrices corresponding to any two sample patients included in the intermediate patient group; S21: Is the number of intermediate grouping results greater than the number threshold? If so, execute S22; if not, execute S18; S22. Adjust the group center sample patients and execute S19.

2. The AI-based bone rehabilitation decision-making support system according to claim 1, characterized in that: The sample analysis module adjusts the group center sample patient, including: S201, obtaining multiple intermediate grouping results; S202. Determine a plurality of optimal group center sample patients from the plurality of group center sample patients according to a plurality of intermediate grouping results through a center adjustment model.

3. The AI-based bone rehabilitation decision-making support system according to claim 1, characterized in that: The sample analysis module determines the target scheme-related factors corresponding to each patient group according to the multi-dimensional information of each sample patient included in the patient group, including: For each of the patient groups, according to the multidimensional information of each sample patient included in the patient group, the correlation coefficient between each regimen-related factor and the patient group is calculated, and according to the correlation coefficient between each regimen-related factor and the patient group, the target regimen-related factor corresponding to the patient group is determined; The sample analysis module determines the target rehabilitation monitoring factor corresponding to the patient group according to the rehabilitation status information of each sample patient included in the patient group, including: For each of the patient groups, according to the rehabilitation status information of each sample patient included in the patient group, determine the factor importance of each rehabilitation detection factor corresponding to the patient group, and according to the factor importance of each rehabilitation detection factor corresponding to the patient group, determine the target rehabilitation monitoring factor corresponding to the patient group; The sample analysis module determines the target rehabilitation evaluation factor corresponding to the patient group according to the rehabilitation progress information of each sample patient included in the patient group, including: For each of the patient groups, the factor importance of each rehabilitation evaluation factor corresponding to the patient group is determined based on the rehabilitation status information of each sample patient included in the patient group, and the target rehabilitation evaluation factor corresponding to the patient group is determined based on the factor importance of each rehabilitation evaluation factor corresponding to the patient group.

4. The AI-based bone rehabilitation decision-making support system according to claim 3 is characterized in that: The sample determination module determines the target patient group of the current patient based on the multi-dimensional information of the current patient and the target regimen related factors corresponding to each patient group, including: For any two of the patient groups, the factor overlap between the two patient groups is calculated according to the target scheme-related factors corresponding to the two patient groups; Determine a plurality of patient units and a screening factor corresponding to each patient unit according to the factor overlap between any two of the patient groups, wherein each of the patient units includes at least two of the patient groups; For each of the patient units, based on the screening factor corresponding to the patient unit, extract a first feature vector from the multi-dimensional information of the current patient, and calculate the matching degree between the current patient and the patient unit according to the first feature vector; Determining a target patient unit according to a matching degree between the current patient and each of the patient units; For each patient group included in the target patient unit, based on the target regimen related factors corresponding to the patient group, extracting a second feature vector from the multidimensional information of the current patient, and calculating the matching degree between the current patient and the patient group included in the target patient unit according to the second feature vector; The target patient group of the current patient is determined according to the matching degree between the current patient and each patient group included in the target patient unit.

5. The AI-based bone rehabilitation decision-making support system according to claim 4, characterized in that: The sample determination module determines similar sample patients based on the multi-dimensional information of the current patient and the target scheme related factors corresponding to each patient group, including: extracting a third feature vector from the multidimensional information of the current patient based on the target regimen-related factor corresponding to the target patient group of the current patient; extracting a fourth feature vector from multidimensional information of each sample patient included in the target patient group of the current patient based on the target regimen-related factor corresponding to the target patient group of the current patient; For each sample patient included in the target patient group of the current patient, calculating the information similarity between the current patient and the sample patient according to the third feature vector and the fourth feature vector corresponding to the sample patient; The similar sample patients are determined according to the information similarity between the current patient and each sample patient included in the target patient group of the current patient.

6. The AI-based bone rehabilitation decision-making support system according to claim 3, characterized in that: The solution generation module is also used for: When the rehabilitation program of the current patient includes physical therapy, the physical therapy parameters corresponding to the current patient are determined according to the multi-dimensional information of the multiple sample patients through the parameter determination model.

7. The AI-based bone rehabilitation decision-making support system according to claim 3, characterized in that: The information collection module at least includes an activity monitoring unit, and the activity monitoring unit includes a plurality of activity monitoring devices. The rehabilitation monitoring plan of the current patient includes the number of activity monitoring devices and the installation position of each activity monitoring device.

8. The AI-based bone rehabilitation decision-making support system according to claim 3, characterized in that: The information collection module at least includes a sleep monitoring unit and an emotion analysis unit, wherein the sleep monitoring unit is used to collect sleep state information of the current patient, and the emotion analysis unit is used to determine the emotional state of the current patient based on the collected sleep state information of the current patient.

9. The AI-based bone rehabilitation decision-making support system according to claim 3, characterized in that: The program adjustment module determines whether to adjust the program according to the rehabilitation progress information of the current patient and the rehabilitation progress information of similar sample patients, including: Calculate the rehabilitation difference value based on the rehabilitation progress information of the current patient and the rehabilitation progress information of similar sample patients; Whether to adjust the plan is determined according to the rehabilitation difference value.

Images