A respiratory rehabilitation training effect evaluation method based on multi-modal data

Abstract

The present application relates to the technical field of respiratory rehabilitation training effect evaluation, and particularly relates to a respiratory rehabilitation training effect evaluation method based on multi-modal data. The method acquires multi-modal diagnosis and treatment data and multi-modal rehabilitation data; acquires reference historical patients of a current patient according to the multi-modal diagnosis and treatment data; divides rehabilitation stages based on the multi-modal rehabilitation data of the reference historical patients, and determines a current rehabilitation stage of the current patient; and acquires a current training effective degree to evaluate the respiratory rehabilitation training effect of the current patient according to the difference between the multi-modal rehabilitation data of the current patient and other rehabilitation stages before the current rehabilitation stage, and the difference between the multi-modal rehabilitation data of the reference historical patients in the current rehabilitation stage and other rehabilitation stages before the current rehabilitation stage. The present application is beneficial to real-time and accurate determination of the respiratory rehabilitation training effect of the current patient by acquiring the current training effective degree, and is further beneficial to timely adjustment of the rehabilitation training scheme, and effectively improves the treatment effect of the patient.

Description

Technical Field

[0001] This invention relates to the field of respiratory rehabilitation training effect evaluation technology, specifically to a method for evaluating the effect of respiratory rehabilitation training based on multimodal data. Background Technology

[0002] Respiratory rehabilitation is an important branch of rehabilitation medicine. Its core objective is to develop a comprehensive rehabilitation plan tailored to the individual needs of patients with lung diseases. This plan, developed through multidisciplinary collaboration, includes diagnosis, treatment, psychological support, and health education. The goal is to stabilize or reverse lung pathological damage, helping patients restore optimal respiratory function in daily life and improve their quality of life. Respiratory rehabilitation training has become a key means of improving patients' lung conditions, and the scientific rigor and accuracy of evaluating the effectiveness of respiratory rehabilitation training directly impact the optimization and adjustment of rehabilitation plans and the patient's recovery process.

[0003] Current methods for assessing the effectiveness of respiratory rehabilitation training use the patient's initial state before rehabilitation training as the sole baseline. The effectiveness is determined by comparing the data from a single rehabilitation training session with the initial baseline. However, in reality, rehabilitation training is a long-term and gradual process. Patients' rehabilitation outcomes exhibit significant stage-specific characteristics. While rehabilitation indicators fluctuate relatively little within the same stage, there are significant differences in indicator improvement between stages. Using only the initial state as a baseline for a single comparison fails to reflect the stage-specific progress in the rehabilitation process. Furthermore, relying solely on the patient's initial baseline cannot achieve cross-group comparisons of rehabilitation outcomes, making it difficult to determine the patient's current level of progress within a similar disease group. This results in a lack of reference value in the assessment results, affecting the targetedness and rationality of rehabilitation plan adjustments, ultimately leading to insufficient accuracy in rehabilitation outcome assessment and failing to meet the clinical needs of personalized rehabilitation guidance. Therefore, existing methods for assessing the effectiveness of respiratory rehabilitation training cannot capture the patient's progress in real time within a specific rehabilitation stage, resulting in an inability to accurately assess the effectiveness of respiratory rehabilitation training in real time. Consequently, timely and sensitive feedback on stage-specific training effects cannot be provided, impacting the patient's treatment outcome. Summary of the Invention

[0004] To address the technical problem that existing methods for evaluating the effectiveness of respiratory rehabilitation training cannot capture patients' progress in real time during specific rehabilitation stages, leading to inaccurate assessments, this invention aims to provide a method for evaluating the effectiveness of respiratory rehabilitation training based on multimodal data. The specific technical solution adopted is as follows: This invention provides a method for evaluating the effectiveness of respiratory rehabilitation training based on multimodal data, the method comprising the following steps: Acquire multimodal diagnostic and treatment data of current and historical patients before respiratory rehabilitation training, as well as multimodal rehabilitation data during respiratory rehabilitation training; Based on the similarity of multimodal diagnosis and treatment data of current patients and historical patients, reference historical patients are obtained for the current patient; based on the changes in multimodal rehabilitation data of reference historical patients, rehabilitation stages are divided to determine the current rehabilitation stage of the current patient; Based on the differences between the current multimodal rehabilitation data of the patient and the multimodal rehabilitation data of other rehabilitation stages before the current rehabilitation stage, and in conjunction with the differences between the current multimodal rehabilitation data of the patient and the multimodal rehabilitation data of other rehabilitation stages before the current rehabilitation stage, the current training effectiveness of the patient can be obtained. Assess the effectiveness of the current respiratory rehabilitation training for the current patient based on the current training effectiveness.

[0005] Furthermore, the method for obtaining the reference historical patients is as follows: The multimodal diagnosis and treatment data of the current patient and each historical patient are transformed into vectors, which serve as the diagnosis and treatment feature vectors for the corresponding patients; elements at the same position in the diagnosis and treatment feature vectors correspond to the same type of diagnosis and treatment data. The cosine similarity between the current patient's and each historical patient's diagnostic and treatment feature vectors is obtained and used as a reference analysis value. The reference analysis values ​​are arranged in descending order to obtain the reference sequence; The historical patients corresponding to the first preset number of reference analysis values ​​in the reference sequence are all used as reference historical patients for the current patient.

[0006] Furthermore, the method for obtaining the rehabilitation stage is as follows: For any reference historical patient, the degree of increase in the daily rehabilitation indicators of that reference historical patient is obtained based on the difference between the multimodal rehabilitation data of that reference historical patient and the previous adjacent day. When the increase in rehabilitation indicators exceeds the preset threshold for the increase in rehabilitation indicators, the corresponding day will be marked as a reference dividing day. Arrange the sequence numbers corresponding to the reference division days according to the time order to obtain the division sequence of the reference historical patients; The mean of the elements at the same position in the partition sequence of all reference historical patients is rounded down and used as the reference element at the corresponding position. The sequence formed by the reference elements is used as the target partitioning sequence; The average number of days corresponding to each element in the target segmentation sequence is taken as the target segmentation day. The segmentation is performed between the target segmentation day and its preceding adjacent day to obtain the recovery stage.

[0007] Furthermore, the method for obtaining the degree of increase in the rehabilitation indicators is as follows: For any day and any rehabilitation data during the respiratory rehabilitation training process of any reference historical patient, the mean of the rehabilitation data for that day of the reference historical patient shall be used as the reference value of the rehabilitation data for that day of the reference historical patient. The difference between the reference value of the rehabilitation data of the reference patient on that day and the reference value of the adjacent day before that day is used as the incremental analysis value of the rehabilitation data of the reference patient on that day. The increase values ​​of each rehabilitation data of the reference historical patient on that day were dimensionless. The dimensionless increase values ​​were summed and normalized. The result was used as the degree of increase of the rehabilitation index of the reference historical patient on that day.

[0008] Furthermore, the method for obtaining the current training effectiveness is as follows: Based on the differences between the current multimodal rehabilitation data of the patient and the multimodal rehabilitation data of other rehabilitation stages before the current rehabilitation stage, obtain the current rehabilitation feature difference vector and training effect analysis value of the patient. Based on the differences in multimodal rehabilitation data between the current rehabilitation stage of the reference historical patient and other previous rehabilitation stages, obtain the comparative rehabilitation feature difference vector of the current rehabilitation stage of the reference historical patient; The result of normalizing the cosine similarity between the current patient's current rehabilitation feature difference vector and the compared rehabilitation feature difference vector is used as the reference similarity weight. The normalized result of the product of the reference similarity weight and the training effect analysis value is taken as the current training effectiveness of the current patient.

[0009] Furthermore, the method for obtaining the current patient's current rehabilitation feature difference vector is as follows: The reference values ​​of all types of rehabilitation data for the current patient each day are constructed into a vector, which serves as the rehabilitation feature vector for the current patient each day; where elements at the same position in the rehabilitation feature vector correspond to the same type of data. For any rehabilitation stage, the average of the rehabilitation feature vectors of all days within that rehabilitation stage is used as the reference vector for that rehabilitation stage. All rehabilitation stages that precede the current rehabilitation stage in time sequence are considered as historical rehabilitation stages; The average of the difference vectors between the current patient's current day's rehabilitation feature vector and the reference vectors of each historical rehabilitation stage is taken as the current patient's current rehabilitation feature difference vector.

[0010] Furthermore, the method for obtaining the training effect analysis value is as follows: After independently removing the original physical dimensions from each element in the current patient's current rehabilitation characteristic difference vector, the average of the processed elements is used as the current patient's current training effect analysis value.

[0011] Furthermore, the method for obtaining the comparison and rehabilitation feature difference vector is as follows: The vector corresponding to the average of the reference vectors of all patients in the current rehabilitation stage is used as the overall historical reference vector for the current rehabilitation stage. The vector obtained by averaging the difference vectors of the overall historical reference vectors of each historical rehabilitation stage and the overall historical reference vector of the current rehabilitation stage is used as the comparison rehabilitation feature difference vector of the current rehabilitation stage of the reference historical patients.

[0012] Furthermore, the method for assessing the current respiratory rehabilitation training effect of the current patient based on the current training effectiveness is as follows: When the current training effectiveness exceeds the preset training effectiveness threshold, assess the effectiveness of the current respiratory rehabilitation training for the current patient. When the current training effectiveness is less than or equal to the preset training effectiveness threshold, the current respiratory rehabilitation training effect of the patient is considered poor.

[0013] Furthermore, the types of data in the multimodal diagnosis and treatment data include, but are not limited to, the types of data in the multimodal rehabilitation data.

[0014] The present invention has the following beneficial effects: This invention first identifies reference historical patients based on the similarity of multimodal diagnostic and treatment data of the current patient and historical patients. This facilitates accurate identification of the current patient's rehabilitation stage and provides a valuable baseline for evaluating the current patient's training effectiveness, avoiding the limitations of relying solely on the current patient's initial state. Next, it accurately divides the rehabilitation stage based on changes in the multimodal rehabilitation data of the reference historical patients, thus accurately determining the current patient's current rehabilitation stage. This facilitates accurate and efficient evaluation of the current patient's respiratory rehabilitation training effectiveness. Furthermore, it assesses the current patient's training effectiveness by comparing the differences between the current patient's multimodal rehabilitation data and the multimodal rehabilitation data of previous rehabilitation stages, combined with the differences in the multimodal rehabilitation data of the reference historical patients. This yields the current training effectiveness level, accurately reflecting the current patient's respiratory rehabilitation training effect and effectively improving the accuracy and scientific rigor of subsequent evaluations. Finally, it accurately evaluates the current patient's respiratory rehabilitation training effect based on the current training effectiveness level, enabling timely and sensitive feedback on the staged training results. This allows for timely adjustments to the rehabilitation training plan, providing reliable support for the dynamic optimization of the clinical rehabilitation plan in real time and effectively improving the patient's treatment outcome. Attached Figure Description

[0015] Figure 1 This is a schematic flowchart illustrating a method for evaluating the effectiveness of respiratory rehabilitation training based on multimodal data, provided in one embodiment of the present invention. Figure 2 A flowchart illustrating a method for obtaining the current training effectiveness according to an embodiment of the present invention; Figure 3 This is a structural diagram of a respiratory rehabilitation training effect evaluation system based on multimodal data, provided in one embodiment of the present invention. Figure 4 This is a schematic diagram of a computer device provided according to an embodiment of the present invention. Detailed Implementation

[0016] The following description, in conjunction with the accompanying drawings, details a specific scheme for evaluating the effectiveness of respiratory rehabilitation training based on multimodal data, provided by this invention.

[0017] Example 1: This invention proposes a method for evaluating the effectiveness of respiratory rehabilitation training based on multimodal data. Please refer to [link / reference]. Figure 1 The diagram illustrates a schematic flowchart of a method for evaluating the effectiveness of respiratory rehabilitation training based on multimodal data, according to an embodiment of the present invention. The method includes the following steps: Step S1: Obtain multimodal diagnostic and treatment data of the current patient and historical patients before respiratory rehabilitation training, as well as multimodal rehabilitation data during respiratory rehabilitation training.

[0018] Specifically, to accurately, comprehensively, and in real-time analyze the current respiratory rehabilitation training effect of patients, this embodiment acquires two types of data through an electronic medical record database: multimodal diagnostic and treatment data before respiratory rehabilitation training and multimodal rehabilitation data during respiratory rehabilitation training. Multimodal diagnostic and treatment data refers to the patient's clinical data before starting respiratory rehabilitation training, used to assess the patient's initial state, including but not limited to medical diagnoses, patient age, patient gender, and baseline lung function indicators (such as forced expiratory volume, vital capacity, and forced vital capacity). Multimodal rehabilitation data refers to data collected periodically or continuously during respiratory rehabilitation training to monitor the training progress and effect, including but not limited to process indicators such as lung function indicators and endurance test duration measured during training. The types of data in multimodal diagnostic and treatment data include, but are not limited to, the types of data in multimodal rehabilitation data. This embodiment sets the acquisition of the patient's multimodal rehabilitation data once every other day. Implementers can set the acquisition interval of the patient's multimodal rehabilitation data according to actual conditions; it is not limited here.

[0019] It should be noted that, to ensure data comparability, the data extracted from current patients and historical patients are consistent in type, structure, and format. Historical patients are those who have fully recovered through respiratory rehabilitation training. This embodiment sets the number of historical patients at 5000, but implementers can adjust the number of historical patients according to actual circumstances; no limitation is imposed here.

[0020] Step S2: Based on the similarity of the multimodal diagnosis and treatment data of the current patient and historical patients, obtain the reference historical patients for the current patient; based on the changes in the multimodal rehabilitation data of the reference historical patients, divide the rehabilitation stages and determine the current rehabilitation stage of the current patient.

[0021] Specifically, respiratory rehabilitation training is a long-term and gradual process that requires multiple, regular, and scientific training sessions to achieve significant improvement. Therefore, the changes in a patient's rehabilitation outcomes exhibit distinct stage-specific characteristics: within the same rehabilitation stage, the fluctuations in various rehabilitation evaluation indicators are relatively small, while significant changes occur between different rehabilitation stages. It is known that patients with similar degrees of lung injury and physiological conditions generally show consistency in their rehabilitation potential and stage evolution patterns when dividing rehabilitation stages. Considering that complete rehabilitation data from historical patients is readily available, clearly defining the different rehabilitation stages, the rehabilitation stage characteristics of historical patients with similar lung injury and physiological conditions can be used to reasonably predict the current patient's rehabilitation stage. This allows for the identification of each rehabilitation stage for the current patient. Furthermore, by extracting the core respiratory indicator characteristics of the current patient at each rehabilitation stage and combining this with the changing trends of different rehabilitation stages, the effectiveness of the current patient's respiratory rehabilitation training can be accurately and in real-time assessed.

[0022] The patient's multimodal diagnostic and treatment data includes the patient's lung function indicators and physiological information. Therefore, in this embodiment, based on the similarity between the multimodal diagnostic and treatment data of the current patient and historical patients, a reference historical patient is obtained for the current patient, that is, a historical patient with similar lung damage and physiological conditions to the current patient. Furthermore, the rehabilitation stage is divided based on the changes in the multimodal rehabilitation data of the reference historical patient to determine the current rehabilitation stage of the current patient, which is conducive to the accurate analysis of the current respiratory rehabilitation training effect of the current patient.

[0023] Preferably, in one possible implementation of this embodiment, the method for obtaining reference historical patients is as follows: First, the multimodal diagnosis and treatment data of the current patient and each historical patient are converted into vectors, serving as the corresponding patient's diagnosis and treatment feature vectors; wherein, elements at the same position in the diagnosis and treatment feature vectors correspond to the same type of diagnosis and treatment data; it should be noted that for non-data type diagnosis and treatment information, a keyword library specifically for the field of respiratory rehabilitation can be constructed, and feature extraction and quantification can be performed using a bag-of-words model, converting it into a vector or numerical data form that can be used for calculation. The method for converting non-data type information into vectors or numerical data is a well-known technique and will not be elaborated further. Furthermore, to ensure the stability of data analysis, this embodiment requires that all data types in the multimodal diagnosis and treatment data be standardized first, so that the value range of each data type is consistent and the dimensions are eliminated; for example, taking vital capacity as an example for analysis, the vital capacity of all historical patients is used as a dataset, and the vital capacity is normalized using the min-max normalization method. The min-max normalization method is a well-known technique and will not be elaborated further. Then, the cosine similarity between the diagnostic feature vectors of the current patient and each historical patient is obtained, and these are all used as reference analysis values. The reference analysis values ​​are arranged in descending order to obtain a reference sequence. The historical patients corresponding to the first preset number of reference analysis values ​​in the reference sequence are all used as reference historical patients for the current patient. In this embodiment, the preset number is set to 30. Implementers can set the size of the preset number according to the actual situation, and it is not limited here. The cosine similarity is well known and will not be described in detail here.

[0024] Preferably, in one possible implementation of this embodiment, the method for obtaining the rehabilitation stage is as follows: For any reference historical patient, based on the difference between the multimodal rehabilitation data of the reference historical patient each day and the previous adjacent day, the degree of increase in the rehabilitation indicators of the reference historical patient each day is obtained; wherein, the method for obtaining the degree of increase in the rehabilitation indicators is as follows: For any day and any rehabilitation data in the respiratory rehabilitation training process of any reference historical patient, the mean value of the rehabilitation data of the reference historical patient on that day is used as the reference value of the rehabilitation data of the reference historical patient on that day; then, the difference between the reference value of the rehabilitation data of the reference historical patient on that day and the reference value of the rehabilitation data of the previous adjacent day is used as the increase analysis value of the rehabilitation data of the reference historical patient on that day; considering that the dimensions of the increase analysis values ​​of different rehabilitation data items are different, the increase analysis values ​​of each rehabilitation data item of the reference historical patient on that day are respectively dimensionless, and the sum of the obtained dimensionless increase analysis values ​​and the result of normalization are used as the degree of increase in the rehabilitation indicators of the reference historical patient on that day. This embodiment divides the increase analysis value of each rehabilitation data point for the day by the preset unit value of each rehabilitation data point (set according to actual situation, not limited here) to achieve dimensionless measurement. This embodiment normalizes the sum of the above increase analysis values ​​using the maximum-minimum normalization method, where the dataset corresponding to the maximum-minimum normalization method is the sum of the daily increase analysis values ​​of rehabilitation indicators for all reference historical patients. It should be noted that the degree of increase in rehabilitation indicators on the first day of the respiratory rehabilitation training process for the reference historical patient is obtained by comparing the multimodal rehabilitation data on the first day of the respiratory rehabilitation training process for the reference historical patient with the multimodal rehabilitation data in the multimodal diagnosis and treatment data of the reference historical patient. The greater the increase in rehabilitation indicators, the more likely the corresponding day is to be the stage division day for the reference historical patient. Therefore, this embodiment sets a preset threshold for the increase in rehabilitation indicators of 0.6. Implementers can set the size of the preset threshold for the increase in rehabilitation indicators according to actual conditions, which is not limited here. When the increase in rehabilitation indicators is greater than the preset threshold, the corresponding day is marked as a reference division day; then, the sequence numbers corresponding to the reference division days are arranged in chronological order to obtain the division sequence of the reference historical patient. It should be noted that the sequence numbers of each day during the respiratory rehabilitation training of the reference historical patient are 1, 2, 3... in chronological order. In order to predict the current rehabilitation stage classification of the patient, this embodiment uses the mean of the elements at the same position in the classification sequence of all reference historical patients as the reference element for the corresponding position, which is rounded down. It should be noted that when calculating the mean of the elements at the same position in the classification sequence of all reference historical patients, if some patients do not have the element corresponding to a certain position due to the short span of the classification sequence, the missing position is excluded from the sum of the numerator of the mean of the corresponding position, and the actual number of patients with the objective element belonging to that position is used as the denominator for the average calculation. The sequence of reference elements is further used as the target segmentation sequence. Finally, the average day corresponding to each element in the target segmentation sequence is used as the target segmentation day. The rehabilitation stage is obtained by dividing the target segmentation day into its preceding adjacent day. For example, if the target segmentation sequence is {5,10,15,19,24}, the rehabilitation stages are, in order, day 1 to day 4, day 5 to day 9, day 10 to day 14, day 15 to day 18, day 19 to day 23, and day 24 until the end of the respiratory rehabilitation training.

[0025] It should be noted that when a reference patient's partition sequence contains consecutive, uninterrupted sequence numbers, the first occurrence of the consecutive, uninterrupted sequence number is retained, while the other consecutive, uninterrupted sequence numbers are deleted. For example, if the original partition sequence of the reference patient is {5,10,11,15,17,23}, then the partition sequence used in subsequent analyses after adjustment will be {5,10,15,17,23}.

[0026] After dividing the rehabilitation stages, the current rehabilitation stage of the current patient is accurately determined by the sequential number of the current day.

[0027] Step S3: Based on the differences between the current multimodal rehabilitation data of the current patient and the multimodal rehabilitation data of other rehabilitation stages before the current rehabilitation stage, and in conjunction with the differences between the current multimodal rehabilitation data of the current patient and the multimodal rehabilitation data of other patients before the current rehabilitation stage, obtain the current training effectiveness of the current patient.

[0028] Specifically, by comparing the multimodal rehabilitation data of a patient at different rehabilitation stages, a precise assessment of the effectiveness of their respiratory rehabilitation training can be achieved. Since the effectiveness of respiratory rehabilitation training gradually improves with each stage of rehabilitation, the greater the improvement in the current patient's multimodal rehabilitation data compared to previous stages, the more significant the current rehabilitation effect. Furthermore, to enhance the objectivity of the assessment, the benchmark for comparing the improvement of the current patient at each rehabilitation stage can refer to the improvement of historical patients at the same rehabilitation stage. Therefore, this embodiment obtains the current training effectiveness level of the current patient based on the difference between the current patient's current multimodal rehabilitation data and the multimodal rehabilitation data of other rehabilitation stages prior to the current stage, combined with the difference in multimodal rehabilitation data of historical patients at the current stage and other previous rehabilitation stages, accurately reflecting the current patient's current respiratory rehabilitation training effect; the greater the current training effectiveness level, the better the current rehabilitation training effect of the current patient.

[0029] Preferably, in one possible implementation of this embodiment, the method for obtaining the current training effectiveness level is described in [reference needed]. Figure 2 The flowchart illustrates a method for obtaining the current training effectiveness provided in this embodiment. The method includes the following steps: Step S201: Based on the differences between the current multimodal rehabilitation data of the current patient and the multimodal rehabilitation data of other rehabilitation stages before the current rehabilitation stage, obtain the current rehabilitation feature difference vector and training effect analysis value of the current patient.

[0030] It is known that the effectiveness of respiratory rehabilitation training should gradually improve with each rehabilitation stage. Therefore, this embodiment obtains the current rehabilitation feature difference vector and training effect analysis value of the current patient based on the differences between the current patient's multimodal rehabilitation data and the multimodal rehabilitation data of other rehabilitation stages before the current stage. This prepares for subsequent analysis of the current patient's respiratory rehabilitation training effect. A higher training effect analysis value indicates a better current respiratory rehabilitation training effect for the patient.

[0031] In one possible implementation of this embodiment, the current rehabilitation feature difference vector of the current patient is obtained as follows: Reference values ​​for all types of rehabilitation data for the current patient each day are constructed into a vector, serving as the daily rehabilitation feature vector for the current patient; elements at the same position in the rehabilitation feature vector correspond to the same type of data; for any rehabilitation stage, the average of the rehabilitation feature vectors for all days within that rehabilitation stage is used as the reference vector for that rehabilitation stage, accurately representing the rehabilitation characteristics of the current patient at that stage, which is beneficial for more accurate and efficient analysis of the current patient's current rehabilitation effect; rehabilitation stages preceding the current rehabilitation stage in time sequence are all considered historical rehabilitation stages; furthermore, the average of the difference vectors between the current patient's current day's rehabilitation feature vector and the reference vectors of each historical rehabilitation stage is used as the current patient's current rehabilitation feature difference vector, accurately and comprehensively representing the current patient's current rehabilitation status. It should be noted that if the current rehabilitation stage is the first rehabilitation stage, i.e. there is no previous historical rehabilitation stage, then the corresponding elements belonging to the same data type as the rehabilitation features are extracted from the patient's initial multimodal diagnosis and treatment feature vector before rehabilitation training, and an initial reference vector of equal length is constructed as the historical rehabilitation stage reference vector for reference comparison.

[0032] In order to comprehensively represent the current training effect of the current patient, the original physical dimensions of each element in the current rehabilitation feature difference vector of the current patient are removed independently (that is, each element is divided by its corresponding unit value), and the average of the processed elements is used as the current training effect analysis value of the current patient.

[0033] Step S202: Based on the differences in multimodal rehabilitation data between the current rehabilitation stage of the reference historical patient and other previous rehabilitation stages, obtain the comparative rehabilitation feature difference vector of the current rehabilitation stage of the reference historical patient.

[0034] To more accurately analyze whether the current respiratory rehabilitation training effect of the current patient conforms to the normal trend, and then obtain the comparative rehabilitation feature difference vector of the current rehabilitation stage of the reference historical patient based on the differences in multimodal rehabilitation data between the current rehabilitation stage and other previous rehabilitation stages, this vector serves as the benchmark for the current rehabilitation feature difference vector analysis of the current patient.

[0035] In one possible implementation of this embodiment, the method for obtaining the comparative rehabilitation feature difference vector is as follows: the vector corresponding to the average of the reference vectors of all historical patients in the current rehabilitation stage is used as the overall historical reference vector of the current rehabilitation stage; then the vector corresponding to the average of the difference vectors of the overall historical reference vectors of each historical rehabilitation stage minus the overall historical reference vector of the current rehabilitation stage is used as the comparative rehabilitation feature difference vector of the current rehabilitation stage of the historical patients.

[0036] Step S203: The result of normalizing the current patient's current rehabilitation feature difference vector with the cosine similarity of the compared rehabilitation feature difference vector is used as a reference similarity weight.

[0037] The more similar the current patient's current rehabilitation feature difference vector is to the comparison rehabilitation feature difference vector of a historical patient at their current rehabilitation stage, the more reasonable the analysis of the current patient's respiratory rehabilitation training effect is. Therefore, this embodiment uses the normalized result of the cosine similarity between the current patient's current rehabilitation feature difference vector and the comparison rehabilitation feature difference vector as a reference similarity weight. The larger the reference similarity weight, the more reasonable the analysis of the current patient's respiratory rehabilitation training effect is. The original value range of the cosine similarity is known to be... In order to eliminate the influence of negative values ​​on subsequent product operations and to make the result have probabilistic interpretation, this embodiment uses the result of adding 1 to the above cosine similarity and dividing by 2 as the normalized result of the above cosine similarity, so that the value of the reference similarity weight is between 0 and 1.

[0038] Step S204: The result of normalizing the product of the reference similarity weight and the training effect analysis value is taken as the current training effectiveness of the current patient.

[0039] It is known that a larger reference similarity weight indicates a more reasonable analysis of the current patient's respiratory rehabilitation training effect; a larger training effect analysis value indicates a better current patient's respiratory rehabilitation training effect. Therefore, this embodiment normalizes the product of the reference similarity weight and the training effect analysis value as the current patient's current training effectiveness, effectively improving the accuracy and scientific nature of subsequent assessments of the current patient's current rehabilitation effect. This embodiment normalizes the product of the reference similarity weight and the training effect analysis value using the min-max normalization method. The dataset corresponding to the min-max normalization method is the product of the reference similarity weight and the training effect analysis value calculated for all historical patients at the current rehabilitation stage.

[0040] It should be noted that, in all the above maximum-minimum normalization methods, in order to avoid the denominator being 0, a preset minimum positive number is added to the denominator as a parameter adjustment factor for summation. In this embodiment, the preset minimum positive number is set to 0.1. Implementers can set the size of the preset minimum positive number according to the actual situation, and there is no limitation here.

[0041] Step S4: Assess the current respiratory rehabilitation training effect on the patient based on the current training effectiveness.

[0042] It is known that the greater the effectiveness of the current training, the better the current respiratory rehabilitation training effect for the patient. Therefore, this embodiment evaluates the current respiratory rehabilitation training effect for the patient based on the current effectiveness of the training.

[0043] This embodiment sets a preset training effectiveness threshold of 0.6. Implementers can set the value of the preset training effectiveness threshold according to actual conditions; no limitation is made here. When the current training effectiveness is greater than the preset threshold, the current respiratory rehabilitation training for the current patient is assessed as effective, the existing training framework is maintained, and monitoring continues. When the current training effectiveness is less than or equal to the preset threshold, the current respiratory rehabilitation training effect for the current patient is assessed as poor, and the current rehabilitation training program needs to be specifically strengthened to improve the patient's respiratory rehabilitation training effect.

[0044] In summary, this embodiment acquires multimodal diagnostic and treatment data and multimodal rehabilitation data; it obtains reference historical patients for the current patient based on the multimodal diagnostic and treatment data; it divides rehabilitation stages based on the multimodal rehabilitation data of the reference historical patients to determine the current rehabilitation stage of the current patient; and it assesses the current respiratory rehabilitation training effect of the current patient by obtaining the current training effectiveness level by combining the differences between the current patient's current rehabilitation stage and other rehabilitation stages before the current rehabilitation stage with the differences between the current rehabilitation stage and other rehabilitation stages of the reference historical patients. This invention, by obtaining the current training effectiveness level, facilitates the real-time and accurate determination of the current patient's respiratory rehabilitation training effect, thereby enabling timely adjustments to the rehabilitation training plan and effectively improving the patient's treatment outcome.

[0045] Example 2: This invention also proposes a respiratory rehabilitation training effect evaluation system based on multimodal data. Please refer to [link / reference]. Figure 3 The diagram illustrates a structure of a respiratory rehabilitation training effect evaluation system based on multimodal data according to an embodiment of the present invention. The system includes: a data acquisition module 10, a rehabilitation stage acquisition module 20, a current training effectiveness acquisition module 30, and an effect evaluation module 40.

[0046] The data acquisition module 10 is used to acquire multimodal diagnosis and treatment data of current patients and historical patients before respiratory rehabilitation training, as well as multimodal rehabilitation data during respiratory rehabilitation training. The rehabilitation stage acquisition module 20 is used to acquire reference historical patients for the current patient based on the similarity of multimodal diagnosis and treatment data of the current patient and historical patients; and to divide the rehabilitation stage based on the changes in the multimodal rehabilitation data of the reference historical patients to determine the current rehabilitation stage of the current patient. The current training effectiveness acquisition module 30 is used to obtain the current training effectiveness of the current patient based on the difference between the current multimodal rehabilitation data of the current patient and the multimodal rehabilitation data of other rehabilitation stages before the current rehabilitation stage, combined with the difference between the current rehabilitation stage of the current patient and the multimodal rehabilitation data of other rehabilitation stages before the current rehabilitation stage in the past. The effectiveness evaluation module 40 is used to evaluate the current respiratory rehabilitation training effect of the current patient based on the current training effectiveness.

[0047] It should be noted that the above embodiments of the respiratory rehabilitation training effect evaluation system based on multimodal data and the method embodiment of the respiratory rehabilitation training effect evaluation based on multimodal data belong to the same concept. The specific implementation process can be found in the method embodiment, and will not be repeated here.

[0048] Example 3: The present invention also proposes a computer device, see [link to relevant documentation]. Figure 4 The computer device includes a memory 401, a processor 402, and a computer program 403 stored in the memory 401 and running on the processor 402. When the processor 402 executes the computer program 403, the computer device can perform any of the aforementioned methods for evaluating the effectiveness of respiratory rehabilitation training based on multimodal data.

Claims

1. A method for evaluating the effectiveness of respiratory rehabilitation training based on multimodal data, characterized in that, The method includes the following steps: Acquire multimodal diagnostic and treatment data of current and historical patients before respiratory rehabilitation training, as well as multimodal rehabilitation data during respiratory rehabilitation training; Based on the similarity of multimodal diagnosis and treatment data of current patients and historical patients, reference historical patients are obtained for the current patient; based on the changes in multimodal rehabilitation data of reference historical patients, rehabilitation stages are divided to determine the current rehabilitation stage of the current patient; Based on the differences between the current multimodal rehabilitation data of the patient and the multimodal rehabilitation data of other rehabilitation stages before the current rehabilitation stage, and in conjunction with the differences between the current multimodal rehabilitation data of the patient and the multimodal rehabilitation data of other rehabilitation stages before the current rehabilitation stage, the current training effectiveness of the patient can be obtained. Assess the effectiveness of the current respiratory rehabilitation training for the current patient based on the current training effectiveness.

2. The method for evaluating the effect of respiratory rehabilitation training based on multimodal data as described in claim 1, characterized in that, The method for obtaining the reference historical patients is as follows: The multimodal diagnosis and treatment data of the current patient and each historical patient are transformed into vectors, which serve as the diagnosis and treatment feature vectors for the corresponding patients; elements at the same position in the diagnosis and treatment feature vectors correspond to the same type of diagnosis and treatment data. The cosine similarity between the current patient's and each historical patient's diagnostic and treatment feature vectors is obtained and used as a reference analysis value. The reference analysis values ​​are arranged in descending order to obtain the reference sequence; The historical patients corresponding to the first preset number of reference analysis values ​​in the reference sequence are all used as reference historical patients for the current patient.

3. The method for evaluating the effect of respiratory rehabilitation training based on multimodal data as described in claim 1, characterized in that, The method for obtaining the rehabilitation stage is as follows: For any reference historical patient, the degree of increase in the daily rehabilitation indicators of that reference historical patient is obtained based on the difference between the multimodal rehabilitation data of that reference historical patient and the previous adjacent day. When the increase in rehabilitation indicators exceeds the preset threshold for the increase in rehabilitation indicators, the corresponding day will be marked as a reference dividing day. Arrange the sequence numbers corresponding to the reference division days according to the time order to obtain the division sequence of the reference historical patients; The mean of the elements at the same position in the partition sequence of all reference historical patients is rounded down and used as the reference element at the corresponding position. The sequence formed by the reference elements is used as the target partitioning sequence; The average number of days corresponding to each element in the target segmentation sequence is taken as the target segmentation day. The segmentation is performed between the target segmentation day and its preceding adjacent day to obtain the recovery stage.

4. The method for evaluating the effect of respiratory rehabilitation training based on multimodal data as described in claim 3, characterized in that, The method for obtaining the degree of increase in the rehabilitation indicators is as follows: For any day and any rehabilitation data during the respiratory rehabilitation training process of any reference historical patient, the mean of the rehabilitation data for that day of the reference historical patient shall be used as the reference value of the rehabilitation data for that day of the reference historical patient. The difference between the reference value of the rehabilitation data of the reference patient on that day and the reference value of the adjacent day before that day is used as the incremental analysis value of the rehabilitation data of the reference patient on that day. The increase values ​​of each rehabilitation data of the reference historical patient on that day were dimensionless. The sum of the dimensionless increase values ​​and the result of normalization were taken as the degree of increase of the rehabilitation indicators of the reference historical patient on that day.

5. The method for evaluating the effect of respiratory rehabilitation training based on multimodal data as described in claim 4, characterized in that, The method for obtaining the current training effectiveness is as follows: Based on the differences between the current multimodal rehabilitation data of the patient and the multimodal rehabilitation data of other rehabilitation stages before the current rehabilitation stage, obtain the current rehabilitation feature difference vector and training effect analysis value of the patient. Based on the differences in multimodal rehabilitation data between the current rehabilitation stage of the reference historical patient and other previous rehabilitation stages, obtain the comparative rehabilitation feature difference vector of the current rehabilitation stage of the reference historical patient; The result of normalizing the cosine similarity between the current patient's current rehabilitation feature difference vector and the compared rehabilitation feature difference vector is used as the reference similarity weight. The normalized result of the product of the reference similarity weight and the training effect analysis value is taken as the current training effectiveness of the current patient.

6. The method for evaluating the effect of respiratory rehabilitation training based on multimodal data as described in claim 5, characterized in that, The method for obtaining the current patient's current rehabilitation feature difference vector is as follows: The reference values ​​of all types of rehabilitation data for the current patient each day are constructed into a vector, which serves as the rehabilitation feature vector for the current patient each day; where elements at the same position in the rehabilitation feature vector correspond to the same type of data. For any rehabilitation stage, the average of the rehabilitation feature vectors of all days within that rehabilitation stage is used as the reference vector for that rehabilitation stage. All rehabilitation stages that precede the current rehabilitation stage in time sequence are considered as historical rehabilitation stages; The average of the difference vectors between the current patient's current day's rehabilitation feature vector and the reference vectors of each historical rehabilitation stage is taken as the current patient's current rehabilitation feature difference vector.

7. The method for evaluating the effect of respiratory rehabilitation training based on multimodal data as described in claim 5, characterized in that, The method for obtaining the training effect analysis value is as follows: After independently removing the original physical dimensions from each element in the current patient's current rehabilitation characteristic difference vector, the average of the processed elements is used as the current patient's current training effect analysis value.

8. The method for evaluating the effect of respiratory rehabilitation training based on multimodal data as described in claim 6, characterized in that, The method for obtaining the comparison and rehabilitation feature difference vector is as follows: The vector corresponding to the average of the reference vectors of all patients in the current rehabilitation stage is used as the overall historical reference vector for the current rehabilitation stage. The vector obtained by averaging the difference vectors of the overall historical reference vectors of each historical rehabilitation stage and the overall historical reference vector of the current rehabilitation stage is used as the comparison rehabilitation feature difference vector of the current rehabilitation stage of the reference historical patients.

9. The method for evaluating the effect of respiratory rehabilitation training based on multimodal data as described in claim 1, characterized in that, The method for evaluating the current respiratory rehabilitation training effect of the current patient based on the current training effectiveness is as follows: When the current training effectiveness exceeds the preset training effectiveness threshold, assess the effectiveness of the current respiratory rehabilitation training for the current patient. When the current training effectiveness is less than or equal to the preset training effectiveness threshold, the current respiratory rehabilitation training effect of the patient is considered poor.

10. The method for evaluating the effect of respiratory rehabilitation training based on multimodal data as described in claim 1, characterized in that, The types of data in the multimodal diagnosis and treatment data include, but are not limited to, the types of data in the multimodal rehabilitation data.

Images