A stroke assessment method and device based on artificial intelligence

Abstract

The present application provides a stroke assessment method and device based on artificial intelligence, which belongs to the field of clinical medicine. The method includes: determining a stroke assessment indicator set, identifying the evaluation level of each indicator in the indicator set based on a pre-trained model and historical evaluation data; determining multiple evaluation dimensions of each indicator based on the historical evaluation data; determining the combination of evaluation dimensions under each evaluation level of each indicator based on the correlation between the evaluation dimension and the corresponding indicator evaluation level; determining the target evaluation indicator from the stroke assessment indicator set based on the information of the person to be assessed, and obtaining the stroke assessment result of the person to be assessed based on the target evaluation indicator and its corresponding evaluation level and evaluation dimension. The stroke assessment method and device based on artificial intelligence provided in the present application can help clinicians complete the examination quickly and accurately.

Description

Technical Field

[0001] The present application relates to the field of clinical medical technology, and in particular to a stroke assessment method and device based on artificial intelligence. Background Art

[0002] Stroke, also known as stroke, is an acute cerebrovascular disease. It is a group of diseases that cause brain tissue damage due to the rupture or blockage of brain blood vessels, which prevents blood from flowing into the brain. A stroke caused by a sudden rupture of a brain blood vessel is called a hemorrhagic stroke, and a stroke caused by a blockage of blood vessels that prevents blood from flowing into the brain is called an ischemic stroke. With the improvement of people's living standards, changes in living habits and dietary structure, as well as the aging of the population and traffic accidents, the incidence of cardiovascular disease has become increasingly serious. Once a stroke occurs, severe cases will die, and mild cases will become disabled. Therefore, it should be prevented, diagnosed and treated as early as possible.

[0003] For ischemic stroke, current guidelines recommend evaluating treatment decisions through the National Institute of Health Stroke Scale (NIHSS) score, clinical symptoms, and auxiliary examination results. Through the score, we can evaluate the severity of the patient's stroke, help identify stroke symptoms early, and provide a basis for the selection of treatment methods. Especially in the early stages of ischemic stroke, head CT cannot detect early ischemic sites. At this time, diagnosis and treatment depend on clinical scores. Early thrombolytic therapy based on NIHSS scores can effectively and quickly restore blood supply to brain tissue, reduce the occurrence of complications such as long-term disability, and improve patient prognosis. The National Institutes of Health Stroke Scale is a very good assessment tool for both doctors and patients. It can help clinicians quickly diagnose and evaluate stroke. Therefore, the NIHSS score is recommended for all patients suspected of ischemic stroke.

[0004] However, the NIHSS score is highly dependent on the doctor's experience and is completed manually by clinicians. The results are greatly affected by subjective factors and have poor homogeneity. In addition, due to the large number of test items, it is easy to miss items or extend the examination time, thus delaying the patient's diagnosis and subsequent treatment. Summary of the invention

[0005] In view of this, the present application provides an artificial intelligence-based stroke assessment method and device that can help clinicians complete examinations quickly and accurately.

[0006] Specifically, the present application is implemented through the following technical solutions:

[0007] The first aspect of the present application provides a stroke assessment method based on artificial intelligence, the method comprising:

[0008] Determine a stroke assessment indicator set, and identify the evaluation level of each indicator in the indicator set based on a pre-trained model and historical assessment data, wherein the evaluation level of each indicator is a numerical level;

[0009] Determine multiple evaluation dimensions for each indicator based on the historical evaluation data, wherein a first evaluation dimension for each indicator is determined based on the historical evaluation data, multiple second evaluation dimensions are selected based on the causal relationship and degree of causal association between each evaluation dimension and the first evaluation dimension, and the first evaluation dimension and the multiple second evaluation dimensions are combined as multiple evaluation dimensions corresponding to the indicator;

[0010] Based on the correlation between the evaluation dimensions and the corresponding indicator evaluation levels, determine the combination of evaluation dimensions at each evaluation level of each indicator, wherein the combination of evaluation dimensions at different evaluation levels of the same indicator is different, and the evaluation dimensions of different indicators are different;

[0011] Based on the information of the person to be evaluated, a target evaluation indicator is determined from the stroke evaluation indicator set, and based on the target evaluation indicator and its corresponding evaluation level and evaluation dimension, a stroke evaluation result of the person to be evaluated is obtained.

[0012] The second aspect of the present application provides a stroke assessment device based on artificial intelligence, the device comprising a determination module, a combination module and an assessment module; wherein:

[0013] The determination module is used to determine a stroke assessment indicator set, and identify the evaluation level of each indicator in the indicator set based on a pre-trained model and historical assessment data, wherein the evaluation level of each indicator is a numerical level;

[0014] The determination module is further configured to determine multiple evaluation dimensions of each indicator based on the historical evaluation data, wherein a first evaluation dimension of each indicator is determined based on the historical evaluation data, multiple second evaluation dimensions are selected based on the causal relationship and degree of causal association between each evaluation dimension and the first evaluation dimension, and the first evaluation dimension and the multiple second evaluation dimensions are combined as multiple evaluation dimensions corresponding to the indicator;

[0015] The combination module is used to determine the combination mode of the evaluation dimensions at each evaluation level of each indicator based on the correlation between the evaluation dimensions and the corresponding indicator evaluation level, wherein the combination modes of the evaluation dimensions at different evaluation levels of the same indicator are different, and the evaluation dimensions of different indicators are different;

[0016] The evaluation module is used to determine the target evaluation index from the stroke evaluation index set based on the information of the person to be evaluated, and obtain the stroke evaluation result of the person to be evaluated based on the target evaluation index and its corresponding evaluation level and evaluation dimension.

[0017] The artificial intelligence-based stroke assessment method and device provided in the present application first determine a stroke assessment indicator set, identify the evaluation level of each indicator in the indicator set based on a pre-trained model and historical assessment data, and then determine multiple evaluation dimensions of each indicator based on the historical assessment data, wherein a first evaluation dimension of each indicator is determined based on the historical assessment data, multiple second evaluation dimensions are selected based on the causal relationship and the degree of causal association between each evaluation dimension and the first evaluation dimension, and the first evaluation dimension and multiple second evaluation dimensions are combined as multiple evaluation dimensions corresponding to the indicator, and then based on the correlation between the evaluation dimensions and the corresponding indicator evaluation level, the evaluation dimension combination method of each indicator at each evaluation level is determined, wherein the evaluation dimension combination method at different evaluation levels of the same indicator is different, and the evaluation dimensions of different indicators are different, and finally, based on the information of the person to be assessed, a target evaluation indicator is determined from the stroke assessment indicator set, and the stroke assessment result of the person to be assessed is obtained based on the target evaluation indicator and its corresponding evaluation level and evaluation dimension. In this way, compared with determining the index of evaluation according to experience and determining the score of the index according to the doctor's sense organ, the stroke evaluation method based on the pre-training model and historical evaluation data provided by the present invention can accurately identify the evaluation index and determine the evaluation grade through the deep learning algorithm, that is, first construct a massive index pool for different types of evaluation personnel to match. On the one hand, each index grade of the index pool is identified according to the processing of historical data by the model, that is, whether it is the evaluation dimension of the index or the scoring method of the index, it is the most matched evaluation combination analyzed according to the situation of the evaluation personnel and the historical evaluation data, the evaluation index is in line with the actual situation of the evaluation personnel, and the rating level is also matched with the situation, the whole evaluation method is automatically determined, and personalized evaluation is realized according to individual circumstances by combining the combination of multiple evaluation dimensions, and the evaluation dimension combination method can be adjusted to improve the accuracy and reliability of the evaluation result, and the evaluation efficiency is higher, and the specific situation of different patients can be better reflected, and more accurate guidance is provided for the diagnosis and treatment of stroke patients. On the other hand, from the beginning of the patient's evaluation, all processes are automated, and scientific and automatic testing is realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A flowchart of embodiment 1 of the stroke assessment method based on artificial intelligence provided in this application;

[0019] Figure 2 A schematic diagram of a stroke detection instrument shown as an exemplary embodiment of the present application;

[0020] Figure 3 This is a schematic diagram of the structure of Example 1 of the artificial intelligence-based stroke assessment device provided in this application. DETAILED DESCRIPTION

[0021] Here, exemplary embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application.

[0022] The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in this article refers to and includes any or all possible combinations of one or more associated listed items.

[0023] It should be understood that although the terms first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0024] The present application provides an artificial intelligence-based stroke assessment method and device that can help clinicians complete examinations quickly and accurately.

[0025] The artificial intelligence-based stroke assessment method and device provided in the present application first determine a stroke assessment indicator set, identify the evaluation level of each indicator in the indicator set based on a pre-trained model and historical assessment data, and then determine multiple evaluation dimensions of each indicator based on the historical assessment data, wherein a first evaluation dimension of each indicator is determined based on the historical assessment data, multiple second evaluation dimensions are selected based on the causal relationship and the degree of causal association between each evaluation dimension and the first evaluation dimension, and the first evaluation dimension and multiple second evaluation dimensions are combined as multiple evaluation dimensions corresponding to the indicator, and then based on the correlation between the evaluation dimensions and the corresponding indicator evaluation level, the evaluation dimension combination method of each indicator at each evaluation level is determined, wherein the evaluation dimension combination method at different evaluation levels of the same indicator is different, and the evaluation dimensions of different indicators are different, and finally, based on the information of the person to be assessed, a target evaluation indicator is determined from the stroke assessment indicator set, and the stroke assessment result of the person to be assessed is obtained based on the target evaluation indicator and its corresponding evaluation level and evaluation dimension. In this way, compared with determining the index of evaluation according to experience and determining the score of the index according to the doctor's sense organ, the stroke evaluation method based on the pre-training model and historical evaluation data provided by the present invention can accurately identify the evaluation index and determine the evaluation grade through the deep learning algorithm, that is, first construct a massive index pool for different types of evaluation personnel to match. On the one hand, each index grade of the index pool is identified according to the processing of historical data by the model, that is, whether it is the evaluation dimension of the index or the scoring method of the index, it is the most matched evaluation combination analyzed according to the situation of the evaluation personnel and the historical evaluation data, the evaluation index is in line with the actual situation of the evaluation personnel, and the rating level is also matched with the situation, the whole evaluation method is automatically determined, and personalized evaluation is realized according to individual circumstances by combining the combination of multiple evaluation dimensions, and the evaluation dimension combination method can be adjusted to improve the accuracy and reliability of the evaluation result, and the evaluation efficiency is higher, and the specific situation of different patients can be better reflected, and more accurate guidance is provided for the diagnosis and treatment of stroke patients. On the other hand, from the beginning of the patient's evaluation, all processes are automated, and scientific and automatic testing is realized.

[0026] Specific embodiments are given below to introduce the technical solution of the present application in detail.

[0027] Figure 1 This is a flowchart of the first embodiment of the stroke assessment method based on artificial intelligence provided in this application. Please refer to Figure 1 , the method provided in this embodiment may include:

[0028] S101. Determine a stroke assessment indicator set, and identify the evaluation level of each indicator in the indicator set based on a pre-trained model and historical assessment data, wherein the evaluation level of each indicator is a numerical level.

[0029] It should be noted that the stroke assessment index set includes multiple indicators. Specifically, the multiple indicators at least include consciousness and understanding indicators, nervous system function indicators, motor function indicators, and sensory and cognitive function indicators. Among them, each specific indicator contains multiple corresponding evaluation dimensions. The relationship between the specific indicators and the corresponding evaluation dimensions is as follows:

[0030] The awareness and understanding indicators include at least awareness level judgment, awareness level questioning and awareness level instruction evaluation dimensions;

[0031] The nervous system function index at least includes gaze, visual field and facial paralysis evaluation dimensions;

[0032] The motor function index at least includes upper and lower limb movement and limb ataxia evaluation dimensions;

[0033] The sensory and cognitive function indicators at least include sensory, language and communication disorders and neglect examination evaluation dimensions.

[0034] It should be noted that the historical evaluation data may include the patient's basic information (e.g., age, gender, medical history, etc.), stroke evaluation indicators and their numerical levels, clinical information, etc. The evaluation level of each indicator in the stroke evaluation indicator set can be determined based on the pre-trained model and historical evaluation data, where the evaluation level of each indicator is a numerical level. For example, taking the consciousness and understanding indicators as an example, their evaluation levels can be 0 to 3 from low to high.

[0035] As an optional embodiment, the pre-trained model is fine-tuned according to historical evaluation data of the same patient type, the input of the pre-trained model is the patient's disease attributes and evaluation indicators, and the output of the pre-trained model is the level of each evaluation indicator;

[0036] Input the type of person to be evaluated and the stroke evaluation index set into the fine-tuned pre-training model to obtain the numerical evaluation level of each index in the stroke evaluation index set;

[0037] The evaluation value range corresponding to each numerical evaluation level is divided based on the evaluation content of each evaluation indicator, wherein the evaluation value range includes the evaluation value range of all evaluation dimensions under each indicator.

[0038] Taking the nervous system function index as an example, the numerical evaluation level output by the model can be 2, 1, and 0, but how these numerical levels correspond to the values ​​detected by the sensor requires determining the evaluation value range at each level. The evaluation value range corresponding to level 2 includes at least the value range of each evaluation dimension under the three evaluation dimensions of gaze, visual field, and facial paralysis, and so on.

[0039] The method provided by the present invention converts non-quantifiable evaluation into quantitative numerical evaluation values, and also establishes a corresponding relationship between the numerical evaluation values ​​and the sensor detection values, thereby reducing the reliance on human experience in the evaluation process and improving the consistency and scientificity of the evaluation.

[0040] S102. Determine multiple evaluation dimensions for each indicator based on the historical evaluation data, wherein a first evaluation dimension for each indicator is determined based on the historical evaluation data, multiple second evaluation dimensions are selected based on the causal relationship and the degree of causal association between each evaluation dimension and the first evaluation dimension, and the first evaluation dimension and the multiple second evaluation dimensions are combined as multiple evaluation dimensions corresponding to the indicator.

[0041] It should be noted that multiple evaluation dimensions of each indicator can be determined based on the relevant information of the stroke assessment indicators and their numerical levels in the historical assessment data. Specifically, the first evaluation dimension of each indicator can be determined preferentially. For example, in one embodiment, the first evaluation dimension of the consciousness and understanding indicators is the consciousness level judgment evaluation dimension, the first evaluation dimension of the nervous system function indicators is the gaze evaluation dimension, the first evaluation dimension of the motor function indicators is the upper and lower limb movement evaluation dimension, and the first evaluation dimension of the sensory and cognitive function indicators is the sensory rating dimension.

[0042] After determining the first evaluation dimension of each indicator, the relevant causal relationship analysis method and the correlation degree analysis method can be used to select multiple second evaluation dimensions associated with the first evaluation dimension, and then the first evaluation dimension and the multiple second evaluation dimensions can be combined as multiple evaluation dimensions corresponding to the indicator. For example, in one embodiment, when the first evaluation dimension of the consciousness and understanding indicator is the consciousness level judgment evaluation dimension, the second evaluation dimension obtained by the causal relationship analysis method and the correlation degree analysis method is the consciousness level question evaluation dimension and the consciousness level instruction evaluation dimension, and finally the first evaluation dimension (consciousness level judgment evaluation dimension) and the two second evaluation dimensions (consciousness level question evaluation dimension and consciousness level instruction evaluation dimension) are combined to obtain the evaluation dimension of the consciousness and understanding indicator.

[0043] Optionally, based on the historical evaluation data of the same patient type, an evaluation dimension with a usage frequency higher than a preset threshold is determined as the first evaluation dimension; an evaluation object relationship graph is constructed based on the causal relationship between the evaluation objects and the evaluation contents corresponding to all evaluation dimensions; and the evaluation dimensions corresponding to the evaluation objects with a correlation degree greater than a preset threshold are screened based on the object relationship graph as the second evaluation dimension. The evaluation object relationship graph constructed based on the causal relationship between the evaluation objects and the evaluation contents corresponding to all evaluation dimensions includes: all evaluation objects and the corresponding evaluation contents in the historical evaluation data of the same patient type; taking the evaluation objects as nodes, determining the connection relationship between each node based on the causal relationship between each evaluation object; and calculating the semantic relevance of the evaluation contents between nodes with connection relationships as the value attribute of the connection edge.

[0044] The method provided by the present invention determines the associated evaluation dimensions according to the cause-effect relationship, and automatically screens out some key dimensions to evaluate personnel. In addition, based on the pre-constructed relationship diagram, the screening and search of objects are simplified, the screening standards are unified, and the scientific nature of the screening is improved. When the relationship diagram is constructed, only the nodes with connection relationships need to calculate the relevance of their contents to avoid building an overly complex network. The entire relationship diagram is only used to characterize the cause-effect relationship between each node, which simplifies the construction and maintenance of the relationship diagram. In later use, by locating a node, each node with a causal relationship can be quickly searched, and the cause-effect relationship between each node and the node is very clear.

[0045] S103. Based on the correlation between the evaluation dimensions and the corresponding indicator evaluation levels, determine the combination method of the evaluation dimensions at each evaluation level of each indicator, wherein the combination methods of the evaluation dimensions at different evaluation levels of the same indicator are different, and the evaluation dimensions of different indicators are different.

[0046] It should be noted that, according to the relevant information of stroke assessment indicators and their numerical levels in the historical assessment data, this information includes multiple evaluation dimensions and their evaluation levels corresponding to each indicator (such as consciousness and understanding indicators, nervous system function indicators, motor function indicators, sensory and cognitive function indicators). The correlation between each evaluation dimension and the evaluation level of the corresponding indicator can be analyzed by statistical methods or other relevant methods. According to the obtained correlation information, a suitable combination of evaluation dimensions can be selected for each indicator at different evaluation levels, that is, a set of evaluation dimensions that best reflects the characteristics of the level can be determined for each evaluation level.

[0047] Specifically, any evaluation level includes multiple evaluation dimension combinations, each evaluation dimension combination corresponds to a personnel type, and one personnel type corresponds to multiple evaluation dimension combinations. The determination of the evaluation dimension combination for each indicator at each evaluation level includes:

[0048] Obtain the pre-assessment medical information of the person to be assessed, and identify the type of the person to be assessed based on the medical information; determine the target indicators in the evaluation indicator set and the combination of evaluation dimensions under each evaluation level of each target indicator according to the type of the person to be assessed. It should be noted that information such as age, gender, medical history, and current symptoms can be obtained from the medical information of the person to be assessed, and then the person to be assessed can be classified based on the medical information. For example, in one embodiment, elderly patients or patients with a specific medical history can be classified as one type, and patients with language disorders can be classified as another type.

[0049] In the specific implementation, it is assumed that there are three types of people to be evaluated, namely, elderly patients with a history of hypertension, patients with symptoms of language disorders, and other patients. At this time, if the evaluation levels are 0 to 2 from low to high, for elderly patients with a history of hypertension, when the evaluation level is 0, the evaluation dimension combination method is the consciousness level judgment evaluation dimension, the gaze evaluation dimension, and the upper and lower limb movement evaluation dimension; when the evaluation level is 1, the visual field evaluation dimension is added on the basis of the original level 0; when the evaluation level is 2, the sensory evaluation dimension is added on the basis of the original level 1. For patients with symptoms of language disorders, when the evaluation level is 0, the evaluation dimension combination method is the consciousness level judgment evaluation dimension, the gaze evaluation dimension, the upper and lower limb movement evaluation dimension, and the visual field evaluation dimension; when the evaluation level is 1, the neglect inspection evaluation dimension is added on the basis of the original level 0; when the evaluation level is 2, the sensory evaluation dimension is added on the basis of the original level 1. For patients with other conditions, when the evaluation level is 0, the combination of evaluation dimensions is consciousness level judgment evaluation dimension, gaze evaluation dimension, upper and lower limb movement evaluation dimension, visual field evaluation dimension and sensation evaluation dimension; when the evaluation level is 1, the sensation evaluation dimension is added on the basis of the original level 0; when the evaluation level is 2, the language and communication disorder evaluation dimension and the neglect examination evaluation dimension are added on the basis of the original level 1.

[0050] S104, determining a target assessment indicator from the stroke assessment indicator set based on the information of the person to be assessed, and obtaining a stroke assessment result of the person to be assessed based on the target assessment indicator and its corresponding assessment grade and assessment dimension.

[0051] It should be noted that the information of the person to be evaluated includes age, gender, medical history and other information. The target evaluation index determined from the stroke evaluation index set based on the information of the person to be evaluated can be determined according to the set rules. For example, elderly patients may pay more attention to the level of consciousness and motor function, while patients with language disorders need to focus on the evaluation of language function. Specifically, the stroke evaluation results of the person to be evaluated based on the target evaluation index and its corresponding evaluation level and evaluation dimension include:

[0052] (1) Determine each evaluation dimension included in the target evaluation indicator based on the target evaluation indicator;

[0053] Specifically, the evaluation dimensions included in the target evaluation indicators can be found in the above introduction and will not be repeated here.

[0054] (2) obtaining a test result of each evaluation dimension according to each evaluation dimension, and performing a dimension score on the test result;

[0055] It should be noted that the scoring process for each evaluation dimension is as follows:

[0056] Select any dimension, and according to the instructions issued by the human-computer dialogue, the sensor collects the response of the person to be evaluated to the instructions; score the selected dimension according to the correlation between the numerical range detected by the sensor and the indicator evaluation level.

[0057] It should be noted that the type of sensor is selected according to actual needs and is not limited in this embodiment. For example, the sensor may be a camera, a test plate, a gripper, a pressure sensor, etc. Specifically, the collected reaction data is analyzed and converted into a numerical interval for comparison and scoring with the index evaluation level. According to the previously defined association relationship, the collected numerical interval is compared with the index evaluation level, and then a corresponding score is given. For example, if the reaction data is within the optimal range of the index evaluation level, a higher score is given; otherwise, a lower score is given.

[0058] (3) Obtain the total score of the evaluation dimension of the target evaluation indicator based on the dimension scoring results of each evaluation dimension, and identify the level corresponding to the total score of the evaluation dimension.

[0059] For example, in one embodiment, the NIHSS score ranges from 0 to 42, with higher scores indicating more severe neurological damage, wherein 0 to 1 indicates normal or near normal, 1 to 4 indicates mild stroke, 5 to 15 indicates moderate stroke, 15 to 20 indicates moderate to severe stroke, and 20 to 42 indicates severe stroke.

[0060] The artificial intelligence-based stroke assessment method provided in this embodiment first determines a stroke assessment indicator set, identifies the evaluation level of each indicator in the indicator set based on a pre-trained model and historical assessment data, and then determines multiple evaluation dimensions of each indicator based on the historical assessment data, wherein a first evaluation dimension of each indicator is determined based on the historical assessment data, multiple second evaluation dimensions are selected based on the causal relationship and the degree of causal association between each evaluation dimension and the first evaluation dimension, and the first evaluation dimension and multiple second evaluation dimensions are combined as multiple evaluation dimensions corresponding to the indicator, and then based on the correlation between the evaluation dimensions and the corresponding indicator evaluation level, the evaluation dimension combination method at each evaluation level of each indicator is determined, wherein the evaluation dimension combination methods at different evaluation levels of the same indicator are different, and the evaluation dimensions of different indicators are different, and finally, based on the information of the person to be assessed, a target evaluation indicator is determined from the stroke assessment indicator set, and the stroke assessment result of the person to be assessed is obtained based on the target evaluation indicator and its corresponding evaluation level and evaluation dimension. In this way, compared with determining the index of evaluation according to experience and determining the score of the index according to the doctor's sense organ, the stroke evaluation method based on the pre-training model and historical evaluation data provided by the present invention can accurately identify the evaluation index and determine the evaluation grade through the deep learning algorithm, that is, first construct a massive index pool for different types of evaluation personnel to match. On the one hand, each index grade of the index pool is identified according to the processing of historical data by the model, that is, whether it is the evaluation dimension of the index or the scoring method of the index, it is the most matched evaluation combination analyzed according to the situation of the evaluation personnel and the historical evaluation data, the evaluation index is in line with the actual situation of the evaluation personnel, and the rating level is also matched with the situation, the whole evaluation method is automatically determined, and personalized evaluation is realized according to individual circumstances by combining the combination of multiple evaluation dimensions, and the evaluation dimension combination method can be adjusted to improve the accuracy and reliability of the evaluation result, and the evaluation efficiency is higher, and the specific situation of different patients can be better reflected, and more accurate guidance is provided for the diagnosis and treatment of stroke patients. On the other hand, from the beginning of the patient's evaluation, all processes are automated, and scientific and automatic testing is realized.

[0061] The specific implementation of the present application is described below. The present method is applied to a stroke detection instrument. Specifically, Figure 2 This is a schematic diagram of a stroke detection instrument shown in an exemplary embodiment of the present application. Figure 2 ,The instrument includes a display screen, a camera, an upper limb test board and a gripper, a lower limb test board and a pressure sensor.

[0062] Combine the following Figure 2 The instrument shown here explains the testing process of the consciousness and understanding index, the nervous system function index, the motor function index, the sensory and cognitive function index and their corresponding evaluation dimensions:

[0063] Awareness and understanding indicators:

[0064] Consciousness level judgment evaluation dimension: Through human-computer dialogue: ask the patient's gender, age, time. When aphasia or obstacles such as placement of endotracheal tube, language barriers or oropharyngeal tracheal trauma / use of bandages prevent complete assessment, pressurize the patient's nail bed (device 4) through the pressure sensor, and observe the patient's facial reaction to stimulation through identification.

[0065] Consciousness level question evaluation dimension: Ask the patient's gender, age, and time through human-computer dialogue.

[0066] Evaluation dimension of consciousness level command: issuing commands through human-computer dialogue, closing eyes, sticking out tongue, and clenching the handgrip (device 3), and judging whether the action is completed correctly by identifying the changes in the observer's facial expression (device 1) and the upper limb movement handgrip (device 3).

[0067] Nervous system function indicators:

[0068] Gaze evaluation dimension: Instruct the patient to follow the video screen (device 2) by issuing instructions through human-computer dialogue, and determine the patient's eye movement by identifying changes in the observer's facial expression (device 1).

[0069] Visual field evaluation dimension: Cursors of different intensities move from the periphery to the center of the display (device 2). During the movement, the patient goes from not seeing the cursor to seeing it. When seeing it, the patient immediately presses the confirmation key on the screen. Connecting all the reaction points on the boundary between not seeing and seeing forms the range of the visual field.

[0070] Facial paralysis evaluation dimension: Ask the patient to smile and use the face image recognition (device 1) function to determine whether the patient has facial paralysis.

[0071] Motor function indicators:

[0072] Upper and lower limb movement evaluation dimension: Among them, upper limb movement is to instruct the patient to lift and hold the handgripper (device 3) with the left upper limb and the right upper limb in turn through human-computer dialogue, measure the handgripper lifting angle and grip strength, and calculate the patient's grip strength index (grip strength (Kg) ÷ body weight (Kg) × 100). The specific evaluation criteria are:

[0073] 1 point if the grip strength index is less than 50% and the lifting angle reaches 90° (45° in supine position) but is less than 10s;

[0074] The grip strength index is less than 50% but the lifting angle is less than 90° (45° in supine position) 2 points;

[0075] The grip strength index is less than 50% and cannot maintain 3 points;

[0076] 4 points if the grip strength index is less than 50% and the hand cannot be lifted at all;

[0077] Among them, the lower limb movement is to instruct the patient to lift and hold the lower limb test board (device 5) with the left lower limb and the right lower limb in turn through human-computer dialogue, and measure the lifting angle. The specific evaluation criteria are:

[0078] 1 minute if the elevation angle is greater than 30° but the holding time is less than 5 seconds;

[0079] 2 points if the elevation angle is less than 30° but the holding time is less than 5 seconds;

[0080] The elevation angle is less than 30° but cannot be maintained at all 3 points;

[0081] Unable to lift 4 points;

[0082] Limb ataxia evaluation dimension: Through human-computer dialogue, the patient is instructed to perform the finger-to-nose test with both upper limbs in turn. The test is calibrated and judged by identifying the changes in the observer's facial expression. If the test can be completed 8 times or more within 10 seconds and the finger-to-nose test is accurate, it is normal.

[0083] Sensory and cognitive function indicators:

[0084] Sensory evaluation dimension: Through human-computer dialogue, the patient was instructed to put both hands into the pressure sensor (device 4) in turn, and then low-frequency current stimulation (5Hz, wave width gradually increased from 0.1ms to 0.5ms) was given. It was normal for the patient to feel pain and muscle contraction when the wave width was ≤0.3ms. 1 point was given for pain and muscle contraction within 0.3-0.4ms, and 2 points were given for pain and muscle contraction exceeding 0.4ms. Both upper limbs were tested in turn.

[0085] Language and articulation disorder assessment dimension: Instruct the patient to read the text (20-30 words) that appears on the screen, and ask the patient to name the items displayed on the screen one by one. Test the patient's speech speed and voice, and compare the patient's speech clarity with standard voice. The ability to understand language can be judged by the patient's response to this test and all the instructions given in the above neurological examination.

[0086] Neglected inspection and evaluation dimensions: 1. Bisection test: Several lines of different lengths appear on the screen (as shown in the figure), and the patient is told to click on the midpoint of each line segment. If the distance between the point selected by the patient and the actual midpoint exceeds 10% of the total length of the line segment, it is abnormal. 2. Image recognition test: Different pictures are displayed on the left and right sides of the screen at the same time, and the patient is told to say the content of all the pictures. If the patient can only say the content of one side of the picture, it is abnormal.

[0087] Preferably, after obtaining the stroke assessment result of the person to be assessed based on the target assessment index and its corresponding assessment level and assessment dimension, the method further includes: comparing the assessment result with the medical treatment result of the person to be assessed; using the comparison result as a prompt to correct the pre-trained model and the combination mode of the assessment dimension; and evaluating the same type of persons to be assessed based on the corrected model. In order to further improve the scientific nature of the assessment indicators, dimensions and combination modes, the method provided by the present invention, after the assessment, further corrects the output of the model and the combination mode of the dimensions based on the actual assessment results and the diagnosis results as prompt information, and optimizes the process of automatically determining the assessment plan.

[0088] Corresponding to the aforementioned embodiment of an artificial intelligence-based stroke assessment method, the present application also provides an embodiment of an artificial intelligence-based stroke assessment device.

[0089] Figure 3 This is a schematic diagram of the structure of the first embodiment of the stroke assessment device based on artificial intelligence provided by this application. Figure 3 The device provided in this embodiment includes a determination module 310, a combination module 320 and an evaluation module 330; wherein,

[0090] The determination module 310 is used to determine a stroke assessment indicator set, and identify the evaluation level of each indicator in the indicator set based on a pre-trained model and historical assessment data, wherein the evaluation level of each indicator is a numerical level;

[0091] The determination module 310 is further configured to determine multiple evaluation dimensions of each indicator based on the historical evaluation data, wherein a first evaluation dimension of each indicator is determined based on the historical evaluation data, multiple second evaluation dimensions are selected based on the causal relationship and degree of causal association between each evaluation dimension and the first evaluation dimension, and the first evaluation dimension and the multiple second evaluation dimensions are combined as multiple evaluation dimensions corresponding to the indicator;

[0092] The combination module 320 is used to determine the combination mode of the evaluation dimensions at each evaluation level of each indicator based on the correlation between the evaluation dimensions and the corresponding indicator evaluation level, wherein the combination modes of the evaluation dimensions at different evaluation levels of the same indicator are different, and the evaluation dimensions of different indicators are different;

[0093] The evaluation module 330 is used to determine a target evaluation indicator from the stroke evaluation indicator set based on the information of the person to be evaluated, and obtain a stroke evaluation result of the person to be evaluated based on the target evaluation indicator and its corresponding evaluation level and evaluation dimension.

[0094] The device of this embodiment can be used to perform Figure 1The steps, specific implementation principles and implementation processes of the method embodiment shown are similar and will not be repeated here.

[0095] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0096] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present application scheme. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0097] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A stroke assessment method based on artificial intelligence, characterized in that: The method comprises: Determine a stroke assessment indicator set, and identify the evaluation level of each indicator in the indicator set based on a pre-trained model and historical assessment data, wherein the evaluation level of each indicator is a numerical level; Determine multiple evaluation dimensions for each indicator based on the historical evaluation data, wherein a first evaluation dimension for each indicator is determined based on the historical evaluation data, multiple second evaluation dimensions are selected based on the causal relationship and degree of causal association between each evaluation dimension and the first evaluation dimension, and the first evaluation dimension and the multiple second evaluation dimensions are combined as multiple evaluation dimensions corresponding to the indicator; Based on the correlation between the evaluation dimensions and the corresponding indicator evaluation levels, determine the combination of evaluation dimensions at each evaluation level of each indicator, wherein the combination of evaluation dimensions at different evaluation levels of the same indicator is different, and the evaluation dimensions of different indicators are different; Determining a target evaluation indicator from the stroke evaluation indicator set based on the information of the person to be evaluated, and obtaining a stroke evaluation result of the person to be evaluated based on the target evaluation indicator and its corresponding evaluation level and evaluation dimension; The indicators at least include consciousness and understanding indicators, nervous system function indicators, motor function indicators, and sensory and cognitive function indicators; wherein the consciousness and understanding indicators at least include consciousness level judgment, consciousness level questioning, and consciousness level instruction evaluation dimensions; the nervous system function indicators at least include gaze, visual field, and facial paralysis evaluation dimensions; the motor function indicators at least include upper and lower limb movement and limb ataxia evaluation dimensions; the sensory and cognitive function indicators at least include sensory, language, communication disorders, and neglect examination evaluation dimensions; The method comprises determining multiple evaluation dimensions for each indicator based on the historical evaluation data, wherein an evaluation dimension with a usage frequency higher than a preset threshold is determined according to the historical evaluation data of the same patient type as the first evaluation dimension; constructing an evaluation object relationship diagram based on the causal relationship between the evaluation objects and the evaluation contents corresponding to all evaluation dimensions; and screening the evaluation dimensions corresponding to the evaluation objects with a correlation degree greater than a preset threshold as the second evaluation dimension based on the object relationship diagram.

2. The method according to claim 1, characterized in that Any of the evaluation levels includes multiple evaluation dimension combinations, each evaluation dimension combination corresponds to a personnel type, and a personnel type corresponds to multiple evaluation dimension combinations. The evaluation dimension combination method for determining each indicator at each evaluation level includes: Obtaining pre-assessment medical information of the person to be assessed, and identifying the type of the person to be assessed based on the medical information; The target indicator in the evaluation indicator set and the combination method of the evaluation dimensions at each evaluation level of each target indicator are determined according to the type of the person to be evaluated.

3. The method according to claim 1, characterized in that The obtaining of the stroke assessment result of the person to be assessed based on the target assessment indicator and its corresponding assessment level and assessment dimension includes: Determine each evaluation dimension included in the target evaluation indicator according to the target evaluation indicator; Obtaining a test result of each evaluation dimension according to each evaluation dimension, and performing dimension scoring on the test result; The total score of the evaluation dimension of the target evaluation indicator is obtained according to the dimension scoring result of each evaluation dimension, and the level corresponding to the total score of the evaluation dimension is identified.

4. The method according to claim 3, characterized in that The dimension scoring process according to each evaluation dimension is as follows: Select any dimension, and based on the instructions issued by the human-computer dialogue, the sensor collects the response of the person to be evaluated to the instructions; The selected dimension is scored according to the correlation between the numerical range detected by the sensor and the indicator evaluation level.

5. The method according to claim 1, characterized in that The identifying the evaluation level of each indicator in the indicator set based on the pre-training model and the historical evaluation data includes: Fine-tuning the pre-trained model according to historical evaluation data of the same patient type, wherein the input of the pre-trained model is the patient's disease attributes and evaluation indicators, and the output of the pre-trained model is the level of each evaluation indicator; Input the type of person to be evaluated and the stroke evaluation index set into the fine-tuned pre-training model to obtain the numerical evaluation level of each index in the stroke evaluation index set; The evaluation value range corresponding to each numerical evaluation level is divided based on the evaluation content of each evaluation indicator, wherein the evaluation value range includes the evaluation value range of all evaluation dimensions under each indicator.

6. The method according to claim 1, characterized in that The evaluation object relationship diagram is constructed based on the causal relationship between the evaluation objects and evaluation contents corresponding to all evaluation dimensions, including: All evaluation objects and corresponding evaluation contents in the historical evaluation data of the same patient type; Taking the evaluation objects as nodes, determining the connection relationship between the nodes according to the causal relationship between the evaluation objects; The semantic relevance of the evaluation content between nodes with connection relationships is calculated as the value attribute of the connection edge.

7. The method according to claim 1, characterized in that After obtaining the stroke assessment result of the person to be assessed based on the target assessment index and its corresponding assessment level and assessment dimension, the method further includes: Comparing the assessment results with the medical treatment results of the person to be assessed; Using the comparison results as a hint, correct the combination of pre-trained models and evaluation dimensions; Evaluate the same type of persons to be evaluated based on the revised model.

8. A stroke assessment device based on artificial intelligence, characterized in that: The device comprises a determination module, a combination module and an evaluation module; wherein, The determination module is used to determine a stroke assessment indicator set, and identify the evaluation level of each indicator in the indicator set based on a pre-trained model and historical assessment data, wherein the evaluation level of each indicator is a numerical level; the indicators at least include consciousness and understanding indicators, nervous system function indicators, motor function indicators, and sensory and cognitive function indicators; wherein the consciousness and understanding indicators at least include consciousness level judgment, consciousness level questioning, and consciousness level instruction evaluation dimensions; the nervous system function indicators at least include gaze, visual field, and facial paralysis evaluation dimensions; the motor function indicators at least include upper and lower limb movement and limb ataxia evaluation dimensions; the sensory and cognitive function indicators at least include sensory, language and communication disorders, and neglect inspection evaluation dimensions; The determination module is further used to determine multiple evaluation dimensions of each indicator based on the historical evaluation data, wherein a first evaluation dimension of each indicator is determined based on the historical evaluation data, multiple second evaluation dimensions are selected based on the causal relationship and the degree of causal association between each evaluation dimension and the first evaluation dimension, and the first evaluation dimension and the multiple second evaluation dimensions are combined as multiple evaluation dimensions corresponding to the indicator; the multiple evaluation dimensions of each indicator are determined based on the historical evaluation data, wherein an evaluation dimension with a usage frequency higher than a preset threshold is determined according to the historical evaluation data of the same patient type as the first evaluation dimension; an evaluation object relationship diagram is constructed based on the causal relationship between the evaluation objects and the evaluation contents corresponding to all evaluation dimensions; and the evaluation dimension corresponding to the evaluation object with a degree of association greater than a preset threshold is screened based on the object relationship diagram as the second evaluation dimension; The combination module is used to determine the combination mode of the evaluation dimensions at each evaluation level of each indicator based on the correlation between the evaluation dimensions and the corresponding indicator evaluation level, wherein the combination modes of the evaluation dimensions at different evaluation levels of the same indicator are different, and the evaluation dimensions of different indicators are different; The evaluation module is used to determine the target evaluation index from the stroke evaluation index set based on the information of the person to be evaluated, and obtain the stroke evaluation result of the person to be evaluated based on the target evaluation index and its corresponding evaluation level and evaluation dimension.

Images