Aging clock assessment method for hemodialysis population based on artificial intelligence

CN118588299BActive Publication Date: 2024-11-05SOUTHEAST UNIV
View PDF 9 Cites 0 Cited by

Patent Information

Application Number
CN202411074343.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-07
Publication Date
2024-11-05
Estimated Expiration
2044-08-07

Smart Images

  • Figure CN118588299B_ABST
    Figure CN118588299B_ABST
Patent Text Reader

Abstract

The present invention belongs to the technical field of aging clock assessment, and specifically relates to an aging clock assessment method for hemodialysis population based on artificial intelligence. The invention collects medical image data of hemodialysis patients, obtains first feature information by using data preprocessing and feature extraction technology, further classifies and processes the first feature information according to the pre-classification conditions, obtains multiple second feature information, and simultaneously obtains the patient's hematological indicators, and combines and calculates with standard indicators to obtain third feature information, and then classifies the patients into multiple assessment groups according to physical sign parameters and medical history information, and finally trains the second feature information and the third feature information under each assessment group in combination with a convolutional neural network model, and obtains an aging clock assessment model under different assessment groups. By applying the aging clock assessment model to actual scenarios, it can accurately assess the aging degree of the hemodialysis population and provide doctors with individualized treatment recommendations.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of aging clock assessment, and in particular relates to an aging clock assessment method for hemodialysis population based on artificial intelligence. Background Art

[0002] With the continuous advancement of medical technology, hemodialysis, as an important renal replacement therapy, has been widely used in the treatment of patients with chronic kidney disease. However, since hemodialysis patients are often accompanied by multiple complications and decreased physiological functions, their aging process is more complicated and rapid than that of healthy people. Therefore, how to accurately assess the degree of aging of the hemodialysis population to guide clinical treatment and patient management has become an urgent problem to be solved.

[0003] Traditional aging assessment methods for hemodialysis populations mainly rely on the clinical experience of doctors and the biochemical indicators of patients. However, this method often has problems such as strong subjectivity and poor accuracy, which will lead to deviations in the aging clock assessment results of hemodialysis patients, which will undoubtedly affect the treatment plans formulated by doctors. Correspondingly, it may also cause irreversible damage to the treatment of patients. Based on this, the present invention proposes an aging clock assessment method for hemodialysis populations based on artificial intelligence to solve the above problems. Summary of the invention

[0004] The purpose of the present invention is to provide an artificial intelligence-based aging clock assessment method for hemodialysis population, which can combine medical image data, hematological indicators, physical sign parameters and medical history information to comprehensively and objectively and accurately assess the aging degree of hemodialysis patients.

[0005] The technical solution adopted by the present invention is as follows:

[0006] The artificial intelligence-based aging clock assessment method for hemodialysis population includes:

[0007] Collecting medical image data of hemodialysis patients, and performing preprocessing and feature extraction on the data to obtain first feature information;

[0008] Acquire a pre-classification condition, and classify the first characteristic information according to the pre-classification condition to obtain a plurality of second characteristic information;

[0009] Obtaining hematological indicators of hemodialysis patients, combining them with standard indicators, and marking the calculation results as third characteristic information;

[0010] Acquiring physical sign parameters and medical history information, and classifying hemodialysis patients into multiple evaluation groups based on the physical sign parameters and medical history information;

[0011] The second characteristic information and the third characteristic information of each evaluation group are aggregated into a training set and a test set, and are trained in combination with a convolutional neural network model to obtain an aging clock evaluation model;

[0012] A target hemodialysis patient is obtained and matched with the evaluation group to which the patient belongs, and then the corresponding aging clock evaluation model is called to obtain the aging clock of the target hemodialysis patient.

[0013] In a preferred embodiment, the step of collecting medical image data of hemodialysis patients, preprocessing the data and extracting features to obtain first feature information includes:

[0014] Acquire CT or MRI medical image data of hemodialysis patients, and then preprocess the image data, wherein the preprocessing includes denoising, enhancement, and segmentation to eliminate artifacts in the image and obtain a reference image;

[0015] Obtaining the required features corresponding to the aging clock, and performing segmentation processing in the reference image to obtain a plurality of feature-attached regions;

[0016] Cross-comparison is performed on the similar characteristic attached regions corresponding to each hemodialysis patient, and the comparison result is output as the first parameter to be screened;

[0017] Obtaining a screening threshold, and comparing the screening threshold with a first parameter to be screened;

[0018] If the first parameter to be screened is greater than the screening threshold, the demand feature in the corresponding feature-attached area is marked as the first feature information;

[0019] If the first parameter to be screened is less than or equal to the screening threshold, the demand feature in the corresponding feature-attached area is marked as invalid feature information.

[0020] In a preferred embodiment, the step of cross-comparing the similar characteristic attached regions corresponding to the respective hemodialysis patients and outputting the comparison results as the first parameter to be screened includes:

[0021] Obtain similar feature-attached regions corresponding to hemodialysis patients;

[0022] Aligning similar feature-attached regions corresponding to each hemodialysis patient one by one to obtain aligned feature-attached regions;

[0023] Obtaining a first measurement function;

[0024] A pixel-level comparison is performed on each aligned feature-attached region, the difference between the pixel values ​​is calculated according to the first measurement function, the difference is quantified, and the quantification result is calibrated as the first parameter to be screened.

[0025] In a preferred embodiment, the step of obtaining the pre-classification condition and classifying the first feature information according to the pre-classification condition to obtain a plurality of second feature information includes:

[0026] Acquiring pre-classification conditions, wherein the pre-classification conditions include body surface information and blood vessel information of a hemodialysis patient;

[0027] Classify the first characteristic information according to the pre-classification condition to divide the hemodialysis patients into different subgroups;

[0028] A demand indicator is obtained, and corresponding second feature information is extracted from a corresponding subgroup according to the demand indicator.

[0029] In a preferred embodiment, after the second characteristic information is output, the body surface information therein is screened, and the specific process includes:

[0030] Acquire a skin region image of a hemodialysis patient, and calculate statistical information of the number of pixels and pixel values ​​of the skin region image;

[0031] Based on statistical information, the skin area is determined according to the pixel value distribution, and the edge detection is performed on the skin area to determine the skin boundary;

[0032] Calculating the skin area of ​​the skin region according to the inflection point of the skin boundary and marking it as the second parameter to be screened;

[0033] Obtaining a first allowable deviation interval, and performing an offset process on each second parameter to be screened according to the first allowable deviation interval to obtain a plurality of first offset intervals;

[0034] Compare the first offset intervals one by one, and classify the first offset intervals with overlapping values ​​into intervals of the same category;

[0035] The proportion of each same category interval in the first offset interval is obtained, and the second to-be-screened parameter corresponding to the first offset interval with the largest proportion is retained as the second characteristic information, and the second to-be-screened parameters corresponding to other first offset intervals are screened out from the second characteristic information.

[0036] In a preferred solution, after the second characteristic information is output, the blood vessel information therein is screened and processed, and the specific process includes:

[0037] Acquire a blood vessel region image, and perform image segmentation on the blood vessel region image to extract a blood vessel network;

[0038] Extracting a vascular calcification region and characteristic parameters of the vascular calcification region from the vascular network, wherein the characteristic parameters include the area and distribution of the calcification region;

[0039] Obtaining a second calculation function, and inputting the characteristic parameters of the vascular calcification area into the second calculation function, and calibrating the output result thereof as a third parameter to be screened;

[0040] Obtaining a second allowable deviation interval, and performing an offset process on each third parameter to be screened according to the second allowable deviation interval to obtain a plurality of second offset intervals;

[0041] Compare each of the second offset intervals one by one, and classify the second offset intervals with overlapping values ​​as intervals with the same attribute;

[0042] The proportion of each same attribute interval in the second offset interval is obtained, and the third to-be-screened parameter corresponding to the second offset interval with the largest proportion is retained as the second characteristic information, and the third to-be-screened parameters corresponding to other second offset intervals are screened out from the second characteristic information.

[0043] In a preferred embodiment, the step of obtaining the hematological index of the hemodialysis patient, combining it with the standard index, and marking the calculation result as the third characteristic information includes:

[0044] Obtain hematological indicators of hemodialysis patients and their corresponding standard indicators;

[0045] Converting the hematological indexes and standard indexes into numerical data;

[0046] Obtain a third calculation function, input the hematological index and the standard index into the third calculation function, and mark the output result as the third characteristic information.

[0047] In a preferred embodiment, the step of aggregating the second characteristic information and the third characteristic information of each evaluation group into a training set and a test set, and training them in combination with a convolutional neural network model to obtain an aging clock evaluation model includes:

[0048] Acquire an evaluation group of hemodialysis patients, and integrate the second characteristic information and the third characteristic information of each evaluation group;

[0049] The integrated second feature information and the third feature information are equally divided into a plurality of training sets and test sets, wherein the training set is used to train the model, and the test set is used to verify the performance of the model, and each of the training sets can be used as a test set, and each of the test sets can also be used as a training set;

[0050] Obtain a convolutional neural network model, and input the training set into the model for training, use the test set to verify the training result, and obtain the verification accuracy;

[0051] The required accuracy is obtained, and after the verification accuracy reaches the required accuracy, the training of the convolutional neural network model is stopped, and the output is an aging clock evaluation model.

[0052] The present invention also provides an artificial intelligence-based hemodialysis population aging clock assessment system, which is applied to the artificial intelligence-based hemodialysis population aging clock assessment method, comprising:

[0053] A data acquisition module, wherein the data acquisition module is used to collect medical image data of hemodialysis patients, and preprocess and extract features of the data to obtain first feature information;

[0054] A data classification module, the data classification module is used to obtain a pre-classification condition, and classify the first feature information according to the pre-classification condition to obtain a plurality of second feature information;

[0055] An index collection module, which is used to obtain hematological indexes of hemodialysis patients, and combine them with standard indexes for calculation, and mark the calculation results as third characteristic information;

[0056] A group classification module, the group classification module is used to obtain physical sign parameters and medical history information, and classify hemodialysis patients into multiple evaluation groups based on the physical sign parameters and medical history information;

[0057] A model building module, wherein the model building module is used to aggregate the second characteristic information and the third characteristic information of each evaluation group into a training set and a test set, and train them in combination with a convolutional neural network model to obtain an aging clock evaluation model;

[0058] An evaluation module is used to obtain a target hemodialysis patient, match the evaluation group to which the patient belongs, and then call a corresponding aging clock evaluation model to obtain the aging clock of the target hemodialysis patient.

[0059] And, an artificial intelligence-based aging clock assessment device for hemodialysis population, including:

[0060] at least one processor;

[0061] and a memory communicatively coupled to the at least one processor;

[0062] Wherein, the memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the above-mentioned artificial intelligence-based aging clock assessment method for hemodialysis population.

[0063] The technical effects achieved by the present invention are:

[0064] The present invention collects medical image data of hemodialysis patients, obtains first feature information by data preprocessing and feature extraction technology, further classifies the first feature information according to the pre-classification condition to obtain multiple second feature information, and simultaneously obtains the patient's hematological indicators, and combines them with standard indicators to obtain third feature information, and then classifies the patients into multiple evaluation groups according to physical sign parameters and medical history information. Finally, the second feature information and the third feature information of each evaluation group are trained in combination with a convolutional neural network model to obtain an aging clock evaluation model for different evaluation groups. By applying the aging clock evaluation model to actual scenarios, the aging degree of the hemodialysis population can be accurately evaluated, and individualized treatment recommendations can be provided to doctors. At the same time, the aging clock evaluation model for different evaluation groups can more comprehensively consider the characteristics of different evaluation groups and improve the accuracy and reliability of the evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is a schematic flow chart of the method of the present invention;

[0066] Figure 2 It is a schematic diagram of the system module of the present invention;

[0067] Figure 3 It is a structural diagram of the device of the present invention. DETAILED DESCRIPTION

[0068] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0069] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0070] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure or characteristic that may be included in at least one implementation of the present invention. The term "in a preferred embodiment" that appears in different places in this specification does not refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0071] See also Figure 1 As shown, the present invention provides an artificial intelligence-based hemodialysis population aging clock assessment method, comprising:

[0072] S1. Collect medical image data of hemodialysis patients, and preprocess and extract features of the data to obtain first feature information;

[0073] S2. Obtaining a pre-classification condition, and classifying the first feature information according to the pre-classification condition to obtain a plurality of second feature information;

[0074] S3, obtaining hematological indicators of hemodialysis patients, and combining them with standard indicators for calculation, and marking the calculation results as third characteristic information;

[0075] S4, obtaining physical sign parameters and medical history information, and classifying hemodialysis patients into multiple evaluation groups based on the physical sign parameters and medical history information;

[0076] S5, summarizing the second feature information and the third feature information of each evaluation group into a training set and a test set, and training them in combination with a convolutional neural network model to obtain an aging clock evaluation model;

[0077] S6. Obtain a target hemodialysis patient and match the evaluation group to which the patient belongs, then call the corresponding aging clock evaluation model to obtain the aging clock of the target hemodialysis patient.

[0078] As described in the above steps S1-S5, when evaluating the aging clock of the hemodialysis population, the medical image data of the hemodialysis patients are first collected. These data include but are not limited to the patients' CT, MRI and other imaging data. After the data are collected, the data needs to be preprocessed, such as denoising, scaling and other operations to eliminate irrelevant information in the data and improve the accuracy of subsequent feature extraction. After the preprocessing is completed, the data is feature extracted to obtain the first feature information. The first feature information includes vascular structure, muscle tissue, etc. After obtaining the first feature information, it is necessary to obtain the pre-classification conditions and classify the first feature information according to these conditions. After the classification, multiple second feature information can be obtained, and then the hematological indicators of the hemodialysis patients, such as hemoglobin, platelets, etc., are collected and combined with the standard indicators. The result of the combined operation can be used as the second The third characteristic information provides the corresponding evaluation basis for the aging clock evaluation. Then, the physical parameters of hemodialysis patients, such as heart rate, blood pressure, etc., and medical history information, such as illness duration, medication status, etc., are obtained. Based on this information, hemodialysis patients can be classified into multiple evaluation groups to achieve more targeted evaluation. Then, the second characteristic information and the third characteristic information under each evaluation group are summarized into training sets and test sets, and trained in combination with the convolutional neural network (CNN) model. After the training is completed, the model is verified using the test set to ensure the effectiveness and accuracy of the model. In actual applications, the target hemodialysis patients are obtained and matched with the evaluation group to which they belong. Then, the corresponding aging clock evaluation model is called to obtain the aging clock of the target hemodialysis patient, which provides a strong basis for clinicians and helps to formulate more accurate treatment plans and health management strategies.

[0079] In a preferred embodiment, the step of collecting medical image data of hemodialysis patients, preprocessing the data and extracting features to obtain first feature information includes:

[0080] S101, obtaining CT or MRI medical image data of a hemodialysis patient, and then preprocessing the image data, wherein the preprocessing includes denoising, enhancement, and segmentation to eliminate artifacts in the image and obtain a reference image;

[0081] S102, obtaining a demand feature corresponding to the aging clock, and performing segmentation processing in the reference image to obtain a plurality of feature-attached regions;

[0082] S103, cross-comparing the similar characteristic attached regions corresponding to the respective hemodialysis patients, and outputting the comparison result as the first parameter to be screened;

[0083] S104, obtaining a screening threshold, and comparing the screening threshold with the first parameter to be screened;

[0084] If the first parameter to be screened is greater than the screening threshold, the demand feature in the corresponding feature-attached area is marked as the first feature information;

[0085] If the first parameter to be screened is less than or equal to the screening threshold, the demand feature in the corresponding feature-attached area is marked as invalid feature information.

[0086] As described in the above steps S101-S104, after the CT or MRI medical imaging data of the hemodialysis patient is acquired, preprocessing is performed, including denoising, enhancement and segmentation. These preprocessing methods help to eliminate artifacts in the image and improve the image quality. After the preprocessing is completed, it is necessary to obtain the demand features corresponding to the aging clock. These features can help us better understand the physiological needs of the patient, thereby providing a basis for diagnosis and treatment and nursing. In order to extract these features, the demand features are segmented in the reference image to obtain multiple feature-attached regions. The similar feature-attached regions corresponding to each hemodialysis patient are cross-compared, and the comparison results are output as the first parameter to be screened. The first parameter to be screened is used to evaluate the patient's demand features, specifically by obtaining a screening threshold and comparing the screening threshold with the first parameter to be screened. During the comparison process, if the first parameter to be screened is greater than the screening threshold, it means that the demand features in the feature-attached region have certain clinical significance, and it can be calibrated as the first feature information. On the contrary, if the first parameter to be screened is less than or equal to the screening threshold, it means that the demand features in the feature-attached region may not contribute much to diagnosis and treatment and nursing work, and it can be calibrated as invalid feature information.

[0087] In a preferred embodiment, the step of cross-comparing the similar characteristic attached regions corresponding to the respective hemodialysis patients and outputting the comparison results as the first parameter to be screened includes:

[0088] Step 1, obtain the same characteristic attached regions corresponding to the hemodialysis patients;

[0089] Step 2, aligning the similar feature-attached regions corresponding to each hemodialysis patient one by one to obtain aligned feature-attached regions;

[0090] Step 3, obtain the first measurement function;

[0091] Step 4: Compare the aligned feature-attached areas at the pixel level, calculate the difference between the pixel values ​​according to the first measurement function, quantify the difference, and mark the quantification result as the first parameter to be screened.

[0092] As described in the above steps Step 1-Step 4, due to the physical condition of each patient, in this implementation, the relevant features of the hemodialysis patients are cross-compared. First, the similar feature affiliated areas corresponding to each hemodialysis patient and the area where the demand feature is located are obtained, and then the similar feature affiliated areas corresponding to each hemodialysis patient are aligned one by one, and then the first measurement function is called to perform pixel-level comparison. The measurement function is used to calculate the difference between the pixel values ​​between the aligned feature affiliated areas, where the expression of the first measurement function is: , where r represents the difference between pixels in the aligned feature-attached area, and n represents the number of pixels in the feature-attached area. x i and y i They respectively represent the pixel values ​​in different feature-attached areas, and then the calculated differences are quantized. This process can be achieved by mapping the differences to a specific interval, such as converting the differences into a value between 0 and 1. In this way, the differences between the various feature-attached areas can be intuitively represented as an easy-to-understand quantitative indicator. Finally, the result after quantization is calibrated as the first parameter to be screened.

[0093] In a preferred embodiment, the step of obtaining the pre-classification condition and classifying the first feature information according to the pre-classification condition to obtain a plurality of second feature information includes:

[0094] S201, obtaining pre-classification conditions, where the pre-classification conditions include body surface information and blood vessel information of the hemodialysis patient;

[0095] S202, classifying the first characteristic information according to the pre-classification condition, and dividing the hemodialysis patients into different subgroups;

[0096] S203: Obtain a demand indicator, and extract corresponding second feature information from a corresponding subgroup according to the demand indicator.

[0097] As described in the above steps S201-S203, after the first characteristic information is output, the pre-classification conditions are obtained. The pre-classification conditions mainly include the body surface information and vascular information of the hemodialysis patient. The body surface information includes but is not limited to the basic parameters of the patient's age, gender, height, weight, etc., while the vascular information mainly includes access methods such as arteriovenous fistula, arteriovenous graft and long-term catheter. The first characteristic information is then classified and processed according to the pre-classification conditions. The hemodialysis patients are divided into different subgroups according to the basic information and vascular conditions of the patients. The purpose is to group patients with similar characteristics into one category for subsequent data analysis. Finally, the demand indicators are obtained and the corresponding second characteristic information is extracted in the corresponding subgroup according to the demand indicators.

[0098] In a preferred embodiment, after the second characteristic information is output, the body surface information therein is screened and processed, and the specific process includes:

[0099] Acquire a skin region image of a hemodialysis patient, and calculate statistical information of the number of pixels and pixel values ​​of the skin region image;

[0100] Based on statistical information, the skin area is determined according to the pixel value distribution, and the edge detection is performed on the skin area to determine the skin boundary;

[0101] Calculating the skin area of ​​the skin region according to the inflection point of the skin boundary and calibrating it as the second parameter to be screened;

[0102] Obtaining a first allowable deviation interval, and performing an offset process on each second parameter to be screened according to the first allowable deviation interval to obtain a plurality of first offset intervals;

[0103] Compare the first offset intervals one by one, and classify the first offset intervals with overlapping values ​​into intervals of the same category;

[0104] The proportion of each same category interval in the first offset interval is obtained, and the second to-be-screened parameter corresponding to the first offset interval with the largest proportion is retained as the second characteristic information, and the second to-be-screened parameters corresponding to other first offset intervals are screened out from the second characteristic information.

[0105] In this embodiment, after the second characteristic information is output, the body surface information contained therein needs to be screened and processed. First, a skin area image of a hemodialysis patient is obtained, and statistical information of the number of pixels and pixel values ​​of the skin area image is counted. Then, based on the statistical information, the skin area is determined according to the pixel value distribution, and edge detection is performed on the skin area to determine the boundary of the skin. After the skin boundary is determined, the skin area of ​​the skin area needs to be calculated, and this area is calibrated as the second parameter to be screened. Then, a preset first allowable deviation interval is called, and each second parameter to be screened is offset according to the first allowable deviation interval to obtain multiple first offset intervals. Then, they are compared one by one, and the first offset intervals with intersecting values ​​are classified as the same category intervals. Finally, the proportion of each same category interval in all first offset intervals is obtained, and the first offset interval with the largest proportion is found, and the corresponding second parameter to be screened is retained as the second characteristic information. At the same time, the second parameters to be screened corresponding to other first offset intervals are screened out from the second characteristic information to avoid accidental factors affecting subsequent analysis results.

[0106] In a preferred embodiment, after the second characteristic information is output, the blood vessel information therein is screened and processed, and the specific process includes:

[0107] Acquire a blood vessel region image, perform image segmentation on the blood vessel region image, and extract a blood vessel network;

[0108] Extracting a vascular calcification region and characteristic parameters of the vascular calcification region from the vascular network, wherein the characteristic parameters include the area and distribution of the calcification region;

[0109] Obtaining a second calculation function, inputting characteristic parameters of the vascular calcification region into the second calculation function, and calibrating the output result thereof as a third parameter to be screened;

[0110] Obtaining a second allowable deviation interval, and performing an offset process on each third parameter to be screened according to the second allowable deviation interval to obtain a plurality of second offset intervals;

[0111] Compare the second offset intervals one by one, and classify the second offset intervals with overlapping values ​​as intervals with the same attribute;

[0112] The proportion of each interval with the same attribute in the second offset interval is obtained, and the third parameter to be screened corresponding to the second offset interval with the largest proportion is retained as the second characteristic information, and the third parameters to be screened corresponding to other second offset intervals are screened out from the second characteristic information.

[0113] In this embodiment, after the second characteristic information is output, the vascular region image is obtained, and the vascular region image is segmented to distinguish the vascular region from other tissue regions, thereby extracting the vascular network. After the vascular network extraction is completed, the vascular calcification region is identified, and the area and distribution of the calcification region are determined, and then the second measurement function is obtained. The second measurement function is used to convert the characteristic parameters of the vascular calcification region into a numerical index, and this index is called the third parameter to be screened, wherein the expression of the second measurement function is: , where u represents the average calcification area, S represents the total area of ​​the calcification area, and D represents the distribution of the calcification area. After obtaining the third parameter to be screened, the preset second allowable deviation interval is called, and each third parameter to be screened is offset according to it to obtain multiple second offset intervals. Finally, each second offset interval is compared, and the second offset intervals with intersections are classified as intervals with the same attribute. After obtaining each interval with the same attribute, their proportion in the second offset interval is calculated to determine which second offset interval has a more representative value, and the third parameter to be screened corresponding to the second offset interval with the largest proportion is retained as the second feature information. At the same time, the third parameters to be screened corresponding to other second offset intervals are screened out from the second feature information to ensure the accuracy of the feature information.

[0114] In a preferred embodiment, the step of obtaining the hematological index of the hemodialysis patient, combining it with the standard index, and marking the calculation result as the third characteristic information includes:

[0115] S301, obtaining hematological indicators of hemodialysis patients and their corresponding standard indicators;

[0116] S302, converting the hematological indexes and standard indexes into numerical data;

[0117] S303, obtaining a third calculation function, and inputting the hematological index and the standard index into the third calculation function, and marking the output result as the third characteristic information.

[0118] As described in the above steps S301-S303, obtaining the hematological indicators of hemodialysis patients involves multiple, such as red blood cell count, white blood cell count, platelet count, hemoglobin concentration, parathyroid hormone, β2 microglobulin, ferritin, transferrin saturation, total iron binding capacity, erythropoietin, etc., and then converting the hematological indicators and standard indicators into numerical data to facilitate subsequent calculations and analysis, unifying the hematological indicators and standard indicators into numerical data format, and then obtaining a third measurement function, and inputting the hematological indicators and standard indicators into the third measurement function, and calibrating its output result as the third characteristic information, wherein the expression of the third measurement function is: , where w represents the third feature information, k b Represents standard indicators, ky Represents the hematological indicators of hemodialysis patients. Based on the above formula, the output third feature information can provide corresponding data support for the subsequent training of the convolutional neural network model.

[0119] In a preferred embodiment, the second feature information and the third feature information of each evaluation group are aggregated into a training set and a test set, and a convolutional neural network model is combined for training to obtain an aging clock evaluation model, including:

[0120] S501, obtaining an evaluation group of hemodialysis patients, and integrating the second characteristic information and the third characteristic information of each evaluation group;

[0121] S502, dividing the integrated second feature information and the third feature information into multiple training sets and test sets, wherein the training set is used to train the model, and the test set is used to verify the performance of the model, and each training set can be used as a test set, and each test set can also be used as a training set;

[0122] S503, obtaining a convolutional neural network model, and inputting the training set into the model for training, using the test set to verify the training result, and obtaining the verification accuracy;

[0123] S504, obtaining the required accuracy, and after verifying that the accuracy reaches the required accuracy, stopping the training of the convolutional neural network model, and outputting the aging clock assessment model.

[0124] As described in the above steps S501-S504, after obtaining the evaluation group, the second feature information and the third feature information under each evaluation group are integrated, and the feature information in different evaluation groups is merged. Then, the integrated second feature information and the third feature information are equally divided into multiple training sets and test sets. These training sets and test sets will be used to train and verify the convolutional neural network model. It is worth noting that each training set can be used as a test set, and each test set can also be used as a training set. This design can ensure the generalization ability of the model during the training process and improve the accuracy of the model. After preparing the training and test sets, it is necessary to obtain a convolutional neural network model and train the training set to learn the mapping relationship between the characteristics of hemodialysis patients and the aging clock. After the training is completed, the test set will be used to verify the training results and calculate the verification accuracy. After the verification accuracy reaches the expected target, it is necessary to obtain the required accuracy. The required accuracy is used to determine whether the model meets the actual application requirements. When the verification accuracy continuously reaches the required accuracy, the training of the convolutional neural network model will be stopped, and this model will be recorded as an aging clock evaluation model for subsequent application in actual scenarios.

[0125] See also Figure 2 , an artificial intelligence-based hemodialysis population aging clock assessment system, applied to the above-mentioned artificial intelligence-based hemodialysis population aging clock assessment method, includes:

[0126] A data acquisition module, which is used to collect medical image data of hemodialysis patients, and preprocess and extract features from the data to obtain first feature information;

[0127] A data classification module, the data classification module is used to obtain a pre-classification condition, and classify the first feature information according to the pre-classification condition to obtain a plurality of second feature information;

[0128] The index collection module is used to obtain the hematological index of the hemodialysis patient, and combine it with the standard index, and mark the calculation result as the third characteristic information;

[0129] A group classification module is used to obtain physical sign parameters and medical history information, and classify hemodialysis patients into multiple evaluation groups based on the physical sign parameters and medical history information;

[0130] A model building module, which is used to aggregate the second characteristic information and the third characteristic information of each evaluation group into a training set and a test set, and train them in combination with a convolutional neural network model to obtain an aging clock evaluation model;

[0131] The evaluation module is used to obtain the target hemodialysis patient and match the evaluation group to which he belongs, and then call the corresponding aging clock evaluation model to obtain the aging clock of the target hemodialysis patient.

[0132] As mentioned above, the evaluation system includes a data acquisition module, a data classification module, an indicator acquisition module, a group classification module, a model building module and an evaluation module. The data acquisition module is responsible for collecting medical image data of hemodialysis patients, such as CT, MRI, etc. These data will serve as the basic information of the evaluation system. During the data collection process, the original data will be preprocessed, such as denoising, scaling, etc., to improve the quality of the data. Next, the data will be feature extracted to obtain the first feature information. These characteristic information will be used as input parameters of the evaluation system. The data classification module is responsible for obtaining the pre-classification conditions and classifying the first characteristic information according to these conditions. The classified data will obtain multiple second characteristic information, which is helpful for further analyzing the health status of hemodialysis patients. The indicator collection module is responsible for collecting the hematological indicators of hemodialysis patients, such as hemoglobin, platelets, etc. In addition, the system also combines these indicators with standard indicators to obtain third characteristic information. The group classification module is used to obtain the patient's physical parameters and medical history information, such as blood pressure, heart rate, weight, etc. Based on this information, hemodialysis patients are classified into multiple evaluation groups, so that in the subsequent analysis and model construction, more accurate aging clock evaluation can be performed for different groups. The model construction module is responsible for summarizing the second and third characteristic information of each evaluation group into training sets and test sets. Subsequently, it is combined with the convolutional neural network model (CNN) for training to obtain the aging clock evaluation model. The evaluation module is used to obtain the target hemodialysis patient and match the evaluation group to which he belongs. Then, the corresponding aging clock evaluation model is called to obtain the aging clock of the target patient, which provides strong support for clinical diagnosis and rehabilitation work and improves the quality of life and prognosis of patients.

[0133] See also Figure 3 , an artificial intelligence-based aging clock assessment device for hemodialysis population, including:

[0134] at least one processor;

[0135] and a memory communicatively coupled to the at least one processor;

[0136] Among them, the memory stores a computer program that can be executed by at least one processor, and the computer program is executed by at least one processor so that the at least one processor can execute the above-mentioned artificial intelligence-based aging clock assessment method for hemodialysis population.

[0137] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0138] The above is only a preferred embodiment of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications should also be considered as the protection scope of the present invention. The structures, devices and operating methods not specifically described and explained in the present invention shall be implemented according to the conventional means in the art unless otherwise specified and limited.

Claims

1. An artificial intelligence-based aging clock assessment method for hemodialysis population, characterized by: include: Collecting medical image data of hemodialysis patients, and performing preprocessing and feature extraction on the data to obtain first feature information; Acquire a pre-classification condition, and classify the first characteristic information according to the pre-classification condition to obtain a plurality of second characteristic information; Obtaining hematological indicators of hemodialysis patients, combining them with standard indicators, and marking the calculation results as third characteristic information; Acquiring physical sign parameters and medical history information, and classifying hemodialysis patients into multiple evaluation groups based on the physical sign parameters and medical history information; The second characteristic information and the third characteristic information of each evaluation group are aggregated into a training set and a test set, and are trained in combination with a convolutional neural network model to obtain an aging clock evaluation model; Obtain a target hemodialysis patient, match the evaluation group to which the patient belongs, and then call the corresponding aging clock evaluation model to obtain the aging clock of the target hemodialysis patient; The step of collecting medical image data of hemodialysis patients, preprocessing the data and extracting features to obtain first feature information includes: Acquire CT or MRI medical image data of hemodialysis patients, and then preprocess the image data, wherein the preprocessing includes denoising, enhancement, and segmentation to eliminate artifacts in the image and obtain a reference image; Obtaining the required features corresponding to the aging clock, and performing segmentation processing in the reference image to obtain a plurality of feature-attached regions; Cross-comparison is performed on the similar characteristic attached regions corresponding to each hemodialysis patient, and the comparison result is output as the first parameter to be screened; Obtaining a screening threshold, and comparing the screening threshold with a first parameter to be screened; If the first parameter to be screened is greater than the screening threshold, the demand feature in the corresponding feature-attached area is marked as the first feature information; If the first parameter to be screened is less than or equal to the screening threshold, the demand feature in the corresponding feature-attached area is marked as invalid feature information; The step of cross-comparing the similar characteristic attached regions corresponding to the respective hemodialysis patients and outputting the comparison results as the first parameter to be screened includes: Obtain similar feature-attached regions corresponding to hemodialysis patients; Aligning similar feature-attached regions corresponding to each hemodialysis patient one by one to obtain aligned feature-attached regions; Get the first calculation function. The expression of the first calculation function is: , where r represents the difference between pixels in the aligned feature-attached area, and n represents the number of pixels in the feature-attached area. x i and y i Respectively represent the pixel values ​​in the regions attached to different features; Comparing each aligned feature-attached region at the pixel level, calculating the difference between the pixel values ​​according to the first measurement function, quantifying the difference, and marking the quantification result as the first parameter to be screened; The step of obtaining the pre-classification condition and classifying the first feature information according to the pre-classification condition to obtain a plurality of second feature information includes: Acquiring pre-classification conditions, wherein the pre-classification conditions include body surface information and blood vessel information of a hemodialysis patient; Classify the first characteristic information according to the pre-classification condition to divide the hemodialysis patients into different subgroups; Obtaining a demand indicator, and extracting corresponding second feature information from a corresponding subgroup according to the demand indicator; After the second characteristic information is output, the body surface information therein is screened and processed, and the specific process includes: Acquire a skin region image of a hemodialysis patient, and calculate statistical information of the number of pixels and pixel values ​​of the skin region image; Based on statistical information, the skin area is determined according to the pixel value distribution, and the edge detection is performed on the skin area to determine the skin boundary; Calculating the skin area of ​​the skin region according to the inflection point of the skin boundary and marking it as the second parameter to be screened; Obtaining a first allowable deviation interval, and performing an offset process on each second parameter to be screened according to the first allowable deviation interval to obtain a plurality of first offset intervals; Compare the first offset intervals one by one, and classify the first offset intervals with overlapping values ​​into intervals of the same category; Obtaining the proportion of each same category interval in the first offset interval, retaining the second to-be-screened parameter corresponding to the first offset interval with the largest proportion as the second characteristic information, and filtering out the second to-be-screened parameters corresponding to other first offset intervals from the second characteristic information; After the second characteristic information is output, the blood vessel information therein is screened and processed, and the specific process includes: Acquire a blood vessel region image, and perform image segmentation on the blood vessel region image to extract a blood vessel network; Extracting a vascular calcification region and characteristic parameters of the vascular calcification region from the vascular network, wherein the characteristic parameters include the area and distribution of the calcification region; A second calculation function is obtained, and the characteristic parameters of the vascular calcification area are input into the second calculation function, and the output result thereof is calibrated as the third parameter to be screened. The expression of the second calculation function is: , where u represents the average calcification area, S represents the total area of ​​the calcification area, and D represents the distribution of the calcification area; Obtaining a second allowable deviation interval, and performing an offset process on each third parameter to be screened according to the second allowable deviation interval to obtain a plurality of second offset intervals; Compare each of the second offset intervals one by one, and classify the second offset intervals with overlapping values ​​as intervals with the same attribute; The proportion of each same attribute interval in the second offset interval is obtained, and the third to-be-screened parameter corresponding to the second offset interval with the largest proportion is retained as the second characteristic information, and the third to-be-screened parameters corresponding to other second offset intervals are screened out from the second characteristic information.

2. The artificial intelligence-based aging clock assessment method for hemodialysis population according to claim 1, characterized in that: The step of obtaining the hematological index of the hemodialysis patient, combining it with the standard index, and marking the calculation result as the third characteristic information includes: Obtain hematological indicators of hemodialysis patients and their corresponding standard indicators; Converting the hematological indexes and standard indexes into numerical data; Obtain a third calculation function, input the hematological index and the standard index into the third calculation function, and mark the output result as the third characteristic information.

3. The artificial intelligence-based aging clock assessment method for hemodialysis population according to claim 1, characterized in that: The step of aggregating the second characteristic information and the third characteristic information of each evaluation group into a training set and a test set, and training them in combination with a convolutional neural network model to obtain an aging clock evaluation model includes: Acquire an evaluation group of hemodialysis patients, and integrate the second characteristic information and the third characteristic information of each evaluation group; The integrated second feature information and the third feature information are equally divided into a plurality of training sets and test sets, wherein the training set is used to train the model, and the test set is used to verify the performance of the model, and each of the training sets can be used as a test set, and each of the test sets can also be used as a training set; Obtain a convolutional neural network model, and input the training set into the model for training, use the test set to verify the training result, and obtain the verification accuracy; The required accuracy is obtained, and after the verification accuracy reaches the required accuracy, the training of the convolutional neural network model is stopped, and the output is an aging clock evaluation model.

4. An artificial intelligence-based hemodialysis population aging clock assessment system, applied to an artificial intelligence-based hemodialysis population aging clock assessment method as claimed in any one of claims 1 to 3, characterized in that: include: A data acquisition module, wherein the data acquisition module is used to collect medical image data of hemodialysis patients, and preprocess and extract features of the data to obtain first feature information; A data classification module, the data classification module is used to obtain a pre-classification condition, and classify the first feature information according to the pre-classification condition to obtain a plurality of second feature information; An index collection module, which is used to obtain hematological indexes of hemodialysis patients, and combine them with standard indexes for calculation, and mark the calculation results as third characteristic information; A group classification module, the group classification module is used to obtain physical sign parameters and medical history information, and classify hemodialysis patients into multiple evaluation groups based on the physical sign parameters and medical history information; A model building module, wherein the model building module is used to aggregate the second characteristic information and the third characteristic information of each evaluation group into a training set and a test set, and train them in combination with a convolutional neural network model to obtain an aging clock evaluation model; An evaluation module is used to obtain a target hemodialysis patient, match the evaluation group to which the patient belongs, and then call a corresponding aging clock evaluation model to obtain the aging clock of the target hemodialysis patient.

5. An artificial intelligence-based aging clock assessment device for hemodialysis patients, characterized by: include: at least one processor; and a memory communicatively coupled to the at least one processor; Wherein, the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the artificial intelligence-based aging clock assessment method for hemodialysis population as described in any one of claims 1 to 3.

Citation Information

Patent Citations

  • Digitalized pulse diagnosis-based quantification method for on-line monitoring and evaluation of human body qi-blood aging

    CN101884527A

  • Method for predicting age by using blood marker

    CN112466402A

  • System for evaluating curative effect of end-stage nephropathy and predicting complication risk

    CN116705326A

  • Vascular senescence clinical evaluation system establishment method and device

    CN116798618A

  • Pathological picture auxiliary diagnosis method and system based on artificial intelligence

    CN118039125A