Machine learning-based vascular disease type identification method, system, medium, terminal and program product

Abstract

The application provides a kind of based on machine learning vascular disease type identification method, system, medium, terminal and program product, method includes: respectively obtaining atherosclerosis patient and abdominal aortic aneurysm patient in multiple time nodes collected ultrasonic imaging data, clinical physiological and biochemical data and lipidomics data;Based on feature extraction network model, ultrasonic imaging features, clinical physiological and biochemical features and lipidomics features are extracted, and multi-modal weighted fusion is carried out;Multi-modal fusion features with time sequence association information are input into machine learning model for training to obtain and deploy vascular disease type identification model, for identifying the vascular disease type of current user to be identified;Vascular disease type includes: atherosclerosis and abdominal aortic aneurysm.The application can realize the accurate and effective differentiation and risk assessment of atherosclerosis and abdominal aortic aneurysm under the same model framework, meet the actual needs of early screening of vascular diseases.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, system, medium, terminal and program product for identifying vascular disease types based on machine learning. Background Technology

[0002] Atherosclerosis (AS) and abdominal aortic aneurysm (AAA) are both common vascular diseases that seriously threaten human health. Although they share some epidemiological similarities and multiple risk factors, such as advanced age, male sex, smoking, and hypertension, they differ significantly in their pathogenesis, pathological progression, and clinical intervention strategies. Specifically, atherosclerosis (AS) is mainly characterized by lipid deposition-driven plaque formation, manifested as the accumulation of neutral lipids (such as cholesterol esters and triglycerides) under the vascular intima, leading to luminal narrowing or occlusion; its intervention strategies mainly focus on lipid-lowering therapy, antiplatelet therapy, and improving vascular endothelial function. In contrast, abdominal aortic aneurysm (AAA) is mainly characterized by the destruction and progressive dilation of the vascular wall structure, involving processes such as smooth muscle cell reduction and matrix degradation; its clinical management focuses more on imaging monitoring and surgical or endovascular repair when a certain diameter or growth rate is reached. Therefore, in clinical practice, accurately distinguishing between the two types of diseases and conducting targeted risk assessments is of great significance for developing individualized intervention strategies.

[0003] Currently, the identification and risk assessment of the aforementioned vascular disease types in clinical practice mainly rely on imaging examinations (such as CT angiography and ultrasound) and traditional biochemical indicators (such as blood lipid levels and inflammation-related indicators). While imaging methods can directly reflect changes in vascular structure, they suffer from high examination costs, radiation exposure risks in some examinations, and limited applicability to early screening in the general population. Traditional biochemical indicators, while having some clinical accessibility, provide limited information and are insufficient to characterize the specific metabolic changes in different vascular diseases, making it difficult to support the accurate differentiation between AS and AAA.

[0004] With the development of artificial intelligence technology, researchers have tried to build diagnostic models for the auxiliary diagnosis of vascular diseases. However, most existing technologies are designed for single vascular diseases and cannot accurately identify and distinguish between AS and AAA, which have similar pathological mechanisms and overlapping clinical manifestations. They also cannot conduct accurate risk assessment for these two diseases and cannot meet the actual needs of early screening for vascular diseases.

[0005] Therefore, it is necessary to provide a machine learning-based method, system, medium, terminal, and program product for identifying vascular disease types to solve the aforementioned problems in the existing technology. Summary of the Invention

[0006] In view of the shortcomings of the prior art described above, the purpose of this application is to provide a machine learning-based method, system, medium, terminal and program product for identifying vascular disease types, in order to solve the technical problems that most of the existing technologies are designed for single vascular diseases, and cannot achieve accurate identification and differentiation between AS and AAA, which have similar pathological mechanisms and overlapping clinical manifestations, nor can they perform accurate risk assessment for these two diseases, thus failing to meet the actual needs of early screening for vascular diseases.

[0007] To achieve the above and other related objectives, a first aspect of this application provides a machine learning-based method for identifying vascular disease types, comprising: acquiring ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data collected at multiple time points from patients with atherosclerosis and patients with abdominal aortic aneurysms, respectively; extracting ultrasound imaging features, clinical physiological and biochemical features, and lipidomics features from the acquired ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data respectively based on a pre-constructed feature extraction network model; performing multimodal weighted fusion of the ultrasound imaging features, clinical physiological and biochemical features, and lipidomics features to obtain multimodal fusion features with temporal correlation information; inputting the multimodal fusion features with temporal correlation information into a machine learning model for training to obtain a vascular disease type identification model; and deploying the vascular disease type identification model to identify the vascular disease type of the current user to be identified; wherein the vascular disease types include: atherosclerosis and abdominal aortic aneurysm.

[0008] In some embodiments of the first aspect of this application, the extraction of ultrasound imaging features, clinical physiological and biochemical features, and lipidomics features from the acquired ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data based on a pre-constructed feature extraction network model includes: extracting ultrasound imaging features from the acquired ultrasound imaging data based on a lightweight convolutional neural network model; and extracting clinical physiological and biochemical features and lipidomics features from the acquired clinical physiological and biochemical data and lipidomics data, respectively, based on a fully connected neural network model.

[0009] In some embodiments of the first aspect of this application, the step of performing multimodal weighted fusion of the ultrasound imaging features, the clinical physiological and biochemical features, and the lipidomics features to obtain multimodal fusion features with temporal correlation information includes: performing multimodal weighted fusion of the ultrasound imaging features, the clinical physiological and biochemical features, and the lipidomics features extracted from multiple time points based on attention feature fusion technology to obtain multimodal fusion features with temporal correlation information.

[0010] In some embodiments of the first aspect of this application, the method further includes: preprocessing the acquired ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data to obtain initial data.

[0011] In some embodiments of the first aspect of this application, the method further includes: dividing the initial data into a training set and a validation set according to a preset ratio, wherein the training set is used to train a machine learning model so that it learns patterns and rules from the training set; and the validation set is used for tuning and parameter selection of the machine learning model to evaluate the model performance.

[0012] In some embodiments of the first aspect of this application, the machine learning model employs a logistic regression model.

[0013] To achieve the above and other related objectives, a second aspect of this application provides a machine learning-based vascular disease type identification system, comprising: a data acquisition module for acquiring ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data collected at multiple time points from patients with atherosclerosis and patients with abdominal aortic aneurysms; a feature extraction module for extracting ultrasound imaging features, clinical physiological and biochemical features, and lipidomics features from the acquired ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data respectively based on a pre-constructed feature extraction network model, and performing multimodal weighted fusion of the ultrasound imaging features, clinical physiological and biochemical features, and lipidomics features to obtain multimodal fusion features with temporal correlation information; a model training module for inputting the multimodal fusion features with temporal correlation information into a machine learning model for training to obtain a vascular disease type identification model; and a model deployment module for deploying the vascular disease type identification model to identify the vascular disease type of the current user to be identified; the vascular disease types include: atherosclerosis and abdominal aortic aneurysm.

[0014] To achieve the above and other related objectives, a third aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method.

[0015] To achieve the above and other related objectives, a fourth aspect of this application provides a computer program product comprising computer program code that, when executed on a computer, causes the computer to implement the method.

[0016] To achieve the above and other related objectives, a fifth aspect of this application provides an electronic terminal, including a memory, a processor, and a computer program stored in the memory; the processor executes the computer program to implement the method.

[0017] As described above, the machine learning-based vascular disease type identification method, system, medium, terminal, and program product of this application have the following beneficial effects:

[0018] This invention first acquires ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data from multiple time points in patients with atherosclerosis and abdominal aortic aneurysm. Then, based on a pre-constructed feature extraction network model, ultrasound imaging features, clinical physiological and biochemical features, and lipidomics features are extracted and multimodal weighted fusion is performed to obtain multimodal fusion features with temporal correlation information. These multimodal fusion features with temporal correlation information are then input into a machine learning model for training, thereby training and constructing a vascular disease type identification model. Finally, the vascular disease type identification model is deployed to identify atherosclerosis and abdominal aortic aneurysm. This application can achieve accurate and effective differentiation and risk assessment of AS and AAA within the same model framework, effectively solving the diagnostic confusion caused by the similar clinical manifestations of these two vascular diseases in the early stages, thus providing a reliable basis for disease classification and the formulation of individualized treatment plans. Attached Figure Description

[0019] Figure 1 The diagram shown is a flowchart of a machine learning-based vascular disease type identification method in one embodiment of this application.

[0020] Figure 2 The diagram shown illustrates the working principle of a machine learning-based vascular disease type identification method in one embodiment of this application.

[0021] Figure 3 The diagram shows the effect size distribution of each predictor variable in the training set in a multifactor logistic regression analysis according to an embodiment of this application.

[0022] Figure 4 The diagram shows the receiver operating characteristic (ROC) curves of a vascular disease type identification model in one embodiment of this application on the training and validation sets.

[0023] Figure 5 The diagram shown is a block diagram of a machine learning-based vascular disease type identification system according to an embodiment of this application.

[0024] Figure 6 The diagram shown is a structural schematic of an electronic terminal according to an embodiment of this application. Detailed Implementation

[0025] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0026] In the embodiments of this application, the terms "first" and "second" are used to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that the terms "first" and "second" do not necessarily imply that they are different.

[0027] It should be noted that, in the embodiments of this application, the words "exemplary" or "for example" indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0028] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0029] Before providing a further detailed description of the present invention, the nouns and terms used in the embodiments of the present invention are explained, and the nouns and terms used in the embodiments of the present invention are subject to the following interpretations:

[0030] <1> Machine learning (ML) is a subset of artificial intelligence (AI) that focuses on algorithms that can “learn” patterns in training data and then make accurate inferences about new data. This pattern recognition ability enables machine learning models to make decisions or predictions without explicit, hard-coded instructions.

[0031] <2> Lightweight Convolutional Neural Network (CNN) models are designed to address the difficulty of deploying traditional CNNs in mobile or embedded devices. These networks significantly reduce resource requirements while maintaining high accuracy by reducing the number of parameters and computational cost, making them suitable for scenarios with limited computing power.

[0032] <3> Fully Connected Neural Network (FCN) model: This is the most basic neural network structure. It adopts a transmission method in which adjacent neurons are fully connected. It consists of an input layer, a hidden layer, and an output layer. The input layer receives the raw data, the hidden layer performs feature extraction through weighted summation and activation functions, and the output layer outputs the results according to different task types.

[0033] <4> Attention Feature Fusion (AFF) is a feature fusion method based on attention mechanisms. It aims to address the limitations of traditional feature fusion methods when there are inconsistencies in semantics and scale. By dynamically adjusting feature weights, it achieves more efficient feature combination and is widely used in short-hop connections, long-hop connections, and multimodal data fusion in deep learning.

[0034] <5> Logistic Regression (LR) is a supervised learning algorithm widely used in classification problems. It uses the sigmoid function to map the output of linear regression to between 0 and 1, thereby predicting the probability of an event occurring.

[0035] <6> Linear interpolation is a method for estimating unknown values ​​between two known data points, assuming that the change between the two points is linear. It is often used in scenarios such as data completion, graphics scaling, and animation transitions.

[0036] <7> Logarithmic transformation technique: This is a common data transformation method, mainly used to process data with a right-tailed distribution. It can change the skewness of the data, making the data distribution closer to a normal distribution. It is only applicable to non-zero and non-negative data, and the transformed values ​​are usually between [0-1].

[0037] <8> Z-score standardization is a linear transformation method that converts data into a form with a mean of 0 and a standard deviation of 1 by subtracting the mean from each data point and then dividing by the standard deviation. This method does not change the original distribution shape of the data. Therefore, if the original data is normally distributed, the standardized data will still be normally distributed; if the original data is not normally distributed, standardization will not automatically turn it into a normal distribution.

[0038] <9> 10-fold cross-validation is a commonly used model evaluation method in machine learning. Its core idea is to divide the dataset into 10 equal subsets and repeat the training and validation process on these 10 subsets. Each time, one subset is selected as the validation set and the rest are used as the training set. This can effectively utilize limited data to evaluate the performance of the model and reduce the randomness of the evaluation results.

[0039] <10> Lipidomics is a comprehensive and systematic analysis and identification of lipids and the molecules that interact with them in organisms, tissues, or cells. It aims to understand the structure and function of lipids and reveal the relationship between lipid metabolism and the physiological and pathological processes of cells, organs, and even the body. Currently, lipidomics has been widely used in important fields such as drug development, molecular physiology, molecular pathology, functional genomics, nutrition, and environment and health.

[0040] In clinical practice, atherosclerosis and abdominal aortic aneurysm share certain similarities in pathological basis and clinical manifestations. On the one hand, patients with abdominal aortic aneurysm often have varying degrees of atherosclerotic changes, and the two share overlapping characteristics in terms of abnormal vascular wall structure, inflammatory response, and lipid metabolism disorders. On the other hand, in the early stages of the disease, the clinical manifestations and some imaging signs of the two may also overlap. Against this backdrop, clinical needs have expanded from "whether there is a risk of vascular disease" to "in the presence of vascular lesions or metabolic abnormalities, determining whether they are more likely to correspond to atherosclerosis or abdominal aortic aneurysm." However, there is still a relative lack of technical solutions for quantitative analysis and differentiation of the differences between the two diseases, and a stable and widely applicable identification method has not yet been formed, thus limiting its application in the fine classification and individualized management of vascular diseases. Meanwhile, most existing omics research focuses on scientific analysis, emphasizing differential expression, clustering, or mechanism exploration, but rarely produces model outputs that can be directly used for clinical diagnosis. Even those methods with some discriminative ability often remain at the level of statistical results presentation, lacking intuitively interpretable outputs such as nomograms and odds ratios, making them difficult for clinicians to quickly apply. Therefore, this application provides a machine learning-based method, system, medium, terminal, and program product for identifying vascular disease types, which can accurately and effectively distinguish between AS and AAA and assess their risks within the same model framework, and has strong clinical interpretability.

[0041] To facilitate understanding of the embodiments of this application, in conjunction with Figure 1 and Figure 2 Detailed explanation. Figure 1 The diagram illustrates a flowchart of a machine learning-based method for identifying vascular disease types according to an embodiment of the present invention. Figure 2 This illustration shows a schematic diagram illustrating the working principle of a machine learning-based vascular disease type identification method according to an embodiment of the present invention. The machine learning-based vascular disease type identification method in this embodiment includes the following steps:

[0042] Step S11: Obtain ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data from multiple time points for patients with atherosclerosis and patients with abdominal aortic aneurysms.

[0043] Specifically, the collection of ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data from patients with atherosclerosis and abdominal aortic aneurysms should follow unified clinical testing standards. These three core modalities should be collected simultaneously at multiple time points, taking into account both static baseline characteristics and dynamic temporal changes, thereby constructing a comprehensive disease characteristic dataset. Static baseline acquisition (T0) data includes: ultrasound imaging data such as abdominal aortic wall thickness, diameter, echo intensity, and plaque morphology; clinical physiological and biochemical data such as high-sensitivity CRP (C-Reactive Protein), blood lipids, and blood glucose; and lipidomics data such as the full spectrum of plasma lipid metabolites. To capture the temporal evolution of vascular diseases, standardized dynamic follow-up time windows are set, collecting these three core modalities at multiple time points. For example, T0 (baseline), T1 (3 months), and T2 (6 months) are set as time points. The three modalities are repeatedly collected using the same acquisition equipment, procedures, and parameters as the baseline, thus forming an individual-level temporal dataset to accurately characterize the progression of vascular diseases.

[0044] In some specific embodiments, for cases of slight data loss at a single time point or for a single indicator, linear interpolation technology is used to perform pre-set simple interpolation to complete the missing data, thereby ensuring data integrity without compromising feature interpretability and model simplicity.

[0045] In some embodiments of this application, the method further includes: preprocessing the acquired ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data respectively to obtain initial data.

[0046] A simplified and differentiated preprocessing rule was adopted for the three types of core modal data, and all standardized definitions were deeply bound to the model parameter table to ensure consistency in cross-center inference. Specifically, differentiated preprocessing was implemented for different modalities. For ultrasound imaging data, outliers were first cleaned, and then Z-score standardization was used to unify the dimensions to adapt it to the model. For clinical physiological and biochemical data, fixed binary classification coding rules were used, such as male = 1, female = 0, and medication history = 1, no = 0, to ensure unambiguity and ease of operation. For lipidomics data, log2 transformation was first performed to eliminate the influence of skewed data distribution, and then Z-score standardization was used to unify the feature dimensions. For dynamic time series data at the three time points T0 / T1 / T2, only three low-complexity time series features—level value, slope of change, and fluctuation coefficient—were extracted to ensure no redundant dimensions were added, balancing model robustness and interpretability.

[0047] In some embodiments of this application, the method further includes: dividing the initial data into a training set and a validation set according to a preset ratio, wherein the training set is used to train a machine learning model so that it learns patterns and rules from the training set; and the validation set is used for tuning and parameter selection of the machine learning model to evaluate model performance.

[0048] Specifically, the original data is divided into training and validation sets according to a predetermined ratio. A standardized evaluation system is constructed based on the validation set to comprehensively verify the model's performance and reliability. Basic performance evaluation is conducted using the training / validation set partitioning combined with 10-fold cross-validation. Core metrics include AUC (Area Under Curve), sensitivity, specificity, positive predictive value, negative predictive value, F1 score, Brier score, and calibration curve. Based on this, specific robustness validation is performed to assess the consistency of diagnostic performance across different follow-up time windows, verify the model's adaptability to time-series data, test robustness under missing data and noise perturbations, and examine the model's anti-interference ability. The performance difference before and after matching is compared to verify the effectiveness of confounding bias control. An external validation set is introduced to evaluate generalization performance, ensuring the model's adaptability to complex clinical scenarios such as multi-center and primary care settings.

[0049] Step S12: Based on the pre-constructed feature extraction network model, extract ultrasound imaging features, clinical physiological and biochemical features, and lipidomics features from the acquired ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data, respectively. Then, perform multimodal weighted fusion on the ultrasound imaging features, clinical physiological and biochemical features, and lipidomics features to obtain multimodal fusion features with temporal correlation information.

[0050] In some embodiments of this application, the extraction of ultrasound imaging features, clinical physiological and biochemical features, and lipidomics features from the acquired ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data based on a pre-constructed feature extraction network model includes: extracting ultrasound imaging features from the acquired ultrasound imaging data based on a lightweight convolutional neural network model; and extracting clinical physiological and biochemical features and lipidomics features from the acquired clinical physiological and biochemical data and lipidomics data based on a fully connected neural network model.

[0051] The pre-built feature extraction network model is a lightweight multi-branch deep network model. This model is only used for learning key features and refining weights, and does not directly output the final clinical diagnostic results. Specifically, the feature extraction network model is divided into three branches: For ultrasound imaging data, this branch uses a lightweight MobileNetV2 convolutional neural network model to extract deep representations of ultrasound images; For clinical physiological and biochemical data and lipidomics data, these two branches both use shallow fully connected neural network models to extract the core features of the corresponding modalities. Thus, a dedicated feature extraction branch network is designed according to the characteristics of different modalities and data types, providing high-purity, dimensionally controllable input features for subsequent machine learning models.

[0052] With the development of multi-omics technologies, lipidomics, as an important branch of metabolomics, can systematically detect and quantitatively analyze various lipid molecules in vivo, and is considered to have the potential to reveal differences in metabolic characteristics among different vascular diseases. However, the high dimensionality, multivariate nature, and complex correlations of lipidomics data limit the application of lipid metabolism information in the precise subtyping of vascular diseases. Extracting key features from a large number of lipid molecules and further translating them into clinical decision-making models for vascular disease classification remains a significant challenge. To address this, this embodiment uses a shallow fully connected neural network model to extract and integrate core lipid biomarkers related to AS and AAA from the acquired lipidomics data, providing high-quality input for subsequent machine learning models.

[0053] In some embodiments of this application, the step of performing multimodal weighted fusion of the ultrasound imaging features, the clinical physiological and biochemical features, and the lipidomics features to obtain multimodal fusion features with temporal correlation information includes: performing multimodal weighted fusion of the ultrasound imaging features, the clinical physiological and biochemical features, and the lipidomics features extracted from multiple time points based on attention feature fusion technology to obtain multimodal fusion features with temporal correlation information.

[0054] Specifically, attention feature fusion technology is used to weight and fuse multiple time points and multimodal features. Based on a lightweight attention mechanism, the weights of different features are automatically learned to obtain the fused multimodal fusion features. These multimodal fusion features contain spatial information, temporal correlation information, and modal correlation information, thereby focusing on high-value temporal correlation information and aggregating to obtain dynamic multimodal risk representations. This efficiently mines cross-modal temporal patterns and provides purified high-quality features for subsequent models.

[0055] Step S13: Input the multimodal fusion features with temporal correlation information into the machine learning model for training to obtain a vascular disease type recognition model.

[0056] In some embodiments of this application, the machine learning model employs a logistic regression model.

[0057] Specifically, the key feature weights purified after the multimodal weighted fusion are transferred to a simplified logistic regression model, and the input dimensions are strictly controlled to achieve clinical interpretability and rapid deployment at the grassroots level. For example, the input lipidomics features are limited to 40 key lipid biomarkers and combined with core clinical covariates to construct a linear risk scoring formula, the specific calculation formula of which is as follows:

[0058] ;Formula (1)

[0059] in, This represents the linear prediction value in the logistic regression model; This is the model intercept (constant term); The number of key lipid biomarkers included in the model; The serial number is the lipid biomarker number; For the first Key lipid biomarker values; In order to be with the first lipid markers The corresponding regression coefficients; The number of core clinical covariates; The core clinical covariate number; For the first One core clinical covariate; With the Core clinical covariates The corresponding regression coefficients.

[0060] Based on the risk scoring formula provided in formula (1), the probability of identifying vascular diseases is further calculated, and the formula is as follows:

[0061] ;Formula (2)

[0062] ;Formula (3)

[0063] in, To determine the probability of identifying an abdominal aortic aneurysm; To identify the probability of atherosclerosis.

[0064] It should be noted that the model intercept, feature coefficients, and standardized parameters are all fixed in the model parameter table to achieve reproducible calculations and zero-threshold deployment. Identification, diagnosis, and reasoning can be completed without high-end computing power, and risk classification results are output simultaneously, which is in line with the reading habits of clinicians.

[0065] Because different types of vascular diseases exhibit variations in lipid metabolism patterns, lipidomics, as a technique capable of systematically detecting and quantitatively analyzing multiple lipid molecules, can characterize the differences in metabolic features among different vascular diseases at a holistic level, providing crucial information for disease type differentiation. Therefore, the vascular disease type identification model constructed in this application can fully utilize the multidimensional molecular information provided by lipidomics to uncover the differences in lipid metabolism between AS and AAA, and apply this information to vascular disease identification and risk prediction, thus broadening the application value of lipid metabolism information in the precise subtyping of vascular diseases.

[0066] Step S14: Deploy the vascular disease type identification model to identify the vascular disease type of the current user to be identified; the vascular disease types include: atherosclerosis and abdominal aortic aneurysm.

[0067] Based on lipidomics research, a dynamic follow-up acquisition mechanism is further introduced, and information on the longitudinal changes of lipidomics, including in-depth characterization of ultrasound imaging, clinical physiological and biochemical indicators, and lipidomics, is integrated to construct a two-level model framework of precise multimodal deep learning and simplified interpretable logistic regression model. This addresses the problems of insufficient differential diagnostic ability, poor model robustness, and weak clinical interpretability in existing technologies, providing clinicians with a low-cost, highly specific, and reproducible auxiliary diagnostic tool.

[0068] To further illustrate the machine learning-based vascular disease type identification method provided in this application, specific embodiments are described below. First, a homogeneous research cohort is constructed to avoid confounding bias from the outset, laying a data foundation for model training. Strict inclusion and exclusion criteria are set for subjects in the AS and AAA groups, excluding confounding samples with severe underlying diseases, recent vascular intervention / surgery, severe data gaps, and loss to follow-up, ensuring cohort purity. A 1:1 propensity score matching (PSM) strategy is used to achieve inter-group balance, with the matching caliper strictly limited to ≤0.02 to avoid residual bias due to matching failure. Matching variables cover all confounding factors, including at least demographic indicators such as age, gender, and BMI; history of medication use such as lipid-lowering, antihypertensive, anti-inflammatory, and anticoagulant / antiplatelet drugs; history of vascular-related diseases such as hypertension, diabetes, coronary heart disease, and stroke; and hypersensitive CR. Core biochemical indicators such as P, complete blood lipid profile, and fasting blood glucose were used. After matching, chi-square test and rank-sum test were used to verify the balance between groups to ensure that there were no statistically significant differences in key covariates (P>0.05), reducing the interference of confounding factors on the model. Ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data were collected from AS patients and AAA patients. Specifically, a total of 242 subjects were recruited, including 121 AS patients and 121 AAA patients. All subjects signed informed consent forms. The inclusion criteria for AS patients were coronary artery stenosis ≥50% confirmed by coronary CTA, excluding aortic disease. The inclusion criteria for AAA patients were abdominal aortic diameter ≥30mm confirmed by CTA, excluding atherosclerosis-related diseases. Fasting (≥8 hours) peripheral venous blood was collected from all subjects. After plasma separation, it was stored at −80℃ for later use to ensure the stability of lipid metabolites. Lipidomics analysis was performed on plasma samples from both AS and AAA patients using a non-targeted lipidomics analysis method based on liquid chromatography-tandem mass spectrometry (LC-MS / MS). Under uniform experimental conditions, all samples were batch-tested to ensure technical consistency between different vascular disease groups and avoid batch effect interference. After peak extraction, alignment, and normalization of the detection data, a lipid metabolite expression matrix was constructed to achieve systematic quantitative analysis of the lipid metabolite profile in each group of samples, with a total of 705 lipid metabolites quantitatively detected.

[0069] Then, data preprocessing and cohort division were performed, specifically: the raw lipidomics data were summed, normalized, and automatically scaled to eliminate dimensional differences and data distribution biases among different lipid metabolites; a random sampling method was used to divide the 242 subjects into a training set (n = 169) and a validation set (n = 73) in a 7:3 ratio to ensure that the proportions of AS and AAA patients in the two groups were consistent and to avoid sampling bias. Meanwhile, a progressive multi-level feature screening strategy was adopted for the training set, and the inter-group differences were initially screened, sparse regularization compression and multivariate discriminant refinement were carried out in sequence. The clinical covariates were then entered into the multivariate discriminant model to form the final simplified prediction scheme. Specifically, the abdominal aortic atherosclerosis (AS) group and abdominal aortic aneurysm (AAA) group to be identified were used in the training set. The lipid metabolites were compared with two independent samples. The Student's t test was used to compare the expression differences between the two groups. The screening condition was P<0.05. 252 differential lipid metabolites were obtained (denoted as set S1) for inter-group differences screening. The difference analysis results showed that there were significant differences between AS and AAA in lipid metabolism profiles. Specifically, (1) in the AS group, the concentrations of triglycerides (TAG) and cholesterol esters (CE) were significantly increased; (2) in the AAA group, the concentrations of phosphatidylcholine (PC) were significantly increased. LASSO sparse feature compression is performed using the obtained differential feature set in the R language environment and the glmnet package. LASSO least absolute shrinkage and selection operator (LASSO) regression is used for feature compression and coefficient estimation. The optimal regularization strength parameter λ (λ = 0.0235053) is determined through K (K = 10) fold cross-validation, and lipid features corresponding to non-zero LASSO coefficients are retained to form a compressed feature set (denoted as set S2, n = 24). PLS-DA multivariate discriminant refinement is performed on the obtained feature set S2, calculating the variable importance projection (VIP) score for each variable. Core lipid features are retained according to a preset threshold to form the optimal lipid feature set for distinguishing between AS and AAA (VIP > 1.5, denoted as set S3, n = 20). It should be understood that Student's t-test is mainly used for normal distributions with small sample sizes (e.g., n < 30) and unknown population standard deviation σ. It uses t-distribution theory to infer the probability of differences occurring, thereby comparing whether the difference between two means is significant. LASSO is a linear regression algorithm that achieves feature selection and model sparsity by introducing an L1 regularization term into the loss function. PLS-DA is a widely used multivariate analysis technique in statistics, the core of which lies in its ability to handle situations where there is high collinearity (i.e., high correlation between variables) among explanatory variables.

[0070] Finally, the optimal lipid feature set S3 obtained above, along with the simplest covariates (such as age, gender, and BMI), were input into a multivariate logistic regression model for training / fitting to obtain the final identification model for probabilistic output in vascular disease identification. The Box-Tidwell test and restricted cubic splines (4 nodules) were used to verify the linearity assumption of the logistic regression model. The results showed that all quantitative predictors met the linearity assumption (P>0.05). The performance of all models was evaluated in the validation set, and accuracy, sensitivity, specificity, precision, F1 score, and AUC were calculated. The 95% confidence interval of AUC was calculated using the Bootstrap method (1000 resampling). The DeLong test was used to compare the AUCs of the models pairwise, and the results showed no significant difference in AUC among all models (P>0.05). Combining model performance and clinical interpretability, the logistic regression model was selected as the master model, and a nomogram was constructed for clinical application.

[0071] The beneficial effects of the machine learning-based vascular disease type identification method provided in this application are as follows: (1) Integrated design of identification and risk assessment: Based on the different lipidomics characteristics of AS and AAA patients, and based on the lipidomics characteristics, a vascular disease type identification model is constructed, which enables effective differentiation and risk assessment of AS and AAA under the same model framework. This effectively solves the problem of diagnostic confusion caused by the similar clinical manifestations of the two diseases in the early stage, and provides a reliable basis for disease classification and the formulation of individualized treatment plans; (2) Robust identification performance: By using a progressive feature screening strategy (t test + LASSO + PLS-DA) in the input feature preparation stage to effectively reduce and optimize the dimensionality of high-dimensional lipidomics data, redundant variables are reduced. The interference of the quantity on the model, and combined with 10-fold cross-validation and multiple model performance verification methods, ensure that the model has good generalization ability and stability, with AUC values ​​≥0.83 and high prediction accuracy; (3) Strong clinical interpretability: The vascular disease type identification model finally constructed can output risk assessment results with clear clinical significance, such as variable contribution and risk probability, and can be presented in a visual form, which is convenient for clinicians to understand and apply, thereby improving the model's scalability in actual diagnosis and treatment scenarios; (4) Low application cost and suitable for early screening of the population: Constructed based on plasma lipidomics characteristics, compared with methods that rely on imaging examinations, it has advantages such as lower cost and wider applicability, and is more suitable for early screening and risk assessment of large-scale populations.

[0072] To verify the beneficial effects of the machine learning-based vascular disease type identification method provided in this application, this application uses the constructed vascular disease type identification model to conduct identification and analysis experiments on atherosclerosis (AS) and abdominal aortic aneurysm (AAA). Figure 3This diagram illustrates the effect size distribution of each predictor variable in a multivariate logistic regression analysis within the vascular disease type identification model constructed based on ultrasound imaging features, clinical physiological and biochemical features, and lipidomics features in the training cohort (n = 169). The horizontal axis represents the odds ratio (OR), displayed on a log scale to enhance the comparability of variables of different magnitudes. The vertical axis represents the lipid metabolites and clinical covariates (including age, sex, and body mass index) included in the model. Each point in the diagram represents the OR estimate of the corresponding variable, and the horizontal error bar represents its 95% confidence interval (95% CI). The dashed line represents the reference line for OR = 1. Figure 4 This paper presents the Receiver Operating Characteristic (ROC) curves and discriminative performance of the vascular disease type identification model in this application on the training and validation sets. The horizontal axis represents the false positive rate (FPR, i.e., 1 − specificity), and the vertical axis represents the true positive rate (TPR, i.e., sensitivity). Analysis shows that 20 differentially expressed lipid metabolites were selected for model construction. The logistic regression model had an AUC of 0.968 (95% CI: 0.946–0.990) on the training set and an AUC of 0.841 (95% CI: 0.747–0.935) on the validation set, demonstrating good predictive ability. Further analysis revealed significantly elevated cholesterol ester (CE) levels in AS patients, while relatively high phosphatidylcholine (PC) levels were observed in AAA patients. Furthermore, the nomogram model constructed based on key lipid metabolites effectively distinguished between AS and AAA. The calibration curves showed good consistency between the predicted risk and actual observations, demonstrating good stability and reliability.

[0073] Figure 5 This is a schematic block diagram of a machine learning-based vascular disease type identification system provided in an embodiment of this application. Figure 5 As shown, the machine learning-based vascular disease type identification system 500 includes:

[0074] The data acquisition module 501 is used to acquire ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data collected at multiple time points from patients with atherosclerosis and patients with abdominal aortic aneurysms.

[0075] The feature extraction module 502 is used to extract ultrasound imaging features, clinical physiological and biochemical features, and lipidomics features from the acquired ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data respectively based on a pre-built feature extraction network model, and to perform multimodal weighted fusion of the ultrasound imaging features, clinical physiological and biochemical features, and lipidomics features to obtain multimodal fusion features with temporal correlation information.

[0076] The model training module 503 is used to input multimodal fusion features with temporal correlation information into the machine learning model for training, so as to obtain a vascular disease type recognition model.

[0077] The model deployment module 504 is used to deploy the vascular disease type identification model to identify the vascular disease type of the current user to be identified; the vascular disease types include: atherosclerosis and abdominal aortic aneurysm.

[0078] It should be understood that the specific process of each module performing the above-mentioned corresponding steps has been described in detail in the above method embodiments, and will not be repeated here for the sake of brevity.

[0079] It should also be understood that the module division in the embodiments of this application is illustrative and only represents a logical functional division; in actual implementation, there may be other division methods. Furthermore, the functional modules in the various embodiments of this application can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0080] Figure 6 This is a schematic block diagram of the electronic terminal provided in an embodiment of this application. Figure 6 As shown, the electronic terminal 600 includes at least one processor 601, a memory 602, at least one network interface 603, and a user interface 605. The various components in the electronic terminal 600 are coupled together via a bus system 604. It is understood that the bus system 604 is used to implement communication between these components. In addition to a data bus, the bus system 604 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in… Figure 6 The general will label all buses as bus systems.

[0081] The user interface 605 may include a monitor, keyboard, mouse, trackball, clicker, button, touchpad, or touch screen.

[0082] It is understood that memory 602 can be volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM) or programmable read-only memory (PROM), which serves as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM) and synchronous static random access memory (SSRAM). The memories described in the embodiments of this invention are intended to include, but are not limited to, these and any other suitable categories of memory.

[0083] In this embodiment of the invention, the memory 602 is used to store various types of data to support the operation of the electronic terminal 600. Examples of this data include: any executable program for operation on the electronic terminal 600, such as the operating system 6021 and application programs 6022; the operating system 6021 contains various system programs, such as the framework layer, core library layer, driver layer, etc., for implementing various basic services and handling hardware-based tasks. The application program 6022 may contain various applications, such as a media player, browser, etc., for implementing various application services. The methods provided in this embodiment of the invention can be included in the application program 6022.

[0084] The methods disclosed in the above embodiments of the present invention can be applied to processor 601, or implemented by processor 601. Processor 601 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in processor 601 or by instructions in the form of software. The processor 601 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 601 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. General-purpose processor 601 may be a microprocessor or any conventional processor, etc. The steps of the accessory optimization method provided in the embodiments of the present invention can be directly reflected as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software module may be located in a storage medium, which is located in memory. The processor reads the information in the memory and combines it with its hardware to complete the steps of the aforementioned method.

[0085] In an exemplary embodiment, the electronic terminal 600 may be used by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), or complex programmable logic devices (CPLDs) to perform the aforementioned method.

[0086] According to the method provided in the embodiments of this application, this application also provides a computer program product, which includes: computer program code, which, when run on a computer, causes the computer to execute... Figures 1 to 4 The method of any of the embodiments shown.

[0087] According to the method provided in the embodiments of this application, this application also provides a computer-readable storage medium storing program code, which, when executed on a computer, causes the computer to perform... Figures 1 to 4 The method of any of the embodiments shown.

[0088] As used in this specification, the terms "component," "module," "system," etc., are used to refer to computer-related entities, hardware, firmware, combinations of hardware and software, software, or software in execution. For example, a component can be, but is not limited to, a process running on a processor, a processor, an object, an executable file, an execution thread, a program, and / or a computer. As illustrated, applications running on computing devices and computing devices can both be components. One or more components may reside in a process and / or an execution thread, and components may be located on a single computer and / or distributed among two or more computers. Furthermore, these components can be executed from various computer-readable media on which various data structures are stored. Components can communicate, for example, via local and / or remote processes based on signals having one or more data packets (e.g., data from two components interacting with another component between a local system, a distributed system, and / or a network, such as the Internet interacting with other systems via signals).

[0089] Those skilled in the art will recognize that the various illustrative logical blocks and steps described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0090] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0091] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0092] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0093] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0094] In the above embodiments, the functions of each functional unit can be implemented entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. A computer program product includes one or more computer instructions (programs). When the computer program instructions (programs) are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., high-density digital video discs (DVDs)), or semiconductor media (e.g., solid-state disks (SSDs)).

[0095] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0096] In summary, existing technologies are mostly designed for single vascular diseases, failing to accurately identify and differentiate between AS and AAA, which have similar pathological mechanisms and overlapping clinical manifestations, and also unable to accurately assess the risk of these two diseases, thus failing to meet the practical needs of early screening for vascular diseases. This invention provides a machine learning-based method, system, medium, terminal, and program product for identifying vascular disease types. It first acquires ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data from multiple time points in patients with atherosclerosis and abdominal aortic aneurysms; then, based on a pre-constructed feature extraction network model, it extracts ultrasound imaging features and clinical physiological and biochemical features, respectively. This invention utilizes biochemical and lipidomics characteristics, employing multimodal weighted fusion to obtain multimodal fusion features with temporal correlation information. These features are then input into a machine learning model for training, thereby constructing a vascular disease type identification model. Finally, this model is deployed to identify two vascular diseases: atherosclerosis (AS) and abdominal aortic aneurysm (AAA). This application achieves accurate and effective differentiation and risk assessment of AS and AAA within the same model framework, effectively solving the diagnostic confusion caused by the similar clinical manifestations of these two vascular diseases in their early stages. This provides a reliable basis for disease classification and the development of individualized treatment plans. Therefore, this application effectively overcomes the various shortcomings of existing technologies and has high industrial application value.

[0097] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A method for identifying vascular disease types based on machine learning, characterized in that, include: Ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data were collected at multiple time points from patients with atherosclerosis and patients with abdominal aortic aneurysm. Based on a pre-built feature extraction network model, ultrasound imaging features, clinical physiological and biochemical features, and lipidomics features are extracted from the acquired ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data, respectively. The ultrasound imaging features, clinical physiological and biochemical features, and lipidomics features are then fused using multimodal weighting to obtain multimodal fused features with temporal correlation information. Multimodal fusion features with temporal correlation information are input into a machine learning model for training to obtain a vascular disease type recognition model; The vascular disease type identification model is deployed to identify the vascular disease type of the current user to be identified; the vascular disease types include: atherosclerosis and abdominal aortic aneurysm.

2. The method for identifying vascular disease types based on machine learning according to claim 1, characterized in that, The pre-constructed feature extraction network model extracts ultrasound imaging features, clinical physiological and biochemical features, and lipidomics features from the acquired ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data, respectively, including: Ultrasound imaging features are extracted from the acquired ultrasound imaging data based on a lightweight convolutional neural network model. Furthermore, clinical physiological and biochemical features and lipidomics features are extracted from the acquired clinical physiological and biochemical data and lipidomics data respectively based on a fully connected neural network model.

3. The method for identifying vascular disease types based on machine learning according to claim 2, characterized in that, The step of performing multimodal weighted fusion of the ultrasound imaging features, the clinical physiological and biochemical features, and the lipidomics features to obtain multimodal fused features with temporal correlation information includes: Attention feature fusion technology is used to perform multimodal weighted fusion of the ultrasound imaging features, clinical physiological and biochemical features and lipidomics features extracted from multiple time points to obtain multimodal fusion features with temporal correlation information.

4. The method for identifying vascular disease types based on machine learning according to claim 1, characterized in that, Also includes: The acquired ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data were preprocessed to obtain initial data.

5. The machine learning-based vascular disease type identification method according to claim 4, characterized in that, Also includes: The initial data is divided into a training set and a validation set according to a preset ratio. The training set is used to train the machine learning model so that it learns patterns and rules from the training set. The validation set is used to fine-tune the machine learning model and select parameters to evaluate the model performance.

6. The method for identifying vascular disease types based on machine learning according to claim 1, characterized in that, The machine learning model used is a logistic regression model.

7. A machine learning-based system for identifying vascular disease types, characterized in that, include: The data acquisition module is used to acquire ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data collected at multiple time points from patients with atherosclerosis and patients with abdominal aortic aneurysms. The feature extraction module is used to extract ultrasound imaging features, clinical physiological and biochemical features, and lipidomics features from the acquired ultrasound imaging data, clinical physiological and biochemical data, and lipidomics data respectively based on a pre-built feature extraction network model, and to perform multimodal weighted fusion of the ultrasound imaging features, clinical physiological and biochemical features, and lipidomics features to obtain multimodal fusion features with temporal correlation information. The model training module is used to input multimodal fusion features with temporal correlation information into the machine learning model for training, so as to obtain a vascular disease type recognition model; The model deployment module is used to deploy the vascular disease type identification model to identify the vascular disease type of the current user to be identified; the vascular disease types include: atherosclerosis and abdominal aortic aneurysm.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 6.

9. A computer program product, characterized in that, The computer program product includes computer program code that, when run on a computer, causes the computer to implement the method as described in any one of claims 1 to 6.

10. An electronic terminal, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the method as described in any one of claims 1 to 6.

Images