Abdominal aortic aneurysm progress prediction method and system based on cross-modal knowledge distillation

Abstract

The invention discloses an abdominal aortic aneurysm progress prediction method and system based on cross-modal knowledge distillation. The method comprises the following steps: collecting blood of a target object; the method comprises the following steps: detecting the concentration of blood protein fingerprint markers of blood of a target object, wherein the blood protein fingerprint markers comprise matrix metalloproteinase-12, calcitonin-related polypeptide-alpha, uromodulin, nerve injury induction protein-1, glycosylated hemoglobin A1c and fibroblast growth factor-9; carrying out abdominal aortic aneurysm risk calculation by calling Student-net of cross-modal knowledge distillation based on the blood protein fingerprint marker concentration; wherein the input of the Teaser-net of the cross-modal knowledge distillation comprises a CTA image feature and a pathological section feature, and a category probability vector output by the Student-net is aligned with a category probability vector output by the Teaser-net; and outputting the risk level of the abdominal aortic aneurysm and the probability of occurrence / progress / rupture of the abdominal aortic aneurysm in the preset year. And the development of the aortic aneurysm disease is accurately predicted.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of artificial intelligence, and particularly relates to an abdominal aortic aneurysm progression prediction method and system based on cross-modal knowledge distillation. BACKGROUND

[0002] An abdominal aortic aneurysm (AAA) is a serious life-threatening vascular disease, which refers to the local sustained expansion of the largest abdominal artery blood vessels in the human body. When the diameter of the blood vessel is increased by more than 50% compared with the normal condition, a weak cystic or fusiform bulge is formed, which is diagnosed as an aneurysm.

[0003] Doctors usually diagnose the disease when the abdominal aorta diameter exceeds 3 cm. Many patients may not feel any discomfort in daily life, but once the aneurysm ruptures, like a balloon being blown up, the patient will suddenly have severe abdominal pain, a sharp drop in blood pressure and massive bleeding. This situation is very dangerous and can endanger life in a short time.

[0004] The annual growth rate of untreated abdominal aortic aneurysm is about 2-8 mm, and the risk of rupture is significantly increased when the diameter is >5 cm, with a mortality rate of 70%-90%. The essence is that the smooth muscle and elastic fibers in the middle layer of the arterial wall are destroyed, resulting in a decrease in structural strength due to replacement by collagen.

[0005] Currently, there is no quantifiable method to effectively predict the development of abdominal aortic aneurysm to prevent sudden rupture and life-threatening consequences. SUMMARY

[0006] To solve the above problems, the purpose of the present application is to provide an abdominal aortic aneurysm progression prediction method and system based on cross-modal knowledge distillation, which can accurately predict the development of aortic aneurysm disease so as to intervene in time and prevent sudden rupture and endanger life.

[0007] The technical solution provided by the present application is as follows: an abdominal aortic aneurysm progression prediction method based on cross-modal knowledge distillation, comprising the following steps: Collecting blood of a target object; Detecting the concentration of blood protein fingerprint markers of the target object, wherein the blood protein fingerprint markers include matrix metalloproteinase-12, calcitonin-related polypeptide-alpha, urine regulatory factor, nerve injury-induced protein-1, glycosylated hemoglobin A1c, and fibroblast growth factor-9. calling the Student-net of the cross-modal knowledge distillation based on the blood protein fingerprint marker concentration to perform abdominal aortic aneurysm risk calculation; wherein the input of the Teacher-net of the cross-modal knowledge distillation comprises CTA image features and pathological section features, and the Student-net outputs a category probability vector and the category probability vector output by the Teacher-net aligns; outputting an abdominal aortic aneurysm risk level and a probability of three types of events of abdominal aortic aneurysm occurrence / progression / breakage in a preset year.

[0008] Preferably, before training the Teacher-net of the cross-modal knowledge distillation, the CTA image features, the pathological section features, the blood protein fingerprint marker concentration, the public database information and the text mining AI score need to be preprocessed: For the CTA image features, 3D-ResNet-18 migration learning realized by PyTorch is adopted, and 256 × 1 full connection is output, and SE-Attention is compressed to 128 dimensions; For the pathological section features, ViT-Tiny is fine-tuned for 20 epochs, and 64 dimensions are taken after Global Average Pooling, wherein the pathological section is processed into an image with a size of 224 × 224, and the image is divided into 16 × 16 patches by ViT-Tiny; For the blood protein fingerprint marker concentration, the public database information and the text mining AI score, Z-standardization and missing value filling are respectively performed.

[0009] Preferably, before training the Teacher-net of the cross-modal knowledge distillation, a feature matrix needs to be constructed: constructing a feature matrix based on the CTA image features, the pathological section features, the blood protein fingerprint marker concentration, the public database information and the text mining AI score wherein is the sample size, x 1~6D represents matrix metalloproteinase-12, calcitonin-related polypeptide-alpha, uromodulin, nerve injury-induced protein-1, glycosylated hemoglobin A1c, fibroblast growth factor-9 concentration data, D represents CTA image aortic diameter level, I represents CTA image convolution features, H represents abdominal aortic aneurysm risk classification score, P represents pathological section feature vector, at least including inflammatory cell infiltration, elastic fiber rupture, smooth muscle cell apoptosis, and microscopic structure information, O represents mining features output by the Transformer text encoder of the public literature; the feature matrix is used to input the public database information and the text mining AI score through a fusion layer together with the CTA image features and / or the pathological section features into a teacher model, to provide supplementary knowledge of the mechanism layer and the baseline layer, thereby enhancing the knowledge transfer effect of the distillation process.

[0010] Preferably, the process of training the Teacher-net and the Student-net of the cross-modal knowledge distillation includes: inputting the CTA image into a 3D-CNN+ self-attention model to output a rupture risk probability ; inputting the pathological section into a ViT-Tiny to output a histopathological risk probability ; calling a DeepSeek-Reasoner API based on the public database information and the text mining AI score to output a clinical risk probability and a public database risk probability ; the fusion layer adopts a Transformer output category probability vector ; aligning the category probability vector output by the Student-net with the category probability vector output by the Teacher-net, and using a distillation loss: wherein σ is a softmax function, T is a distillation temperature, α is a loss weight coefficient, CE is a cross-entropy loss, KL represents a KL divergence, y true is a real label; the Teacher-net receives a fusion risk vector output by the CTA and the pathological features and a multi-layer attention weight wherein is the attention distribution of the lth layer, the element takes a value of 0-1, indicating the importance allocation of different feature regions, satisfying row normalization, represents an abdominal aortic aneurysm occurrence probability, represents an abdominal aortic aneurysm progression probability, represents an abdominal aortic aneurysm rupture probability.

[0011] Preferably, the process of training the Teacher-net and the Student-net for the cross-modal knowledge distillation further comprises: applying a target temperature to obtain soft labels , and calculating divergence with the Student-net output . Measuring the L2 distance of the attention of the Student-net and the attention of the Teacher-net , focusing on key high-risk patterns has achieved attention transfer.

[0012] Preferably, the attention transfer of the cross-modal knowledge distillation further comprises: extracting key attention maps from the multi-layer attention weights of the Teacher-net introducing a corresponding attention module in the Student-net, calculating its attention output using L2 distance as a loss function to quantify the difference in attention distribution between the two add this loss to the total distillation loss to prompt the Student-net to adjust parameters in training to minimize the distance, thereby transferring the attention of the Teacher-net to key features.

[0013] Preferably, the process of training the Teacher-net and the Student-net for the cross-modal knowledge distillation further comprises: applying a temperature T=4 to obtain soft labels , and calculating divergence with the Student-net output . At the same time, use the real label y to calculate the cross entropy CE to ensure the classification accuracy Freeze the Teacher-net, and only update the Student-net parameters through back propagation , until the verification set ΔAUC<0.01 converges, where ΔAUC represents the change in the area under the ROC curve, used to evaluate the performance improvement of the binary classification model.

[0014] Based on the same idea, the present application also provides an abdominal aortic aneurysm progression prediction system based on cross-modal knowledge distillation, comprising: a collection module for collecting blood of a target object a detection module for detecting the concentration of blood protein fingerprint markers in the blood of the target object, the blood protein fingerprint markers including matrix metalloproteinase-12, calcitonin-related polypeptide-alpha, uromodulin, nerve injury-induced protein-1, glycosylated hemoglobin A1c, and fibroblast growth factor-9. a risk calculation module, which invokes the Student-net of the cross-modal knowledge distillation to calculate the risk of abdominal aortic aneurysm based on the blood protein fingerprint marker concentration; wherein the input of the Teacher-net of the cross-modal knowledge distillation comprises CTA image features and pathological section features, and the Student-net outputs a category probability vector a category probability vector output by the Teacher-net alignment; an output module, which outputs the risk level of abdominal aortic aneurysm and the probabilities of three types of events of abdominal aortic aneurysm occurrence / progression / breaking in a preset year.

[0015] Based on the same concept, the present application further provides an electronic device, comprising: a memory, which is used to store a processing program; a processor, which, when executing the processing program, realizes the abdominal aortic aneurysm progression prediction method based on cross-modal knowledge distillation according to any one of the above.

[0016] Based on the same concept, the present application further provides a readable storage medium, which stores a processing program, and the processing program, when executed by a processor, realizes the abdominal aortic aneurysm progression prediction method based on cross-modal knowledge distillation according to any one of the above.

[0017] The present application has the following advantages and positive effects compared with the prior art due to the adoption of the above technical solutions: The present scheme only needs blood samples, completely avoids the limitations of imaging examination, realizes truly non-invasive, low-risk, low-cost, and strong repeatability dynamic monitoring, greatly improves patient compliance and screening / monitoring frequency. Cross-modal knowledge distillation realizes "small model, great wisdom", Teacher-Net (teacher network): input contains rich CTA image features (anatomical structure, tumor size and shape, thrombus, wall thickness, etc.) and pathological section features (structural damage at the cellular level, inflammatory infiltration, calcification, etc.). These modal data have a huge amount of information, and can build a very powerful prediction model, but its acquisition is highly dependent on professional equipment, invasive operation (biopsy) and expert interpretation, and it is extremely difficult to apply clinically. Student-Net (student network): only use easily accessible blood marker concentrations as input. Student-Net can learn and inherit the complex discrimination ability contained in Teacher-Net (based on image + pathology) only with blood data, and its prediction accuracy is much higher than that of traditional models trained only with blood data, and is infinitely close to Teacher-Net using the gold standard multi-modal data. The structure of Student-Net is simpler than that of Teacher-Net which fuses multiple modalities, has small computational complexity, and is easier to integrate into a clinical laboratory information system or a portable device to realize rapid, real-time bedside or laboratory risk assessment. BRIEF DESCRIPTION OF DRAWINGS

[0018] The specific embodiments of the present application will be further described in detail below in combination with the drawings, in which: Figure 1 The flow chart of the abdominal aortic aneurysm progression prediction method based on cross-modal knowledge distillation of the present application; Figure 2 The clinical sample quantification of the abdominal aortic aneurysm protein fingerprint of an embodiment of the present application; Figure 3 The structure schematic diagram for CTA feature extraction; Figure 4 The encoding flow chart of pathological section ViT-Tiny; Figure 5 Cross-modal independence verification of the present application; Figure 6 The Receiver Operating Characteristic (ROC) curve diagram of the model of the present application; Figure 7 The Calibration curve diagram of the reliability of the model of the present application; Figure 8 The Net Benefit Curve (NBC) diagram of the model of the present application. DETAILED DESCRIPTION

[0019] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. The advantages and features of the present invention will become more apparent from the following description. It should be noted that the drawings are greatly simplified and use non-precise ratios, and are only used to facilitate and clearly illustrate the embodiments of the present invention.

[0020] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the various components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0021] First embodiment like Figure 1 As shown, this embodiment provides a method for predicting abdominal aortic aneurysm progression based on cross-modal knowledge distillation, comprising the following steps: Collecting blood from the target subject; Performing a blood protein fingerprint marker concentration test on the target subject's blood, wherein the blood protein fingerprint markers include matrix metalloproteinase-12, calcitonin-related polypeptide-α, uromodulin, nerve injury-inducible protein-1, glycated hemoglobin A1c, and fibroblast growth factor-9; Based on the concentration of the blood protein fingerprint marker, the Student-net of the cross-modal knowledge distillation is called to calculate the risk of abdominal aortic aneurysm; wherein the input of the Teacher-net of the cross-modal knowledge distillation includes CTA image features and pathological section features, and the category probability vector output by the Student-net Class probability vector output by Teacher-net alignment; Output the abdominal aortic aneurysm risk level and the probability of three types of events: abdominal aortic aneurysm occurrence / progression / rupture in the preset year.

[0022] Traditional abdominal aortic aneurysm monitoring and risk assessment is highly dependent on imaging examinations (such as ultrasound, CTA), this scheme only needs blood samples, completely avoids the limitations of imaging examinations, realizes truly non-invasive, low-risk, low-cost, and strong repeatability of dynamic monitoring, greatly improves patient compliance and screening / monitoring frequency. Cross-modal knowledge distillation realizes "small model, big wisdom", Teacher-Net (teacher network): input contains rich CTA image features (anatomical structure, aneurysm size and shape, thrombus, wall thickness, etc.) and pathological section features (structural damage at the cellular level, inflammation infiltration, calcification, etc.). These modal data have a huge amount of information, and can build a very powerful prediction model, but its acquisition is highly dependent on professional equipment, invasive operation (biopsy) and expert interpretation, and it is extremely difficult to apply clinically. Student-Net (student network): only use easily accessible blood marker concentrations as input. Knowledge distillation process: by aligning the class probability vectors (i.e. the probability distribution of predicting "occurrence / progression / rupture") output by Student-Net and Teacher-Net, Student-Net not only learns the mapping from blood data to labels during training, but also learns to mimic the "judgment" or "knowledge" made by Teacher-Net based on multi-modal rich information. Student-Net can learn and inherit the complex discrimination ability of Teacher-Net (based on image + pathology) based on blood data alone, and its prediction accuracy is much higher than that of traditional models trained only with blood data, and is close to that of Teacher-Net using gold standard multi-modal data. The structure of Student-Net is simpler than that of Teacher-Net which fuses multiple modalities, and the calculation amount is small, and it is easier to integrate into a clinical laboratory information system or a portable device to realize rapid, real-time bedside or laboratory risk assessment. Without the need to obtain expensive CTA and invasive pathological sections during prediction, blood testing alone can achieve a prediction effect close to the use of these advanced information. The output result is not only a simple "high risk / low risk" label, but also a stratification of the current overall risk (such as low, medium, high), which is convenient for rapid classification and management. Probability of three events in a preset year: directly quantify the specific probability of abdominal aortic aneurysm occurrence (for high-risk population), progression (significant aneurysm enlargement), and rupture within a specific time window (such as 1 year, 3 years, 5 years) in the future. This output directly meets the core needs of clinical decision-making.

[0023] The present application uses cross-modal knowledge distillation technology to "compress" and "migrate" the complex disease progression information contained in high-order images and pathological sections to a lightweight model that only needs routine blood testing (specific proteome) to run.

[0024] The training model features include: (1) Imaging group: aortic diameter, wall calcification volume, intramural hematoma signal, vascular remodeling morphology; (2) Pathology group: elastic fiber rupture ratio, smooth muscle cell apoptosis rate, and inflammatory cell infiltration degree; (3) Proteomics: 6 validated blood protein fingerprint markers; (4) Public database information (from public resources): including differential pathway enrichment score, protein interaction network node degree centrality, and reported functional mutation sites, which are standardized, missing value filled, and feature scaled. After key-value alignment based on patient ID or feature label with clinical / experimental data, they can be used for training; (5) Text mining AI score: attention score generated by literature abstract and full text through Transformer encoder, and keyword surge score.

[0025] Table 1 - Features for training the model:

[0026] The five-dimensional features (molecular layer, organ macroscopic layer, tissue microscopic layer, mechanism layer, and baseline layer) are statistically independent through orthogonal constraints to reduce redundant information and improve generalization performance.

[0027] In the model data input stage, images and pathology need to be subjected to deep feature extraction (due to high data dimension and complex structure), while the remaining dimensional data has been normalized, encoded, and missing value processed in the early stage, and can be directly used as structured variables to input the fusion layer.

[0028] Inclusion and exclusion criteria: study subjects were selected from clinical visit population, stratified according to abdominal aortic aneurysm size, including abdominal aortic diameter < 30 mm, abdominal aortic aneurysm 30-55 mm, 55-75 mm, and > 75 mm population. The study population was followed up every year. The study complies with the requirements of the Declaration of Helsinki, and written consent was obtained from each subject before baseline investigation and each follow-up investigation. Inclusion criteria: ① 65 years old or older; ② abdominal aortic diameter < 30 mm; ③ abdominal aortic aneurysm 30-55 mm; ④ abdominal aortic aneurysm 55-75 mm; ⑤ abdominal aortic aneurysm > 75 mm. Exclusion criteria: ① incomplete study-related data; ② not signed informed consent; ③ suffering from other major diseases.

[0029] Protein quantification uses Luminex 200 instrument and PeproTech commercial antibody microbead kit (LH-AAA-6Plex-2025) to detect six proteins in the same batch. Standard curve R 2 > 0.995; lower limit of detection (LOD) 0.1-0.5 ng / mL-1 .

[0030] Referring to Figure 2 , the clinical sample quantification of abdominal aortic aneurysm protein fingerprints is shown.

[0031] Preferably, the CTA image features, pathological section features, blood protein fingerprint marker concentrations, public database information, and text mining AI scores need to be preprocessed before training the Teacher-net for cross-modal knowledge distillation: For the CTA image features, 3D-ResNet-18 migration learning implemented by PyTorch is adopted, outputting 256x1 full connection, and SE-Attention is compressed to 128 dimensions; For the pathological section features, ViT-Tiny is fine-tuned for 20 epochs, and 64 dimensions are taken after Global Average Pooling, wherein the pathological section is processed into an image of 224x224 size, and the image is divided into 16x16 patches by ViT-Tiny; For the blood protein fingerprint marker concentration, public database information, and text mining AI score, Z-standardization and missing value filling are performed respectively.

[0032] Directly use the general feature extraction capability of the pre-trained model on large 3D medical image datasets (such as CT, MRI), avoid the high cost of training from scratch and the risk of overfitting with small samples. 3D convolution can effectively capture the spatial volume information of the blood vessel structure (such as aneurysm shape, thrombus distribution, and pipe wall calcification). Self-adaptive learning of the importance weight of different channel features allows the model to focus on the image area most relevant to abdominal aortic aneurysm pathology (such as weak pipe wall and inflammation area), suppresses irrelevant background noise, and compresses to 128 dimensions, which greatly reduces the feature dimension while retaining discriminative information, reduces the computational burden of subsequent fusion, and improves the model generalization ability. Use lightweight Vision Transformer (ViT-Tiny) to fine-tune on medical pathology images. The self-attention mechanism of ViT can model the global long-distance dependence relationship of the section (such as the range of inflammatory cell infiltration and the pattern of extracellular matrix destruction), overcoming the limitations of the local receptive field of CNN.

[0033] Figure 3 A CTA feature extraction structure diagram is shown, Figure 4 A pathological section ViT-Tiny encoding flowchart is shown. The features of protein fingerprints, public database features, text mining scores, and clinical risk scores are processed by Z-standardization, missing value filling, etc. in the feature engineering stage, and can be directly spliced with deep features in the fusion layer without the need for secondary convolution or Transformer encoding.

[0034] Preferably, a feature matrix needs to be constructed before training the Teacher-net for the cross-modal knowledge distillation: Constructing a feature matrix based on the CTA image features, the pathological section features, the blood protein fingerprint marker concentrations, public database information, and text mining AI scores , wherein is the sample size, , x 1~6 represents matrix metalloproteinase-12, calcitonin-related polypeptide-alpha, uromodulin, nerve injury-induced protein-1, glycosylated hemoglobin A1c, fibroblast growth factor-9 concentration data, D represents CTA image aortic diameter level, I represents CTA image convolution features, H represents abdominal aortic aneurysm risk classification score, P represents pathological section feature vector, at least containing inflammatory cell infiltration, elastic fiber rupture, smooth muscle cell apoptosis, and microscopic structure information, and O represents mining features output by the Transformer text encoder of the public literature; the feature matrix is used to input the public database information and the text mining AI scores into the teacher model through a fusion layer together with the CTA image features and / or the pathological section features, to provide complementary knowledge of the mechanism layer and the baseline layer, thereby enhancing the knowledge transfer effect of the distillation process.

[0035] By constructing a structured feature matrix as the input core of the Teacher-Net, systematic integration and knowledge complementarity of multi-source heterogeneous data are achieved, and the knowledge breadth and depth of the "teacher" model in the cross-modal knowledge distillation are significantly improved. The data islands are broken, and the traditional fragmented image, pathology, biochemistry, clinical, and literature data are unified in a structured matrix (n x k), providing a multi-dimensional disease view for the Teacher-Net. The public database information (H, etc.) and the text mining features (O) are combined with the core image / pathology features (I / P) through a special fusion layer. The text mining features contain molecular mechanisms, pathway associations, and pathophysiological logic reported in the literature (such as "MMP-12 high expression promotes elastic fiber degradation"). It helps the Teacher-Net to understand the biological causal chain between blood markers (x1~x6) and image / pathology performance (I / P), and improves the explainability and generalization ability of prediction. The clinical risk score (H) and the public database information (such as age, comorbidities) represent the clinical consensus and epidemiological evidence. It provides a priori risk benchmark for the model to correct individual bias (such as higher baseline risk for elderly patients), making the prediction more consistent with the rules of clinical practice. The Teacher-Net not only learns statistical associations, but also absorbs the inherent logic of domain knowledge, forming a "medical expert-level" judgment ability. The injected mechanism and baseline knowledge make the probability vector output by the Teacher-Net contain richer pathological logic and clinical context, providing a higher quality imitation target for the Student-Net.

[0036] Matrix covering from molecule to cell to tissue to organ to clinic full-scale information, so that Teacher-Net can learn the complex progression patterns that blood markers cannot reveal independently (such as the "local inflammation outbreak" shown by pathology indicating short-term rupture risk). Text features (O) explain the mechanism, clinical features (H) set the baseline, and images / pathology (I / P) provide direct evidence, and blood markers (x) reflect the real-time state - multi-source information confirms and complements each other, reducing single-modal noise interference. By embedding literature and clinical knowledge into image / pathology feature representation through the fusion layer, the final knowledge representation implicitly contains cross-domain associations. When Student-Net imitates the probability vector, it indirectly absorbs these embedded complementary knowledge. More comprehensive and robust feature input enables Teacher-Net to achieve prediction accuracy close to the theoretical upper limit, laying the foundation for distillation. Student-Net can reproduce the logical judgment achieved by multi-modal fusion by imitating the probability distribution of the "enhanced" Teacher, breaking through the input modal restriction.

[0037] Five-dimensional data (images, pathology, blood, clinical, literature) and mechanism layer (O) and baseline layer (H) knowledge create a "super knowledge base" for Teacher-Net. This not only greatly improves the prediction ability of the teacher model itself, but more importantly, through the distillation process, these implicit cross-domain knowledge (especially text and clinical knowledge) are efficiently transferred to Student-Net, which only relies on blood markers, ultimately achieving the revolutionary effect of "blood test for multi-modal wisdom" - this is the core value of cross-modal knowledge distillation in the medical field.

[0038] Input feature matrix X:

[0039] The integrated feature matrix is used in the training process of the teacher model to fuse CTA, pathology, protein fingerprint, public database information, and text mining AI score modal features to generate comprehensive teacher knowledge: where n is the sample size, .

[0040] In cross-modal knowledge distillation, this matrix ensures that the knowledge extracted by the teacher model from multi-modal data is distilled into the student model that only uses protein fingerprints, where public database information and text mining AI scores are input into the teacher model through the fusion layer along with CTA / pathology features, providing complementary knowledge of the mechanism layer and the baseline layer, thereby enhancing the knowledge transfer effect of the distillation process.

[0041] High-dimensional image and pathological data are extracted by deep network features (3D-ResNet, ViT-Tiny), and low-dimensional structured data is normalized, missing value filled and encoded in the feature engineering stage to avoid training instability caused by input distribution difference between different modalities. Orthogonal constraint and cross-modal independence verification (HSIC / CKA / VIF) are introduced to ensure that the five modalities still maintain low correlation after fusion, reducing the risk of overfitting. The model outputs the probability distribution of abdominal aortic aneurysm occurrence / progression / breakage. The multi-dimensional output is adjusted by distillation temperature (T=4) and probability distribution alignment, so that the student model maintains >95% consistency with the teacher model in three types of prediction.

[0042] Preferably, the process of training the Teacher-net and Student-net for the cross-modal knowledge distillation comprises: Inputting the CTA image into a 3D-CNN (3D-Convolutional Neural Networks) + self-attention model to output the risk probability of rupture ; Inputting the pathological section into ViT-Tiny to output the histopathological risk probability ; Calling DeepSeek-Reasoner API based on public database information and text mining AI score to output clinical risk probability and public database risk probability ; The fusion layer adopts Transformer to output the category probability vector ; Aligning the category probability vector output by the Student-net with the category probability vector output by the Teacher-net Using distillation loss: , where σ is the softmax function, T is the distillation temperature, α is the loss weight coefficient, CE is the cross-entropy loss, KL represents the KL divergence, y true is the true label; The Teacher-net receives the CTA and pathological feature output fusion risk vector and its multi-layer attention weight , where is the attention distribution of the lth layer, the element takes value 0-1, indicating the importance allocation of different feature regions, satisfying row normalization, represents the abdominal aortic aneurysm occurrence probability, represents the abdominal aortic aneurysm progression probability, represents the abdominal aortic aneurysm rupture probability.

[0043] In this embodiment, the focal point of the blood vessel wall local weak area (such as asymmetric bulging, intramural hematoma), capture the spatial characteristics of the risk of mechanical rupture, quantify the density of inflammatory cell infiltration, elastic fiber fracture index, etc., reflect the biological activity (microscopic driving force of progression / rupture), integrate epidemiological regularity (such as "diameter>5cm+smoking history=high rupture risk"), literature mechanism (such as "NINJ1 mediated inflammatory cascade"), output the statistical risk supported by evidence-based medicine and the risk associated with pathological mechanism. Three experts independently evaluate the risk from the aspects of physical structure, biological activity and clinical evidence to avoid single modality bias and form a stereoscopic risk assessment system. Using the self-attention mechanism of Transformer, the decision weights of different experts under different samples are automatically learned (such as: focusing on imaging experts when the risk of rupture is high, and focusing on pathological experts in the early stage of progression). The Student-net probability distribution is forced to soften the distribution of the Teacher-net (T>1 probability is smoother, implying the relationship between categories), ensuring the correctness of the basic label. The Teacher-net additionally outputs multi-layer spatio-temporal attention weights: : The attention distribution of the first layer network at time t (0-1 weight, row normalized). Student-net not only learns the probability output, but also approximates by constraining its internal attention distribution, inheriting the "thinking mode" of the teacher in analyzing the problem (such as: which marker combination triggers a high-risk warning). t can be associated with follow-up time points, allowing the model to understand the time dependence of risk factors (such as a sudden increase in HbA1c indicating progression within 3 months).

[0044] In view of the characteristics of the same patient image with lesion segment and healthy segment, the convolution kernel size is introduced to adapt to the vascular cross-sectional scale in the CTA feature extraction stage, thereby enhancing the sensitivity to local morphological differences. The pathology modality optimizes the Patch Embedding and attention head configuration based on ViT-Tiny to adapt to the color channels and high resolution of pathological sections, improving the recognition accuracy of inflammatory cell aggregation areas and elastic fiber fracture areas.

[0045] The Teacher-net, as a "super expert", integrates images, pathology, literature, and clinical data to output a risk vector with biological mechanisms and clinical operability. The attention weight Explicitly annotated decision basis to meet the explainability requirements of medical AI. Student-net realizes "mind cloning", and can reproduce the decision logic (through attention alignment) and conclusion (through probability alignment) of Teacher based on blood markers. Breakthrough "black box" restrictions: doctors can understand why blood indicators predict risks through the attention weights of Student-net. Through three core technologies of multi-expert collaborative decision-making, attention-enhanced distillation, and spatio-temporal modeling, Student-net becomes a "mind clone" of Teacher-net - it can output accurate predictions comparable to multi-modal fusion models using only blood tests, and has clinically explainable decision logic.

[0046] A "cross-modal knowledge distillation" two-stage framework is adopted. Since blood protein detection is the most accessible in practical applications, while CTA and pathological information is costly to obtain. By having the student model with only protein input learn from the teacher model that uses CTA + pathological features, the diagnostic ability of images / pathology is preserved, and lightweight inference is achieved at the deployment end. For the diverse data types in this study, we introduce both probability distribution alignment and attention weight transfer in the knowledge distillation loss. By applying a temperature parameter to the fusion output of the teacher model to generate soft labels, and calculating the KL divergence with the student model output, we achieve probability distribution matching and knowledge transfer, so that the student model learns the consistency of the teacher model's multi-class probability distribution, ensuring accuracy in multi-dimensional outputs such as occurrence, progression, and rupture) and attention weight transfer, and adjusting the distillation temperature to balance the multi-class probability output.

[0047] It should be noted that Student-net is deployed at the application end, with input: only input (MMP12, CALCA, UMOD, NINJ1, GHBA1C, FGF9 six protein molecules in blood standardization expression quantity). The structure includes: the first layer is Conv1D(kernel=3, filters=32) → BatchNorm → ReLU → MaxPool1D; the second layer is Conv1D(kernel=3, filters=64) → BatchNorm → ReLU → GlobalAvgPool; the third layer is FC-1(128 units, ReLU, Dropout=0.2); Softmax(3). Output: (abdominal aortic aneurysm occurrence / progression / breakage three-class probability). Student-net fully inherits the high-dimensional knowledge of images / pathology with only 6-dimensional input, achieving lightweight deployment with inference delay ≤2ms and model volume 0.5MB.

[0048] Preferably, the process of training the Teacher-net and the Student-net for the cross-modal knowledge distillation further comprises: Applying target temperature to obtain soft labels, and calculating divergence with the output of the Student-net; Measuring the L2 distance of the attention of the Student-net and the attention of the Teacher-net , focusing on key high-risk patterns has achieved attention transfer.

[0049] Through the two core technologies of soft label knowledge rectification and attention focus transfer, the depth and fidelity of knowledge transfer in cross-modal distillation are significantly improved. The target temperature T>1 is applied to the original output probability distribution σ(z_t) of the Teacher-net to generate soft labels. The output distribution σ(z_s) of the Student-net is forced to approximate the soft label distribution (by minimizing the KL divergence). Hard labels (such as [0, 0, 1]) only indicate the final class, while soft labels (such as [0.1, 0.2, 0.7]) reveal the relative confidence of the teacher model for each class. Soft labels provide a smooth transition between classes, avoiding overfitting of the Student to noisy data and improving robustness in real clinical data. The high-risk feature patterns (such as "focal elastin fiber rupture + neutrophil cluster") identified by the Teacher through images / pathology are encoded as attention hotspots. Although the Student only sees blood data, its attention mechanism will automatically lock onto the corresponding high-risk biomarker combination. The Student shows the same pathological logic as the Teacher when predicting - activating high-risk warnings when specific marker combinations appear. The attention heat map of the Teacher can be mapped back to the image / pathology area, and the attention weight of the Student can be directly associated with the blood markers, making the prediction of both models visualized and meeting the needs of clinically reliable AI.

[0050] Preferably, the attention transfer of the cross-modal knowledge distillation further comprises: Extracting key attention maps from the multi-layer attention weights of the Teacher-net; Introducing corresponding attention modules in the Student-net to calculate its attention output; Using L2 distance as the loss function to quantify the difference in attention distribution between the two; Add this loss to the total distillation loss to prompt the Student-net to adjust its parameters to minimize the distance during training, thereby transferring the attention of the Teacher-net to key features.

[0051] ​Through cross-modal hard alignment of attention maps, pixel-level transfer of teacher model diagnostic thinking is achieved, making breakthrough progress in explainability reinforcement and high-risk feature transfer. The teacher's locked anatomical / pathological high-risk area based on multi-modal data is encoded as a transferable heat map. Shallow attention focuses on basic features such as texture, while deep attention focuses on semantic features such as rupture precursors, achieving full-dimensional knowledge transfer.

[0052] Preferably, the process of training the Teacher-net and Student-net of the cross-modal knowledge distillation further comprises: For Apply temperature T=4 to obtain soft labels , the Student-net output Calculate divergence; At the same time, calculate the cross-entropy CE using the real label y to ensure classification accuracy; Freeze the Teacher-net, and only update the Student-net parameters through backpropagation , until the verification set ΔAUC<0.01 converges, where ΔAUC represents the change in the area under the ROC curve, used to evaluate the performance improvement of the binary classification model.

[0053] The selection of distillation temperature T is based on experience adjustment, usually T>1 to soften the probability distribution, making it easier for the student model to learn the relative probability relationship of the teacher model rather than hard labels; in experiments, different T values such as 1, 2, 4, 8 are tested through grid search or cross-validation, and T=4 is selected as the best value because it performs best in balancing softening degree and classification sharpness, avoiding distillation failure due to too small T or too uniform probability due to too large T.

[0054] Referring to Figure 5 , cross-modal independence verification includes: HSIC (Hilbert-Schmidt Independence Criterion): Except for the same mode, the cross-modal HSIC value is close to 0.00~0.03, indicating that the modal features are basically independent in the feature space, and the information complementarity is strong when fused.

[0055] CKA (Centered Kernel Alignment): Except for the diagonal line and a few combinations, the CKA value is generally 0.00~0.03, consistent with the results of HSIC, verifying the decoupling of multi-modal features in the representation space.

[0056] VIF (Variance Inflation Factor): The VIF value of all variables is <3, much lower than the commonly used threshold of 5, indicating that there is no multicollinearity, and each feature is relatively independent in explaining the result in prediction.

[0057] The three types of indexes consistently prove from three aspects of nonlinear independence, feature representation similarity and multicollinearity that the five modalities used are independent and complementary in the feature level, which helps to improve the generalization ability and stability of the fusion model and reduce the risk of overfitting.

[0058] Referring to Figure 6 , Figure 7 , Figure 8 , respectively, are the ROC curve diagram, reliability calibration diagram and net benefit curve of the model of the application. Discrimination: AUC 0.93, AUPRC 0.88, external 0.91 / 0.85 (95%CI). Calibration: Brier 0.072, CalibrationSlope 0.98. Clinical: DCA in the threshold 2-20% interval, the net benefit increases by 6-15%. Figure 6 The area under the ROC curve (AUC) in the application is close to 1.0, indicating that the model has excellent discrimination ability. The Youden index, at multiple thresholds, the true positive rate (sensitivity) and the true negative rate (specificity) are excellent, further verifying the ability of the model to distinguish positive and negative samples. The model performs outstandingly in identifying high-risk and low-risk individuals. Figure 7 The Brier coefficient value in the application is 0.07, which is much lower than the critical value of 0.25, indicating that the predicted probability is highly consistent with the actual observation. The calibration slope value is 0.98, close to the ideal 1.0, indicating that the model does not have the problem of systematic overestimation or underestimation of risk. The probability value predicted by the model is very reliable and can be directly used for clinical decision support. Figure 8 The net benefit curve (clinical benefit) threshold range is in the risk threshold range of 0%-10%, the net benefit (NetBenefit) of the model is always higher than the "full intervention" and "no intervention" strategies, and is significantly better than the traditional ultrasound method. The maximum net benefit (MAXNB) is 0.75, while the traditional ultrasound is only 0.10, indicating that under the same false positive rate, the model can reduce unnecessary interventions (such as biopsy) by about 85%. The model can significantly reduce over-treatment and improve decision-making efficiency in clinical applications.

[0059] Compared with the traditional CNN model, the input amount of information is reduced, and compared with the single Transformer model, the interpretability and robustness are improved, and the deployment cost is reduced. On the external validation set, the AUC is improved to 0.91 (about 6-8% higher than the single protein model), and the inference delay is ≤2ms, the model volume is only 0.5MB, suitable for bedside deployment.

[0060] Second embodiment Based on the same concept, the application further provides an abdominal aortic aneurysm progression prediction system based on cross-modal knowledge distillation, comprising: The acquisition module is configured to acquire blood of a target object. a detection module, detecting blood protein fingerprint marker concentration of blood of the target object, the blood protein fingerprint marker including matrix metalloproteinase-12, calcitonin-related polypeptide-alpha, uromodulin, nerve injury-induced protein-1, glycosylated hemoglobin A1c, fibroblast growth factor-9; a risk calculation module, invoking the Student-net of the cross-modal knowledge distillation to calculate abdominal aortic aneurysm risk based on the blood protein fingerprint marker concentration; wherein the input of the Teacher-net of the cross-modal knowledge distillation includes CTA image features and pathological section features, and the class probability vector output by the Student-net is aligned with the class probability vector output by the Teacher-net . an output module, outputting abdominal aortic aneurysm risk level and abdominal aortic aneurysm occurrence / progression / breakdown event probability in a preset year.

[0061] Based on the same concept, the present application also provides an electronic device, comprising: a memory, the memory being used to store a processing program; a processor, the processor realizing the abdominal aortic aneurysm progression prediction method based on cross-modal knowledge distillation of any one of the above when executing the processing program.

[0062] Based on the same concept, the present application also provides a readable storage medium, the readable storage medium storing a processing program, the processing program realizing the abdominal aortic aneurysm progression prediction method based on cross-modal knowledge distillation of any one of the above when executed by a processor.

[0063] The abdominal aortic aneurysm progression prediction method based on cross-modal knowledge distillation can be stored in a computer-readable storage medium if it is realized in the form of program instructions and sold or used as an independent product. Based on such understanding, the technical solutions of the embodiments can be embodied in the form of software, and the computer software is stored in a storage medium, including a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in various embodiments of the present disclosure. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (Read-only memory, ROM), a random access memory (Random access memory, RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0064] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific execution of the system and the device described above can refer to the corresponding process in the foregoing method embodiment.

[0065] The embodiments of the present application are described in detail above with reference to the accompanying drawings, but the present application is not limited to the above-described embodiments. Even if various changes are made to the present application, if the changes belong to the scope of the claims of the present application and the equivalent technology thereof, they still fall within the protection scope of the present application.

Claims

1. A method for predicting abdominal aortic aneurysm progression based on cross-modal knowledge distillation, characterized by: The following steps are involved: Collecting blood from the target subject; Performing a blood protein fingerprint marker concentration test on the target subject's blood, wherein the blood protein fingerprint markers include matrix metalloproteinase-12, calcitonin-related polypeptide-α, uromodulin, nerve injury-inducible protein-1, glycated hemoglobin A1c, and fibroblast growth factor-9; Based on the concentration of the blood protein fingerprint marker, the Student-net of the cross-modal knowledge distillation is called to calculate the risk of abdominal aortic aneurysm; wherein the input of the Teacher-net of the cross-modal knowledge distillation includes CTA image features and pathological section features, and the category probability vector output by the Student-net Class probability vector output by Teacher-net alignment; Output the abdominal aortic aneurysm risk level and the probability of three types of events: abdominal aortic aneurysm occurrence / progression / rupture in the preset year; Among them, before training the Teacher-net of the cross-modal knowledge distillation, it is necessary to construct a feature matrix: construct a feature matrix based on the CTA image features, the pathological section features, the blood protein fingerprint marker concentration, public database information and text mining AI score ,in is the sample size, , x 1~6 represents the concentration data of matrix metalloproteinase-12, calcitonin-related polypeptide-α, uromodulin, nerve injury-induced protein-1, glycated hemoglobin A1c, and fibroblast growth factor-9, D represents the aortic diameter grade of CTA images, I represents the convolutional features of CTA images, H represents the abdominal aortic aneurysm risk grading score, P represents the pathological section feature vector, which at least contains microstructural information including inflammatory cell infiltration, elastic fiber rupture, and smooth muscle cell apoptosis, and O represents the mined features output by the Transformer text encoder of the public literature; the feature matrix is ​​used to input the public database information and text mining AI scores into the teacher model together with the CTA image features and / or pathological section features through the fusion layer, providing supplementary knowledge for the mechanism layer and the baseline layer, thereby enhancing the knowledge transfer effect of the distillation process.

2. The abdominal aortic aneurysm progression prediction method based on cross-modal knowledge distillation according to claim 1, characterized in that: Before training the Teacher-net for cross-modal knowledge distillation, the CTA image features, the pathological section features, the blood protein fingerprint marker concentration, public database information, and text mining AI scores need to be preprocessed: For the CTA image features, 3D-ResNet-18 transfer learning implemented in PyTorch was used, with a 256 × 1 fully connected output and SE-Attention compression to 128 dimensions; ViT-Tiny was used to fine-tune the pathological slice features for 20 epochs, and global average pooling was performed to obtain a 64-dimensional image. The pathological slices were processed into 224×224 images, and ViT-Tiny segmented the images into 16×16 tiles. Z-standardization and missing value filling were performed on the blood protein fingerprint marker concentration, public database information and text mining AI score.

3. The abdominal aortic aneurysm progression prediction method based on cross-modal knowledge distillation according to claim 1, characterized in that: The process of training the Teacher-net and Student-net of the cross-modal knowledge distillation includes: Input CTA images into the 3D-CNN+self-attention model and output the rupture risk probability ; Input the pathological slice into ViT-Tiny and output the tissue pathology risk probability ; Call DeepSeek-Reasoner API based on public database information and text mining AI scores to output clinical risk probability and public database risk probability ; The fusion layer uses Transformer to output category probability vector ; The category probability vector output by Student-net Class probability vector output by Teacher-net Alignment, using distillation loss: , where σ is the softmax function, T is the distillation temperature, α is the loss weight coefficient, CE is the cross entropy loss, KL represents the KL divergence, and y true is the true label; Teacher-net receives CTA and pathological features and outputs a fusion risk vector and its multi-layer attention weights ,in is the attention distribution of the lth layer, and the element value is 0-1, which indicates the importance distribution of different feature areas and satisfies row normalization. The probability of abdominal aortic aneurysm, represents the probability of progression of abdominal aortic aneurysm, Represents the probability of abdominal aortic aneurysm rupture.

4. The abdominal aortic aneurysm progression prediction method based on cross-modal knowledge distillation according to claim 3, characterized in that: The process of training the Teacher-net and Student-net of the cross-modal knowledge distillation further includes: right Apply the target temperature to obtain the soft label and calculate the divergence with the Student-net output; Measuring Student-net Attention With Teacher-net Attention L2 distance, focusing on key high-risk patterns has achieved attention transfer.

5. The abdominal aortic aneurysm progression prediction method based on cross-modal knowledge distillation according to claim 4, characterized in that: The attention transfer of the cross-modal knowledge distillation further includes: Extract key attention maps from the multi-layer attention weights of Teacher-net; Introduce the corresponding attention module in Student-net and calculate its attention output; Use L2 distance as the loss function to quantify the difference in attention distribution between the two; Adding this loss to the total distillation loss encourages Student-net to adjust parameters during training to minimize this distance, thereby migrating Teacher-net's attention to key features.

6. The abdominal aortic aneurysm progression prediction method based on cross-modal knowledge distillation according to claim 4, characterized in that: The process of training the Teacher-net and Student-net of the cross-modal knowledge distillation further includes: right Apply temperature T=4 to obtain soft label , and Student-net output Calculate divergence; At the same time, the cross entropy CE is calculated using the true label y to ensure classification accuracy; Freeze Teacher-net, and backpropagation only updates Student-net parameters , until the validation set ΔAUC < 0.01 converges, where ΔAUC represents the change in the area under the ROC curve, which is used to evaluate the performance improvement of the binary classification model.

7. A system for predicting abdominal aortic aneurysm progression based on cross-modal knowledge distillation, characterized by: include: a collection module, for collecting blood from a target subject; a detection module for detecting the concentration of blood protein fingerprint markers on the blood of the target subject, wherein the blood protein fingerprint markers include matrix metalloproteinase-12, calcitonin-related polypeptide-α, uromodulin, nerve injury-inducible protein-1, glycated hemoglobin A1c, and fibroblast growth factor-9; The risk calculation module calls the Student-net of the cross-modal knowledge distillation to calculate the risk of abdominal aortic aneurysm based on the concentration of the blood protein fingerprint marker; wherein the input of the Teacher-net of the cross-modal knowledge distillation includes CTA image features and pathological section features, and the category probability vector output by the Student-net Class probability vector output by Teacher-net alignment; Output module, outputs the abdominal aortic aneurysm risk level and the probability of three types of events: abdominal aortic aneurysm occurrence / progression / rupture in the preset year; Among them, before training the Teacher-net of the cross-modal knowledge distillation, it is necessary to construct a feature matrix: construct a feature matrix based on the CTA image features, the pathological section features, the blood protein fingerprint marker concentration, public database information and text mining AI score ,in is the sample size, , x 1~6 represents the concentration data of matrix metalloproteinase-12, calcitonin-related polypeptide-α, uromodulin, nerve injury-induced protein-1, glycated hemoglobin A1c, and fibroblast growth factor-9, D represents the aortic diameter grade of CTA images, I represents the convolutional features of CTA images, H represents the abdominal aortic aneurysm risk grading score, P represents the pathological section feature vector, which at least contains microstructural information including inflammatory cell infiltration, elastic fiber rupture, and smooth muscle cell apoptosis, and O represents the mined features output by the Transformer text encoder of the public literature; the feature matrix is ​​used to input the public database information and text mining AI scores into the teacher model together with the CTA image features and / or pathological section features through the fusion layer, providing supplementary knowledge for the mechanism layer and the baseline layer, thereby enhancing the knowledge transfer effect of the distillation process.

8. An electronic device, characterized in that: include: a memory for storing a processing program; A processor, wherein when executing the processing program, the processor implements the abdominal aortic aneurysm progression prediction method based on cross-modal knowledge distillation as described in any one of claims 1 to 6.

9. A readable storage medium, characterized in that: The readable storage medium stores a processing program, and when the processing program is executed by the processor, it implements the abdominal aortic aneurysm progression prediction method based on cross-modal knowledge distillation as described in any one of claims 1 to 6.

Images