Prediction method and system for onset risk of liver cancer

Abstract

The invention provides a liver cancer onset risk prediction method and system, and belongs to the technical field of liver cancer onset risk prediction. The method comprises the following steps: carrying out longitudinal multi-omics data acquisition on a target individual to generate an individualized multi-omics time sequence data set; constructing an individualized liver cancer evolution graph based on the data set; according to the individualized liver cancer evolution graph, quantifying a liver microenvironment pressure field in combination with a medical imaging technology, and performing space-time coupling on a pressure field quantification result and a multi-omics time sequence data set to generate a liver cancer dynamic risk evolution trajectory; through longitudinal multi-omics data acquisition and individualized liver cancer evolution graph construction, a liver microenvironment pressure field is quantified in combination with a medical imaging technology, dynamic and accurate prediction of the liver cancer occurrence risk is realized, and the prediction accuracy is remarkably improved.

Description

TECHNICAL FIELD

[0001] The application provides a prediction method and system for liver cancer incidence risk, and belongs to the technical field of liver cancer incidence risk prediction. BACKGROUND

[0002] Liver cancer is a malignant tumor with high incidence and poor prognosis worldwide. Its early symptoms are occult, and most patients are in the middle and advanced stages when diagnosed, missing the best intervention opportunity. The current liver cancer incidence risk prediction method has obvious limitations: traditional prediction mostly relies on a single indicator (such as AFP detection, ultrasonic examination) or static multi-omics analysis, lacking the capture of the dynamic evolution law of molecular events; the correlation between liver microenvironment stress field (such as liver fibrosis degree, hemodynamic parameter) and molecular events is not effectively fused, and the risk assessment is one-sided; the warning threshold and intervention scheme are mostly general standards, lacking individualized adaptability, which easily leads to missed diagnosis, misdiagnosis or over-treatment. Therefore, there is an urgent need for a prediction method that can dynamically integrate multi-omics time series data and microenvironment characteristics and accurately identify risk key nodes to improve early warning efficiency and provide more targeted risk assessment and intervention basis for clinical practice. SUMMARY

[0003] The application provides a prediction method and system for liver cancer incidence risk to solve the problems mentioned in the background.

[0004] The prediction method for liver cancer incidence risk provided by the application comprises: S1: longitudinal multi-omics data collection is performed on a target individual to generate an individualized multi-omics time series dataset; and an individualized liver cancer evolution map is constructed based on the dataset; S2: according to the individualized liver cancer evolution map, the liver microenvironment stress field is quantified by combining medical imaging technology, the stress field quantification result is coupled with the multi-omics time series dataset in space-time, and a liver cancer dynamic risk evolution track is generated; S3: based on the liver cancer dynamic risk evolution track, a risk acceleration period is identified by using a machine learning algorithm; feature extraction is performed on the risk acceleration period to generate a risk acceleration period feature dataset; S4: the risk acceleration period feature dataset is compared and analyzed with known liver cancer occurrence cases, the spatial distribution characteristics of the liver microenvironment stress field are combined, potential hot spot areas and high-risk subgroups of liver cancer occurrence are identified, the individualized liver cancer evolution map is dynamically updated based on the hot spot area and high-risk subgroup information, and a refined liver cancer risk warning map is generated; S5: according to the refined liver cancer risk warning map, an individualized liver cancer occurrence risk index is calculated; different levels of warning thresholds are set based on the risk index, when the risk index exceeds the corresponding threshold, liver cancer early warning data is generated, and personalized prevention intervention suggestions are provided.

[0005] The prediction system for liver cancer incidence risk provided by the present application comprises: one or more processors; a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method of any one of the above.

[0006] The present application has the following advantages: By longitudinal multi-omics data collection and individualized liver cancer evolution map construction, combined with medical imaging technology to quantify liver microenvironment stress field, dynamic and accurate prediction of liver cancer incidence risk is achieved, which significantly improves the prediction accuracy; machine learning algorithm is used to identify risk acceleration period and extract feature data set, effectively capturing key signals before liver cancer occurs, reducing the possibility of misdiagnosis and missed diagnosis; by identifying potential hotspots and high-risk subgroups of liver cancer occurrence, dynamically updating individualized liver cancer evolution map, generating refined liver cancer risk warning map, early warning of liver cancer is realized, which saves valuable treatment time for patients; based on individualized liver cancer risk index, different levels of warning thresholds are set to avoid excessive medical treatment and unnecessary intervention, improve the utilization efficiency of medical resources; according to the warning data, individualized prevention intervention plan is formulated, and the clinical intervention effect database is optimized to provide more accurate and effective prevention measures for patients, which helps to reduce the risk of liver cancer. BRIEF DESCRIPTION OF DRAWINGS

[0007] Figure 1 The flowchart of an embodiment of the method of the present application is shown. DETAILED DESCRIPTION

[0008] The preferred embodiments of the present application are described below in conjunction with the accompanying drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application, and are not used to limit the present application.

[0009] An embodiment of the present application, as shown in Figure 1 the prediction method for liver cancer incidence risk comprises the following steps: S1: longitudinal multi-omics data collection is performed on the target individual, the data includes genomic, transcriptomic, proteomic and metabolomic data, and individualized multi-omics time series data set is generated; based on the data set, individualized liver cancer evolution map is constructed, and the time sequence and interaction network of key molecular events in the liver cancer occurrence process are determined; S2: According to the individualized liver cancer evolution map, combine medical imaging technology (such as MRI, CT or ultrasonic elastography) to quantify the liver microenvironment stress field, including but not limited to the degree of liver fibrosis, hemodynamic parameters and extracellular matrix stiffness; the stress field quantization result is coupled with the multi-omics time series data set to generate a liver cancer dynamic risk evolution trajectory, which intuitively reflects the trend of change of liver microenvironment stress with time and the relevance to molecular events; S3: Based on the dynamic risk evolution trajectory of liver cancer, use machine learning algorithms (such as LSTM network in deep learning or time series analysis model) to identify the risk acceleration period, that is, the time window of rapid rise of liver cancer risk; feature extraction is performed on the risk acceleration period, including but not limited to risk rising rate, key molecular event activation threshold and liver microenvironment stress field mutation point, to generate a risk acceleration period feature data set; S4: Compare and analyze the risk acceleration period feature data set with clinically known liver cancer cases, combine the spatial distribution characteristics of the liver microenvironment stress field, identify the potential hot spot area and high-risk subpopulation of liver cancer occurrence; based on the hot spot area and high-risk subpopulation information, dynamically update the individualized liver cancer evolution map to generate a refined liver cancer risk warning map, which is used to intuitively show the high-risk area, time window and possible molecular mechanism of liver cancer occurrence; S5: According to the refined liver cancer risk warning map, calculate the individualized liver cancer risk index, which considers the characteristics of the risk acceleration period, the quantization results of the liver microenvironment stress field and the clinical relevant information; based on the risk index, set different levels of warning threshold, when the risk index exceeds the corresponding threshold, generate liver cancer early warning data and provide personalized prevention intervention suggestions, including lifestyle adjustment, regular screening plan and targeted preventive treatment.

[0010] The working principle and effect of the above technical solution are: Through the spatio-temporal coupling of longitudinal multi-omics data and medical imaging, combined with machine learning to identify the risk acceleration period, the key time window of liver cancer occurrence can be captured in advance, avoiding the lag of traditional static evaluation, making the risk judgment more in line with individual disease changes, and improving the accuracy of individualized liver cancer risk prediction; Relying on the comparison of risk acceleration period features and clinical cases, combined with the positioning of potential hot spot areas by liver microenvironment stress field, it reduces the misjudgment caused by single data evaluation, especially can accurately distinguish high-risk and low-risk populations, avoids over-treatment of low-risk individuals or misses high-risk individuals, and reduces the probability of liver cancer misdiagnosis and missed diagnosis; The individualized liver cancer evolution map can be dynamically updated according to hot area information and high-risk subpopulation information, intuitively display a risk area, a time window and a molecular mechanism, provide clear decision basis for doctors, no longer rely on static reports, improve flexibility of clinical judgment, and enhance dynamic nature and clinical practicability of risk assessment. The risk index is used to formulate a targeted screening plan and an intervention scheme, so that all people do not need to adopt high-frequency screening or universal intervention, the screening cost of healthy people is reduced, resource consumption and physical burden caused by blind drug use are reduced, and unnecessary waste of medical resources is reduced. Through early warning data and individualized intervention, the disease can be intervened in time in a risk rising stage, and the progression of the disease to liver cancer is delayed or even blocked, the survival quality of patients is improved, and the probability of liver cancer developing to a late stage is reduced.

[0011] In another embodiment of the application, step S1 further comprises: S11: using multi-modal molecular detection technology (such as whole exome sequencing, targeted transcriptome sequencing, protein chip and non-targeted metabolomics detection), longitudinal data of a target individual for 1-3 years are collected, including collecting original multi-omics data of a genome, a transcriptome, a proteome and a metabolome, to generate an individual original multi-omics data set; and the individual original multi-omics data set is preprocessed to generate a standardized multi-omics time series data set; S12: based on the generated standardized multi-omics time series data set, a time correlation analysis algorithm (such as a dynamic time warping algorithm) is used to mine time correlation relationships among different omics data, to determine core molecular indicators related to liver cancer occurrence, and to generate a multi-omics core correlation data set; S13: according to the generated multi-omics core correlation data set, an individualized liver cancer evolution framework including a molecular event occurrence time axis is constructed in combination with a tumor evolution theory, an occurrence order of key molecular events is preliminarily defined, and a preliminary liver cancer evolution framework is generated; S14: through the generated preliminary liver cancer evolution framework, a network construction algorithm (such as a weighted gene co-expression network analysis) is used to draw an interaction network among key molecular events, to clarify a regulation relationship of the molecular events, and to generate an individualized liver cancer evolution map; S15: based on the generated individualized liver cancer evolution map, a verified and corrected individualized liver cancer evolution map is generated through clinical pathological data (such as liver biopsy results).

[0012] The working principle and effects of the above technical solutions are as follows: By collecting full-dimensional data covering genome to metabolome longitudinally for 1-3 years, the one-time collection bias is avoided; meanwhile, the data redundancy and interference are eliminated through pretreatment, the error caused by poor data quality in subsequent analysis is reduced, a reliable data foundation is laid for subsequent research, and the integrity and standardization of multi-omics data are improved; By means of time series correlation analysis, the time correlation of different omics data is mined, instead of isolated analysis of single data, so that the indicators related to liver cancer occurrence can be more accurately locked, the interference of invalid indicators on the construction of subsequent evolution framework is reduced, and the precision of screening of core molecular indicators related to liver cancer is improved; Combined with tumor evolution theory, a framework containing time axis is built, the order of molecular event occurrence is determined, and the regulation relationship is clarified through interactive network, so that the originally scattered molecular information forms a systematic and orderly system, the confusion in understanding the mechanism of liver cancer occurrence is reduced, the individual disease development rule is easier to grasp, and the logic of individualized liver cancer evolution analysis is enhanced; Through pathological data such as liver biopsy for verification and correction, the atlas is not only a theoretical result generated by algorithm, but also a tool that fits the actual condition, the deviation between the atlas and the clinical practice is reduced, a more valuable basis for subsequent risk assessment and early warning is provided, and the clinical credibility of individualized liver cancer evolution atlas is improved.

[0013] In another embodiment of the application, step S13 further comprises: Based on the generated multi-omics core correlation data set, a molecular event screening algorithm (such as differential expression analysis, pathway enrichment analysis) is used to extract key molecular events (such as TP53 gene mutation, WNT / β-catenin pathway activation, IL-6 inflammatory factor high expression) directly related to the occurrence and development of liver cancer, and the key molecular events are classified and labeled according to gene mutation-pathway regulation-phenotype change to generate a list of key molecular events of liver cancer; According to the generated list of key molecular events of liver cancer, combined with the classical tumor evolution theory (such as multi-stage and multi-step evolution model: normal liver tissue-chronic inflammation-liver fibrosis-liver cirrhosis-pre-cancerous lesion-liver cancer), through event-stage matching algorithm, the key molecular events are attributed to the corresponding evolution stage (such as IL-6 high expression is attributed to chronic inflammation period, TP53 mutation is attributed to pre-cancerous lesion period), and a key molecular event-evolution stage correspondence table is generated; According to the generated key molecular event-evolution stage correspondence table, referring to the longitudinal data time span of the target individual, the expected occurrence time node of each molecular event (such as the chronic inflammation period event is expected to occur at the 6th-12th month of data collection, and the pre-cancerous lesion period event is expected to occur at the 24th-30th month) is labeled, and a preliminary molecular event time axis is generated; Based on the generated preliminary molecular event timeline, combined with the stage progression logic of tumor evolution theory, the basic structure of an individualized liver cancer evolution framework is constructed and divided into modules, including a normal state module, an inflammatory response module, a fibrosis module, a cirrhosis module, and a precancerous early warning module. Each module is associated with the key molecular events and time nodes of the corresponding stage, generating a draft of the basic framework structure. Based on the generated draft framework, an evolutionary logic verification algorithm (such as reverse tracing verification: verifying from precancerous events whether there are supporting early inflammatory / fibrotic events) is used to identify and correct mismatches in event stages (such as misclassifying late-stage liver cancer mutations as fibrotic stages), generate a logically verified framework structure, and supplement the transition relationships between modules (such as the key triggering event for the transition from the inflammation module to the fibrosis module: TGF-β pathway activation), clearly mark the priority of the occurrence order of each molecular event, and generate a preliminary liver cancer evolution framework.

[0014] The working principle and effects of the above technical solution are as follows: By extracting events directly related to liver cancer through differential expression analysis and pathway enrichment analysis, and classifying and labeling them according to gene mutation, pathway regulation and phenotypic changes, irrelevant molecular information can be accurately removed, reducing the interference of messy data on subsequent framework construction, avoiding the confusion of framework logic due to mixed events, and improving the targeting of key molecular event screening. Combining classical tumor evolution theory and event-stage matching algorithms, such as classifying IL-6 high expression to the chronic inflammation stage and TP53 mutation to the precancerous lesion stage, can clearly identify the stage to which an event belongs, avoid problems such as "misclassifying late-stage liver cancer mutations to the fibrosis stage", reduce the bias caused by stage mismatch to the framework, and reduce the probability of mismatch between molecular events and evolution stages. By referencing the longitudinal data time span of the target individual to label event nodes, such as marking precancerous lesion events in the 24th-30th month of data collection, the framework is no longer a general template, but more in line with the rhythm of individual data, reducing the disconnect between the general framework and the actual situation of individuals, and enhancing the individualized adaptability of the evolutionary framework. By retrospectively tracing and verifying mismatches, and supplementing the transitional events between modules (such as TGF-β pathway activation connecting inflammation and fibrosis modules), the order of events is clarified, making the framework structure more coherent, reducing subsequent risk assessment errors caused by logical gaps, and improving the logical rigor of the evolutionary framework.

[0015] In another embodiment of the present invention, step S2 further includes: S21: Based on the generated validated individualized liver cancer evolution atlas, identify the liver tissue regions corresponding to the key stages of liver cancer development (such as cirrhotic nodule areas and inflammatory infiltration areas), formulate targeted medical imaging acquisition plans (such as MRI focused scanning of key areas and CT dynamic contrast-enhanced scanning), and generate targeted image acquisition plans; according to the generated targeted image acquisition plans, use MRI, CT, and ultrasound elastography techniques to acquire image data, obtain raw image data related to the structure, blood flow, and stiffness of liver tissue, and generate a raw liver image dataset. S22: Based on the generated raw liver image dataset, key regions of the liver microenvironment are extracted using image segmentation algorithms (such as the U-Net deep learning segmentation model). Then, through image quantification tools (such as CT value measurement and elastic modulus calculation), the degree of liver fibrosis, hemodynamic parameters, and extracellular matrix stiffness data are obtained to generate a quantitative dataset of the liver microenvironment pressure field. S23: Based on the generated liver microenvironment pressure field quantification dataset, combined with the standardized multi-omics time series dataset generated in S11, a spatiotemporal coupling algorithm (such as spatiotemporal convolutional neural network) is used to establish the time-space correspondence between pressure field parameters and molecular events, and generate a multi-omics-pressure field spatiotemporal coupling dataset. S24: For the generated multi-omics-pressure field spatiotemporal coupled dataset, use trajectory modeling algorithms (such as hidden Markov models) to simulate the changing trend of liver microenvironment pressure over time, as well as the correlation between this trend and molecular events, and generate a preliminary dynamic risk evolution trajectory of liver cancer. S25: Based on the generated preliminary dynamic risk evolution trajectory of liver cancer, combined with the longitudinal follow-up data of the target individual (such as changes in liver function indicators and the time of symptom onset), the trajectory is optimized and adjusted to ensure that the trajectory can accurately reflect the individual's risk changes, and a validated dynamic risk evolution trajectory of liver cancer is generated.

[0016] The working principle and effects of the above technical solution are as follows: Targeted treatment plans are developed for specific areas in critical stages of liver cancer (such as cirrhotic nodule areas and inflammatory infiltration areas) to avoid blind scanning of the whole liver, reduce unnecessary radiation exposure, reduce image data redundancy, and allow subsequent data processing to focus more on the core areas, thereby improving efficiency and accuracy of liver image acquisition. Using quantitative tools such as U-Net segmentation of key regions and CT value measurement to replace subjective judgment, parameters such as the degree of liver fibrosis and stiffness can be accurately obtained, reducing the error of manual assessment, making the pressure field data more consistent with the actual microenvironment state, providing an accurate basis for subsequent analysis, and improving the reliability of liver microenvironment pressure field data. By establishing the time-space correspondence between the two through the spatiotemporal coupling algorithm, we can no longer view multi-omics data and stress field data in isolation, reduce the one-sidedness of heavy molecules and slight environment, and more comprehensively grasp the intrinsic connection of liver cancer occurrence, and enhance the correlation between molecular events and microenvironmental changes. Hidden Markov models are used to simulate the correlation between stress changes and molecular events, transforming complex data into traceable trends, reducing the difficulty for doctors to understand the data, facilitating the rapid capture of risk change patterns, and improving the intuitiveness of risk change trends. By combining longitudinal follow-up data (such as liver function and symptoms) to optimize the trajectory, we can avoid the trajectory remaining at the theoretical level, reduce the deviation from the actual condition of individuals, make subsequent risk assessments more targeted, and improve the individual adaptability of the risk evolution trajectory.

[0017] In one embodiment of the present invention, step S3 further includes: S31: The generated dynamic risk evolution trajectory of liver cancer after validation is preprocessed by a preprocessing algorithm to generate a preprocessed dynamic risk trajectory dataset of liver cancer. Then, a risk change prediction model is trained by machine learning algorithms (such as deep learning LSTM network and time series ARIMA model). The model identifies the time intervals in the trajectory where the risk rises sharply and generates preliminary risk acceleration period time window data. S32: Based on the generated preliminary risk acceleration period time window data, combined with the individualized liver cancer evolution map generated in S15, extract the key molecular events activated within the time window, record the activation thresholds of molecular events (such as gene expression thresholds and protein concentration thresholds), and generate a risk acceleration period molecular event feature dataset. S33: Using the preliminary risk acceleration period time window data generated in S31, combined with the liver microenvironment pressure field quantitative dataset generated in S22, analyze the abrupt change points (such as sudden increase in stiffness and sudden decrease in blood flow) and the rate of risk increase of the pressure field parameters within this time window, and generate a risk acceleration period pressure field characteristic dataset. S34: Based on the risk acceleration period molecular event feature dataset generated in S32 and the risk acceleration period pressure field feature dataset generated in S33, feature fusion algorithms (such as principal component analysis and feature splicing) are used to integrate the two types of feature data to generate a preliminary risk acceleration period feature dataset. Invalid features (such as parameters that are not significantly related to risk) are removed through cross-validation (such as K-fold cross-validation), and feature dimensions and weights are optimized to generate the final risk acceleration period feature dataset.

[0018] The working principle and effects of the above technical solution are as follows: By preprocessing trajectory data to purify it, and then using machine learning models such as LSTM and ARIMA to capture the intervals where the risk increases sharply, it can more accurately pinpoint the key time window than manual trajectory analysis, reduce the chance of missing the risk acceleration period or misjudging the time range, provide an accurate time benchmark for subsequent feature extraction, and improve the accuracy of risk acceleration period identification. It combines evolutionary maps to extract molecular event activation thresholds (such as gene expression and protein concentration thresholds) and stress field data to capture parameter mutation points (such as sudden increase in stiffness and sudden decrease in blood flow), avoiding the one-sidedness of relying on only a single dimension of features, reducing misjudgment of risk features due to incomplete information, and enhancing the comprehensiveness of risk acceleration phase features. By integrating molecular and pressure field features through a feature fusion algorithm, and then using K-fold cross-validation to remove invalid parameters (such as indicators that are not related to risk), the interference of irrelevant data on subsequent analysis is reduced. At the same time, the feature dimensions and weights are optimized to make the feature set more focused on core risk information, reduce the error of subsequent case comparison, and improve the reliability of the final feature set. The optimized final risk acceleration feature set removes redundant information and has a more streamlined dimension. When comparing with clinical cases, there is no need to process messy data, which improves analysis efficiency, reduces the waste of computing resources caused by excessive feature dimensions, and reduces the complexity of feature processing.

[0019] In another embodiment of the present invention, step S34 further includes: Based on the risk acceleration period molecular event feature dataset generated by S32 and the risk acceleration period pressure field feature dataset generated by S33, data dimension verification (such as time series length matching and sample size consistency check) is performed. The numerical scale of the two types of features is unified through data standardization processing (such as Min-Max normalization) to generate an aligned and standardized multi-source feature dataset. Based on the generated aligned and normalized multi-source feature dataset, feature fusion algorithms (such as attention mechanism weighted fusion and multi-kernel learning fusion) are used to map molecular event features and stress field features to the same feature space, retain key correlation information (such as the temporal coupling relationship between molecular activation and stress mutation), and generate a preliminary fused feature dataset. Based on the generated preliminary fused feature dataset, features that are not significantly associated with the risk of liver cancer (such as features with an absolute value of correlation coefficient < 0.15) are identified through feature importance assessment (such as XGBoost feature scoring and Pearson correlation coefficient calculation), and are marked as invalid features to be removed, thus generating a list of feature importance assessments. Based on the generated feature importance assessment list, K-fold cross-validation (such as 5-fold cross-validation) is used to iteratively filter the preliminary fused feature dataset. Each validation removes 30% of the least important features until the feature dimension is reduced to 50%-60% of the original dimension, generating a simplified feature dataset. Based on the generated simplified feature dataset, feature weight optimization algorithms (such as elastic network regularization and particle swarm optimization) are used to adjust the weight distribution of the remaining features (such as increasing the weight of the top 20% of features by 20%), strengthen the contribution of core features to risk prediction, and generate a weighted optimized feature dataset. Based on the generated weighted optimized feature dataset, and combined with follow-up data of clinical liver cancer cases (such as the time difference between the risk acceleration period and the actual onset), the dataset is validated. When the risk prediction accuracy of the feature set is ≥80%, it is determined as the final risk acceleration period feature dataset.

[0020] The working principle and effects of the above technical solution are as follows: Dimension verification is performed by matching time series lengths and checking sample size consistency. Then, Min-Max normalization is used to unify the numerical scale of molecular events and pressure field features. This avoids fusion bias caused by different formats and scales of the two types of data, reduces basic interference in subsequent feature processing, and allows risk features from different sources to be combined smoothly, thereby improving the compatibility of multi-source feature data. When using attention mechanisms and multi-core learning to fuse features, the temporal coupling relationship between molecular activation and stress mutation is deliberately preserved to prevent the two types of features from existing in isolation. At the same time, invalid features are accurately identified through XGBoost scores and Pearson coefficients to reduce redundant information occupation, allowing the feature set to focus more on the core information that is truly related to liver cancer risk, and enhancing the correlation and effectiveness between features. After iteratively eliminating 30% of low-importance features through 5-fold cross-validation, the dimensionality is compressed to 50%-60% of the original, and the computational efficiency is significantly improved after simplification. Then, by adjusting the weights through elastic network regularization and particle swarm optimization, the contribution of the top 20% of core features is increased, important features are prevented from being weakened, the error of missing key information in subsequent risk prediction is reduced, and the computational burden and error of feature processing are reduced. By combining clinical follow-up data for verification, we ensure an accuracy rate of ≥80%, making the feature set not just a theoretical result generated by the algorithm, but a practical tool that fits the actual incidence of liver cancer. This reduces the disconnect between theoretical features and clinical practice, provides a reliable basis for subsequent case comparisons and risk assessments, and improves the clinical credibility of the final feature set.

[0021] In another embodiment of the present invention, step S4 further includes: S41: Based on the generated final risk acceleration phase feature dataset, retrieve the clinically known liver cancer case database (containing multi-omics data, imaging data and pathological results of cases), use a case matching algorithm (such as cosine similarity matching) to screen liver cancer cases with similar characteristics to the target individual, and generate a similar liver cancer case dataset. S42: Based on the generated similar liver cancer case dataset, combined with the original liver image dataset generated in S21, use image comparison analysis algorithms (such as lesion area overlap calculation) to locate areas in the liver of the target individual that are similar to the lesion area features of the case, and generate preliminary potential hotspot area data for liver cancer. S43: Based on the generated preliminary potential hotspot area data of liver cancer, combined with the liver microenvironment pressure field quantification dataset generated in S22, analyze the spatial distribution characteristics of the pressure field in the hotspot area (such as high pressure value cluster area and abnormal pressure gradient area), verify and correct the range of the hotspot area, and generate verified potential hotspot area data of liver cancer. S44: Based on the similar liver cancer case dataset generated in S41, extract the clinical characteristics (such as age, type of underlying liver disease, and history of alcohol consumption) of high-risk groups of liver cancer in the cases. Combine the demographic and clinical data of the target individuals and use a subgrouping algorithm (such as K-means clustering) to determine the high-risk subgroup type to which the target individuals belong, and generate individualized high-risk subgroup identification results. S45: Based on the validated potential hotspot region data of liver cancer generated in S43 and the individualized high-risk subgroup identification results generated in S44, the validated individualized liver cancer evolution map generated in S15 is dynamically updated, and information on the distribution of molecular events in hotspot regions and the risk association information of high-risk subgroups is supplemented to generate a preliminary refined liver cancer risk warning map. S46: Based on the preliminary refined liver cancer risk warning map generated in S45, the risk area labeling and molecular mechanism explanation in the map are corrected through clinical expert review (such as inviting more than 3 chief physicians in the field of liver cancer to score), and the final refined liver cancer risk warning map is generated.

[0022] The working principle and effects of the above technical solution are as follows: By screening similar liver cancer cases based on the characteristics of the final risk acceleration phase, we can avoid blindly comparing all cases, reduce the interference of irrelevant cases on the analysis, and provide more relevant reference for subsequent hot spot area positioning and high-risk subgroup identification. This will reduce the judgment bias caused by case mismatch and improve the targeting and effectiveness of case matching. First, preliminary localization is achieved through image comparison analysis, and then verification and correction are made by combining pressure field quantitative data. This avoids relying solely on subjective judgment of the area based on images, reduces mislabeling or omission of hotspot areas, makes the labeling of risk areas more consistent with the actual condition, improves the accuracy of locating potential hotspot areas of liver cancer, and enhances the individualized adaptability of high-risk subgroup identification. By extracting the clinical characteristics of high-risk groups in the cases and combining them with the target individual data to divide them into subgroups using K-means clustering, instead of applying general subgroup standards, we can reduce the error of one-size-fits-all judgments and more accurately identify the risk group type to which an individual belongs.

[0023] When dynamically updating the evolutionary map, information on molecular events in hotspot regions and associations with high-risk subgroups is added. The map is then reviewed and revised by more than three liver cancer experts to avoid the map relying solely on algorithm generation and becoming detached from clinical reality. This reduces biases in risk region labeling and molecular mechanism interpretation, making the early warning map more clinically valuable and improving its precision and clinical reliability.

[0024] In another embodiment of the present invention, step S41 further includes: Based on the generated final risk acceleration period feature dataset, core feature dimensions (such as molecular event activation threshold, risk rise rate, and stress field mutation point parameters) are extracted. Feature standardization processing (such as Z-score transformation) is used to unify the feature numerical range, eliminate dimensional differences, and generate a standardized target individual risk feature set. We retrieved clinically known liver cancer case databases and screened complete cases that included accelerated risk phase characteristics, multi-omics data, imaging data, and pathological results (excluding cases with missing key information) to form an effective subset of cases in the case database. Based on the generated standardized target individual risk feature set, the core dimensions of case matching are determined (such as molecular event type matching, risk escalation rate deviation ≤20%, and stress field parameter similarity ≥60%), and case matching dimensions and threshold rules are formulated. Based on the generated subset of valid cases in the case library and the established matching rules, the risk acceleration period characteristics of each valid case are standardized to generate a standardized case risk feature set. Using a case matching algorithm (such as cosine similarity matching), the similarity between the generated standardized target individual risk feature set and each generated standardized case risk feature set is calculated to generate a target-case similarity matrix; Set a similarity threshold (e.g., similarity ≥ 0.7), select cases that meet the threshold from the generated target-case similarity matrix, integrate the multi-omics, imaging and pathological data of these cases, and generate a similar liver cancer case dataset.

[0025] The working principle and effects of the above technical solution are as follows: By using Z-score conversion to unify the feature value range of target individuals and cases, matching biases caused by differences in the units of molecular event thresholds and pressure field parameters are eliminated, allowing risk features from different sources to be compared on the same scale, reducing misjudgments caused by different data formats, laying an accurate foundation for subsequent screening of similar cases, and improving the accuracy of case matching. Cases lacking key information are removed from the database, and only complete cases containing characteristics of the accelerated risk period, multi-omics and pathological data are retained. This avoids analytical bias caused by using incomplete cases as references, makes the effective subset of cases in the case library more valuable, improves the reliability of subsequent matching, and reduces interference from invalid cases. Clearly define core dimensions and thresholds such as molecular event type matching and risk rate deviation ≤20% to avoid the arbitrariness of case selection based on subjective experience, make case matching systematic, reduce the misselection or omission of similar cases due to vague standards, improve the standardization of the screening process, and enhance the rigor of the matching rules. By quantifying the feature association between the target individual and the case through cosine similarity calculation, and then filtering by a threshold of ≥0.7, the selected cases are ensured to be highly consistent with the risk characteristics of the target individual, reducing the inclusion of irrelevant cases, providing a reference basis that fits the individual for subsequent hotspot area positioning and high-risk subgroup identification, and improving the relevance of similar cases.

[0026] In another embodiment of the present invention, step S45 further includes: Based on the validated potential hotspot region data of liver cancer generated in S43, the individualized high-risk subgroup identification results generated in S44, and the validated individualized liver cancer evolution map generated in S15, data format standardization processing was adopted (such as converting the spatial coordinates of hotspot regions into map-compatible grid coordinates and converting the features of high-risk subgroups into structured attribute fields) to eliminate data format differences and generate an aligned multi-source basic dataset. Based on the generated aligned multi-source basic dataset, the core information of potential hotspot regions of liver cancer after validation (such as region boundaries, corresponding molecular event types and expression levels) is extracted. Through the map coordinate mapping algorithm, the distribution of hotspot regions and associated molecular events is labeled to the corresponding spatial positions of the validated individualized liver cancer evolution map, generating an evolution map with hotspot region labels. Based on the generated evolutionary map with hotspot region annotations, combined with the individualized high-risk subgroup identification results (such as subgroup type and risk association factors) in the aligned multi-source base dataset, a risk association rule embedding algorithm (such as associating "hepatitis B cirrhosis subgroup + hotspot region TP53 mutation" as a high-risk combination) is used to supplement the risk association information between subgroups and molecular events and hotspot regions, and generate an evolutionary map that integrates high-risk subgroup information. Based on the generated evolutionary map that integrates information on high-risk subgroups, a logical relationship verification algorithm is used to verify the consistency between the distribution of molecular events in hotspot regions and the risk information of high-risk subgroups (such as checking contradictory data that "low-risk subgroups correspond to high mutation hotspot regions"). Duplicate or conflicting information is removed, and the logical deviations of risk associations in the map are corrected to generate an evolutionary map after logical verification. Based on the generated evolutionary map after logical verification, risk level labels are added (e.g., high, medium, and low risk areas are labeled with red, yellow, and blue respectively), and brief explanations of molecular mechanisms are added (e.g., hotspot areas highly express VEGF: promoting angiogenesis and accelerating tumor progression), enhancing the readability of the map information and generating an information-enhanced preliminary early warning map. Based on the generated information-enhanced preliminary early warning map, a map integrity check (such as confirming that all hotspot areas and high-risk subgroup association information have been included) is performed to ensure that no key information is omitted, and finally a preliminary refined liver cancer risk early warning map is generated.

[0027] The working principle and effects of the above technical solution are as follows: By standardizing the format, the coordinates of hotspot areas and the characteristics of high-risk subgroups are converted into a map-compatible format, eliminating integration barriers caused by data format differences, reducing map update chaos caused by data incompatibility, and allowing hotspot area and subgroup information to be smoothly integrated into the evolution map, thereby improving the fusion and adaptability of multi-source data. By marking hotspot areas and related molecular events to their corresponding spatial locations on the map, the ambiguity of risk areas described only in words is avoided. Doctors can intuitively see the distribution of high-risk areas, reduce the time cost of locating risk areas, improve the efficiency of risk identification, and enhance the spatial risk orientation of the map. By using rule embedding algorithms to establish risk combinations of subgroups, molecular events, and hotspot regions (such as hepatitis B cirrhosis subgroups + TP53 mutation hotspots), we can avoid presenting the three types of information in isolation and reduce the one-sidedness of understanding the risk mechanism. Then, we can use logical verification to check contradictory data (such as low-risk subgroups corresponding to high mutation regions) to reduce the logical bias of the graph and improve the logicality of risk association. Using red, yellow, and green to indicate risk levels and adding explanations of molecular mechanisms transforms complex data into intuitive information, preventing doctors from making judgments due to excessive time spent interpreting professional data, and making the warning chart easier to apply in clinical practice; the completeness check ensures that no key information is omitted, improves the overall reliability of the chart, and enhances its clinical readability.

[0028] In another embodiment of the present invention, step S5 further includes: S51: Extract core risk information (such as the number of hotspot areas, risk level of high-risk subgroups, and activation degree of key molecular events) from the generated final refined liver cancer risk warning map. Combine the quantitative results of the liver microenvironment pressure field generated in S22 with clinically relevant information (such as liver function classification and viral load). Use a weighted summation algorithm (weights are determined through clinical data training) to calculate the liver cancer risk value of the target individual and generate an individualized liver cancer risk index. S52: Based on the generated individualized liver cancer risk index, referencing the clinical liver cancer incidence probability thresholds (such as low risk <5%, medium risk 5%-20%, high risk >20%), and combining the longitudinal risk change trend of the target individual (from the evolutionary trajectory of S25), set different levels of risk warning thresholds (such as low risk warning threshold, medium and high risk intervention threshold), and generate an individualized risk warning threshold system. S53: Based on the generated individualized risk warning threshold system, compare the relationship between the individualized liver cancer risk index generated in S51 and the threshold: if the index exceeds the corresponding threshold, the warning mechanism is triggered, the warning level and the triggering reason (such as the expansion of hotspot areas or the activation of molecular events) are recorded, and early warning data for liver cancer onset is generated. S54: Based on the generated early warning data of liver cancer (or the risk index if no warning is triggered), combined with the final risk acceleration period feature dataset generated in S35, formulate preliminary intervention directions (such as lifestyle adjustment and medical intervention) for the risk sources (such as metabolic abnormalities and inflammation-driven factors), and generate preliminary personalized prevention and intervention plans. S55: Based on the generated preliminary personalized prevention and intervention plan, retrieve the clinical intervention effect database (including the effectiveness and side effect data of different intervention measures), use the plan optimization algorithm (such as multi-objective optimization algorithm), adjust the specific content of the intervention measures (such as screening frequency and drug dosage), and generate the optimized personalized prevention and intervention plan. S56: Based on the generated optimized personalized prevention and intervention plan, combined with the early warning data of liver cancer generated in S53, a standardized report (including risk index, warning results, and details of intervention measures) is compiled and uploaded to the clinical management system to provide doctors with a basis for formulating treatment plans and complete the output of liver cancer risk prediction and intervention recommendations.

[0029] The working principle and effects of the above technical solution are as follows: Extract the core information of the early warning map, combine the results of pressure field quantification with clinical data, and calculate the risk index using weights determined by clinical training. This avoids the one-sidedness of assessment based on single data, reduces risk misjudgment caused by incomplete data, makes the index more consistent with the actual condition of individuals, and improves the accuracy of individualized liver cancer risk assessment. By referencing clinical probability thresholds and combining them with longitudinal risk change trends, individualized thresholds are set instead of applying a uniform standard. This reduces the chances of missed warnings for high-risk individuals due to excessively low thresholds and false warnings for low-risk individuals due to excessively high thresholds, making warnings more accurate and enhancing the adaptability of warning thresholds. When an alert is triggered, the level and cause are recorded, allowing doctors to intuitively understand the source of the risk, reducing confusion when interpreting simple alert signals, eliminating the need to trace the cause further, improving the efficiency of clinical judgment, and enhancing the practicality of the alert. The initial direction is determined based on the source of risk, and then optimized by combining the clinical intervention effect database, such as adjusting the screening frequency and drug dosage, to avoid the blindness of general solutions, reduce the waste of medical resources and physical burden caused by ineffective interventions, and enhance the pertinence of intervention programs. The standardized report includes risk index, early warning results, and intervention details. It is directly uploaded to the management system, eliminating the need for doctors to manually organize scattered data, reducing data integration time, and enabling them to develop treatment plans more quickly based on the report, thereby improving work efficiency and clinical application efficiency.

[0030] Another embodiment of the present invention provides a prediction system for the risk of developing liver cancer, comprising: One or more processors; Memory, used to store one or more programs. Wherein, when the one or more programs are executed by the one or more processors, the one or more processors are made to implement the method described in any one of the above.

[0031] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for predicting the risk of liver cancer, characterized in that, include: S1: Collect longitudinal multi-omics data from the target individuals to generate individualized multi-omics time-series datasets; An individualized liver cancer evolution atlas was constructed based on the dataset. S2: Based on the individualized liver cancer evolution map, the liver microenvironment stress field is quantified by combining medical imaging technology. The stress field quantification results are spatially and temporally coupled with multi-omics time series datasets to generate a dynamic risk evolution trajectory of liver cancer. S3: Based on the dynamic risk evolution trajectory of liver cancer, machine learning algorithms are used to identify the risk acceleration phase; features are extracted from the risk acceleration phase to generate a risk acceleration phase feature dataset; S4: Compare and analyze the risk acceleration phase feature dataset with clinically known liver cancer cases, and combine the spatial distribution characteristics of the liver microenvironment pressure field to identify potential hotspot areas and high-risk subgroups of liver cancer; based on the information of hotspot areas and high-risk subgroups, dynamically update the individualized liver cancer evolution map to generate a refined liver cancer risk warning map. S5: Calculate an individualized liver cancer risk index based on the refined liver cancer risk warning map; Different warning thresholds are set based on the risk index. When the risk index exceeds the corresponding threshold, early warning data for liver cancer is generated, and personalized prevention and intervention suggestions are provided.

2. The method for predicting the risk of liver cancer according to claim 1, characterized in that, S1 includes: S11: Employing multimodal molecular detection technology, longitudinal data collection of target individuals is conducted over a period of 1-3 years to generate raw multi-omics datasets for individuals, and preprocessing is performed to generate standardized multi-omics time-series datasets. S12: Based on the generated standardized multi-omics time series dataset, time series correlation analysis algorithm is used to explore the temporal correlation relationship between different omics data and generate a multi-omics core correlation dataset. S13: Based on the generated multi-omics core association dataset and combined with tumor evolution theory, construct an individualized liver cancer evolution framework that includes a timeline of molecular events, preliminarily define the order of occurrence of key molecular events, and generate a preliminary liver cancer evolution framework. S14: Using the generated preliminary liver cancer evolution framework, a network construction algorithm is used to draw the interaction network between key molecular events, clarify the regulatory relationship of molecular events, and generate an individualized liver cancer evolution map; S15: Based on the generated individualized liver cancer evolution atlas, it is validated and corrected using clinical pathological data to generate a validated individualized liver cancer evolution atlas.

3. The method for predicting the risk of liver cancer according to claim 2, characterized in that, S13 includes: Based on the generated multi-omics core association dataset, key molecular events directly related to the occurrence and development of liver cancer are extracted and classified and labeled according to gene mutation, pathway regulation and phenotypic changes to generate a list of key molecular events of liver cancer. Based on the generated list of key molecular events in liver cancer, and combined with classical tumor evolution theory, the key molecular events are assigned to the corresponding evolutionary stages using an event-stage matching algorithm, thus generating a key molecular event-evolutionary stage correspondence table. Based on the generated key molecular event-evolutionary stage correspondence table, and referring to the longitudinal data time span of the target individual, the expected occurrence time node of each molecular event is marked to generate a preliminary molecular event timeline; Based on the generated preliminary molecular event timeline, combined with the stage progression logic of tumor evolution theory, the basic structure of the individualized liver cancer evolution framework is constructed and divided into modules. Each module is associated with the key molecular events and time nodes of the corresponding stage, generating a draft of the basic framework structure. Based on the generated framework infrastructure draft, an evolutionary logic verification algorithm is used to identify and correct event stage mismatches, generate a logically verified framework structure, and supplement the transition relationships between modules to generate a preliminary liver cancer evolution framework.

4. The method for predicting the risk of liver cancer according to claim 1, characterized in that, S2 includes: S21: Based on the generated validated individualized liver cancer evolution map, identify the liver tissue regions corresponding to the key stages of liver cancer development, formulate targeted medical imaging acquisition plans, and generate targeted imaging acquisition plans; according to the generated targeted imaging acquisition plans, collect imaging data, obtain raw imaging data, and generate a raw liver imaging dataset. S22: Extract key regions of the liver microenvironment from the generated raw liver image dataset, and then use image quantification tools to obtain data on the degree of liver fibrosis, hemodynamic parameters and extracellular matrix stiffness, and generate a quantitative dataset of the liver microenvironment pressure field. S23: Based on the generated liver microenvironment pressure field quantitative dataset, combined with the standardized multi-omics time series dataset generated in S11, a spatiotemporal coupling algorithm is used to establish the temporal-spatial correspondence between pressure field parameters and molecular events, and generate a multi-omics-pressure field spatiotemporal coupling dataset. S24: Using the generated multi-omics-pressure field spatiotemporal coupled dataset, a trajectory modeling algorithm is applied to simulate the changing trend of liver microenvironment pressure over time, as well as the correlation between this trend and molecular events, to generate a preliminary dynamic risk evolution trajectory for liver cancer. S25: Based on the generated preliminary dynamic risk evolution trajectory of liver cancer, combined with the longitudinal follow-up data of the target individuals, the trajectory is optimized and adjusted to generate a validated dynamic risk evolution trajectory of liver cancer.

5. The method for predicting the risk of liver cancer according to claim 1, characterized in that, S3 includes: S31: The generated dynamic risk evolution trajectory of liver cancer after validation is preprocessed by a preprocessing algorithm to generate a preprocessed dynamic risk trajectory dataset of liver cancer. A risk change prediction model is trained by a machine learning algorithm. The model identifies the time intervals in the trajectory where the risk rises sharply and generates preliminary risk acceleration period time window data. S32: Based on the generated preliminary risk acceleration period time window data, combined with the individualized liver cancer evolution map generated in S15, extract the key molecular events activated within the time window, record the activation threshold of the molecular events, and generate a risk acceleration period molecular event feature dataset. S33: Using the preliminary risk acceleration period time window data generated in S31, combined with the liver microenvironment pressure field quantitative dataset generated in S22, analyze the mutation points and risk rise rate of the pressure field parameters within this time window, and generate a risk acceleration period pressure field characteristic dataset. S34: Based on the results generated by S32 and S33, a feature fusion algorithm is used to integrate the two types of feature data to generate a preliminary risk acceleration period feature dataset. Invalid features are then removed through cross-validation, and the feature dimensions and weights are optimized to generate the final risk acceleration period feature dataset.

6. The method for predicting the risk of liver cancer according to claim 5, characterized in that, S34 includes: The results generated by S32 and S33 are validated for data dimensions. The numerical scale of the two types of features is unified through data standardization, and an aligned and standardized multi-source feature dataset is generated. Based on the generated aligned and normalized multi-source feature dataset, a feature fusion algorithm is used to map molecular event features and pressure field features to the same feature space, retain key correlation information, and generate a preliminary fused feature dataset. Based on the generated preliminary fused feature dataset, features that are not significantly associated with the risk of liver cancer are identified through feature importance assessment and marked as invalid features to be removed, thus generating a feature importance assessment list; Based on the generated feature importance assessment list, K-fold cross-validation is used to iteratively filter the preliminary fused feature dataset. Each validation removes 30% of the least important features until the feature dimension is reduced to 50%-60% of the original dimension, generating a simplified feature dataset. Based on the generated simplified feature dataset, a feature weight optimization algorithm is used to adjust the weight distribution of the remaining features and generate a weighted optimized feature dataset. Based on the generated weighted optimized feature dataset, and validated by follow-up data of clinical liver cancer cases, when the risk prediction accuracy of the feature set is ≥80%, it is determined as the final risk acceleration phase feature dataset.

7. The method for predicting the risk of liver cancer according to claim 1, characterized in that, S4 includes: S41: Based on the generated final risk acceleration phase feature dataset, retrieve the clinically known liver cancer case database, use a case matching algorithm to screen liver cancer cases with similar characteristics to the target individual, and generate a similar liver cancer case dataset. S42: Based on the generated similar liver cancer case dataset and combined with the original liver image dataset generated in S21, the image comparison analysis algorithm is used to locate the area in the liver of the target individual that is similar to the lesion area of ​​the case, and generate preliminary potential hot spot area data of liver cancer. S43: Based on the generated preliminary potential hotspot area data of liver cancer, combined with the liver microenvironment pressure field quantitative dataset generated in S22, the spatial distribution characteristics of the pressure field of the hotspot area are analyzed, the range of the hotspot area is verified and corrected, and the verified potential hotspot area data of liver cancer is generated. S44: Based on the similar liver cancer case dataset generated by S41, extract the clinical characteristics of high-incidence groups of liver cancer in the cases, combine the demographic and clinical data of the target individuals, use the subgroup segmentation algorithm to determine the high-risk subgroup type to which the target individuals belong, and generate individualized high-risk subgroup identification results. S45: Based on the validated potential hotspot region data of liver cancer generated in S43 and the individualized high-risk subgroup identification results generated in S44, the validated individualized liver cancer evolution map generated in S15 is dynamically updated, and information on the distribution of molecular events in hotspot regions and the risk association information of high-risk subgroups is supplemented to generate a preliminary refined liver cancer risk warning map. S46: Based on the preliminary refined liver cancer risk warning map generated in S45, the map content is revised through clinical expert review to generate the final refined liver cancer risk warning map.

8. The method for predicting the risk of liver cancer according to claim 7, characterized in that, S41 includes: Based on the generated final risk acceleration period feature dataset, core feature dimensions are extracted, feature standardization is used to unify the feature value range, eliminate dimensional differences, and generate a standardized target individual risk feature set. Retrieve clinically known liver cancer case databases, filter complete cases from the database, and form an effective subset of cases in the case database; Based on the generated standardized target individual risk feature set, the core dimensions of case matching are determined, and case matching dimension and threshold rules are formulated. Based on the generated subset of valid cases in the case library and the established matching rules, the risk acceleration period characteristics of each valid case are standardized to generate a standardized case risk feature set. A case matching algorithm is used to calculate the similarity between the generated standardized target individual risk feature set and the generated standardized case risk feature set, and to generate a target-case similarity matrix. By setting a similarity threshold, cases that meet the threshold requirements are selected from the generated target-case similarity matrix. Multi-omics, imaging and pathological data of these cases are integrated to generate a similar liver cancer case dataset.

9. The method for predicting the risk of liver cancer according to claim 1, characterized in that, S5 includes: S51: Extract core risk information from the generated final refined liver cancer risk warning map, combine it with the quantitative results of the liver microenvironment pressure field generated in S22 and clinical related information, and use a weighted summation algorithm to calculate the liver cancer risk value of the target individual and generate an individualized liver cancer risk index. S52: Based on the generated individualized liver cancer risk index, referencing the clinical liver cancer incidence probability threshold, and combining the longitudinal risk change trend of the target individual, set different levels of risk warning thresholds to generate an individualized risk warning threshold system. S53: Based on the generated individualized risk warning threshold system, compare the relationship between the individualized liver cancer risk index generated in S51 and the threshold: if the index exceeds the corresponding threshold, the warning mechanism is triggered, the warning level and triggering reason are recorded, and early warning data for liver cancer onset is generated. S54: Based on the generated early warning data of liver cancer, combined with the final risk acceleration phase feature dataset generated in S35, formulate preliminary intervention directions for the sources of risk and generate preliminary personalized prevention and intervention plans. S55: Based on the generated preliminary personalized prevention and intervention plan, retrieve the clinical intervention effect database, use the plan optimization algorithm to adjust the specific content of the intervention measures, and generate an optimized personalized prevention and intervention plan. S56: Based on the generated optimized personalized prevention and intervention plan, combined with the early warning data of liver cancer generated in S53, a standardized report is compiled and uploaded to the clinical management system to provide doctors with a basis for formulating treatment plans and to complete the output of liver cancer risk prediction and intervention suggestions.

10. A system for predicting the risk of liver cancer, characterized in that, include: One or more processors; Memory, used to store one or more programs. Wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the method of any one of claims 1 to 9.

Images