Method and system for predicting 3-year postoperative recurrence of liver cancer

Abstract

The invention discloses a liver cancer postoperative three-year recurrence prediction method and system, and relates to the technical field of recurrence prediction, and the method comprises the steps: obtaining the related data of a prediction object, carrying out the preprocessing of the related data of the prediction object, and obtaining the processed related data of the prediction object, the related data of the prediction object comprises preoperative multi-stage enhanced CT image information, pathological data and liver related clinical data of the prediction object; performing feature screening on the processed related data of the prediction object to obtain indexes with significant relevance, inputting the indexes with significant relevance into a pre-established liver cancer postoperative three-year recurrence prediction model, and outputting to obtain a trained liver cancer postoperative three-year recurrence prediction model; performing performance optimization on the trained liver cancer postoperative three-year recurrence prediction model through cross validation and an independent validation set to obtain an optimal liver cancer postoperative three-year recurrence prediction model, and realizing liver cancer postoperative three-year recurrence prediction based on the optimal liver cancer postoperative three-year recurrence prediction model.

Description

Technical Field

[0001] The present invention relates to the technical field of recurrence prediction, and in particular to a method and system for predicting the recurrence of liver cancer 3 years after surgery. Background Art

[0002] Microvascular invasion (MVI) of liver cancer is an important factor affecting the postoperative recurrence and survival prognosis of patients with hepatocellular carcinoma, and is also a biomarker that needs to be considered in clinical decision-making. However, the existing MVI diagnostic methods mainly rely on postoperative pathological examinations, and its acquisition usually requires invasive surgery, making it difficult to comprehensively assess the patient's risk of recurrence before surgery. At the same time, although traditional imaging examination methods (such as CT and MRI) can provide basic imaging characteristics of tumors, their ability to reflect the biological behavior of tumors at the microscopic level is limited, making it difficult to effectively predict the occurrence of MVI.

[0003] In recent years, radiomics technology has demonstrated its potential in reflecting tumor microenvironment characteristics and heterogeneity by extracting high-throughput quantitative features from medical images. At the same time, the introduction of machine learning methods provides a powerful tool for the screening and modeling of radiomics features. However, prediction models based solely on radiomics features often ignore the clinical and pathological characteristics of liver cancer patients, which may lead to insufficient prediction performance. In addition, traditional machine learning models have certain "black box" problems in feature selection and model interpretation, which limits their application and promotion in clinical practice. Summary of the invention

[0004] In order to solve the deficiencies mentioned in the above background technology, the purpose of the present invention is to provide a method and system for predicting the recurrence of liver cancer 3 years after surgery.

[0005] In the first aspect, the purpose of the present invention can be achieved by the following technical scheme: a method for predicting the recurrence of liver cancer 3 years after surgery, the method comprising the following steps:

[0006] Acquiring prediction object related data, preprocessing the prediction object related data to obtain processed prediction object related data, wherein the prediction object related data includes preoperative multi-phase enhanced CT image information, pathological data and liver-related clinical data of the prediction object;

[0007] Performing feature screening on the processed prediction object related data to obtain indicators with significant correlation, inputting the indicators with significant correlation into a pre-established liver cancer recurrence prediction model 3 years after surgery, and outputting a trained liver cancer recurrence prediction model 3 years after surgery;

[0008] The performance of the trained liver cancer recurrence prediction model 3 years after surgery was optimized through cross-validation and independent validation sets to obtain the optimal liver cancer recurrence prediction model 3 years after surgery. Based on the optimal liver cancer recurrence prediction model 3 years after surgery, the recurrence prediction of liver cancer 3 years after surgery was realized.

[0009] In combination with the first aspect, in certain implementations of the first aspect, the method further includes: the preprocessing of the prediction object related data includes data standardization, missing value processing and image resolution consistency.

[0010] In combination with the first aspect, in some implementations of the first aspect, the method further includes: preprocessing the prediction object related data:

[0011] The final extracted imaging features need to be normalized. The normalization formula is as follows:

[0012] Xn normalized=(Xn–Xmin) / (Xmax–Xmin);

[0013] Xn represents any numeric variable, Xn normalized represents the normalized value of the numeric variable Xn, Xmax represents the maximum value of the numeric variable, and Xmin represents the minimum value of the numeric variable;

[0014] Continuous variables in clinical and imaging characteristics need to be binarized, and the expression of the restricted cubic spline function is as follows:

[0015] RCS(X)=β 0 X+β 1 S 1 +…+β k-2 S k-2

[0016] Where Si is the cubic component falling in the i-th node.

[0017] In combination with the first aspect, in certain implementations of the first aspect, the method further includes: in the serum index:

[0018] The APRI index formula is:

[0019] APRI=(AST(IU / L) / ULN)×100 / (Platelet count(10^9 / L))

[0020] The FIB-4 formula is:

[0021] FIB-4=age(years)×AST(IU / L) / (Platelet count(10^9 / L)×ALT(IU / L)^1 / 2)

[0022] The mALBI formula is:

[0023] ALBI=(log10 bilirubin×0.66)+(albumin×-0.085)

[0024] In the formula, APRI, FIB-4 and ALBI are commonly used scoring systems in the hepatology department, which are used to evaluate liver fibrosis, cirrhosis and liver function. APRI is the abbreviation of Aspartate Aminotransferase-to-Platelet Ratio Index, that is, the aspartate aminotransferase to platelet ratio index; FIB-4 is the abbreviation of Fibrosis-4 Index; ALBI is the abbreviation of Albumin-Bilirubin, that is, albumin-bilirubin index.

[0025] In combination with the first aspect, in some implementations of the first aspect, the method further includes: the process of performing feature screening on the processed prediction object related data is further fine screening of the screened indicators through lasso regression, and the cost function of the lasso regression 5-fold cross validation is:

[0026]

[0027] is the mean square error term, m is the number of samples, and y i is the actual value of the i-th sample, h θ (x i ) is the model's predicted value for the i-th sample;

[0028] is the L1 regularization term, λ is the regularization parameter, θ j is the jth parameter of the model and n is the total number of parameters.

[0029] In combination with the first aspect, in certain implementations of the first aspect, the method further includes: a process of inputting the significantly correlated indicator into a pre-established model for predicting liver cancer recurrence 3 years after surgery, and outputting a trained model for predicting liver cancer recurrence 3 years after surgery:

[0030] The Logistic regression algorithm is used for construction, where:

[0031] Divide the data set of indicators with significant correlation into a training sample set and a test sample set;

[0032] Logistic regression method was used to establish a prediction model for microvascular invasion of liver cancer, and model parameters were set, including regularization strength and whether to perform feature standardization.

[0033] The training sample set was input into the prediction model of liver cancer recurrence 3 years after surgery to complete the training;

[0034] Input the test sample set into the prediction model of liver cancer recurrence 3 years after surgery, and output the relevant data of liver cancer recurrence status 3 years after surgery;

[0035] The receiver operating characteristic (ROC) curve was established based on the data on the recurrence status of liver cancer 3 years after surgery.

[0036] The pre-established model for predicting liver cancer recurrence 3 years after surgery is as follows:

[0037] The feature set of the joint model is X, which includes radiomics features and imaging features, and the target variable is y, where the microvascular invasion of liver cancer has two classification labels of 0 and 1. When constructing the joint model, the CatBoost algorithm is used and the hyperparameters are optimized through grid search. The model can be expressed as:

[0038]

[0039] Θ opt is the optimal hyperparameter combination obtained through grid search.

[0040] In combination with the first aspect, in certain implementations of the first aspect, the method further includes: optimizing the performance of the trained model for predicting liver cancer recurrence 3 years after surgery through cross-validation and independent validation sets, and evaluating the effect by calculating the AUC value, wherein the calculation process of the AUC value is as follows:

[0041] Calculate the True Positive Rate (TPR) and False Positive Rate (FPR): According to different thresholds, calculate the TPR and FPR under each threshold to obtain a series of points (FPR, TPR). The calculation formulas of TPR and FPR are as follows:

[0042] TPR (True Positive Rate): TPR = TP / (TP+FN)

[0043] FPR (false positive rate): FPR = FP / (FP+TN)

[0044] Among them, TP represents true positive, FP represents false positive, TN represents true negative, and FN represents false negative.

[0045] Draw the ROC curve: Connect these points to form the ROC curve.

[0046] Calculate area: Use the trapezoidal rule to calculate the area under the ROC curve. The specific formula is:

[0047]

[0048] Where i represents the i-th point on the ROC curve. Specifically, FPR i and TPR i They represent the false positive rate and true positive rate of the i-th point respectively. The n in the formula is the number of points on the ROC curve.

[0049] In a second aspect, in order to achieve the above-mentioned purpose, the present invention discloses a system for predicting the recurrence of liver cancer 3 years after surgery, comprising:

[0050] A data processing module is used to obtain prediction object related data, pre-process the prediction object related data, and obtain processed prediction object related data, wherein the prediction object related data includes preoperative multi-phase enhanced CT image information, pathological data and liver-related clinical data of the prediction object;

[0051] The model training module is used to perform feature screening on the processed prediction object related data to obtain indicators with significant correlation, input the indicators with significant correlation into the pre-established liver cancer recurrence prediction model 3 years after surgery, and output the trained liver cancer recurrence prediction model 3 years after surgery;

[0052] The recurrence prediction module is used to optimize the performance of the trained liver cancer recurrence prediction model 3 years after surgery through cross-validation and independent validation sets, to obtain the optimal liver cancer recurrence prediction model 3 years after surgery, and to achieve liver cancer recurrence prediction 3 years after surgery based on the optimal liver cancer recurrence prediction model 3 years after surgery.

[0053] In another aspect of the present invention, in order to achieve the above-mentioned purpose, a terminal device is disclosed, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the memory stores a computer program capable of running on the processor, and when the processor loads and executes the computer program, the method for predicting the recurrence of liver cancer 3 years after surgery as described above is adopted.

[0054] In another aspect of the present invention, in order to achieve the above-mentioned purpose, a computer-readable storage medium is disclosed, in which a computer program is stored. When the computer program is loaded and executed by a processor, the method for predicting the recurrence of liver cancer 3 years after surgery is adopted.

[0055] Beneficial effects of the present invention:

[0056] The beneficial effects of the present invention are mainly reflected in the following aspects:

[0057] Improved prediction accuracy: By integrating preoperative multi-phase enhanced CT imaging information, pathological data, and liver-related clinical data, the present invention can extract features related to liver cancer recurrence 3 years after surgery from multiple dimensions. This fusion of multimodal data significantly enhances the model's ability to identify recurrence risk, and can more accurately assess a patient's recurrence risk compared to prediction methods based on a single data source.

[0058] Optimizing feature screening and model building: The present invention uses advanced feature screening technology to screen out indicators that are significantly associated with postoperative recurrence of liver cancer from massive data, effectively reducing the impact of data noise on prediction results. At the same time, the performance of the model is optimized through cross-validation and independent validation sets, further improving the stability and generalization ability of the model, ensuring that it can maintain a high prediction accuracy on different data sets.

[0059] Improve the efficiency of clinical decision support: The present invention provides an efficient and accurate scientific and quantitative recurrence prediction risk assessment tool for liver cancer after surgery. By predicting the risk of recurrence in advance, doctors can arrange the patient's follow-up plan and intervention measures more reasonably, thereby improving the utilization efficiency of medical resources and improving the patient's postoperative management effect.

[0060] Reduce misdiagnosis and missed diagnosis rates: The prediction model of the present invention is developed based on a large amount of clinical data and advanced machine learning technology, which can effectively reduce the misdiagnosis and missed diagnosis that may occur in traditional diagnostic methods. By accurately predicting the risk of recurrence, it helps to detect signs of recurrence early, take timely intervention measures, and improve the patient's survival rate and quality of life.

[0061] In summary, the present invention significantly improves the accuracy and reliability of prediction of postoperative recurrence of liver cancer through technological innovation and multimodal data fusion. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0063] Figure 1 It is a schematic flow chart of the method of the present invention;

[0064] Figure 2 The present invention relates to a receiver operating curve (ROC) and an area under the curve (AUC) diagram of a prediction model for 3-year postoperative recurrence of liver cancer based on joint analysis of preoperative multi-phase enhanced CT images and clinical data in a training group and a validation group;

[0065] Figure 3The present invention relates to a Kaplan-Meier curve diagram of high, medium and low risk groups for a method for predicting the recurrence of liver cancer 3 years after surgery based on a combined analysis of preoperative multi-phase enhanced CT images and clinical data;

[0066] Figure 4 The present invention relates to a nomogram schematic diagram of a method for predicting the recurrence of liver cancer 3 years after surgery based on combined analysis of preoperative multi-phase enhanced CT images and clinical data;

[0067] Figure 5 It is a schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0068] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] Embodiment 1:

[0070] like Figure 1 As shown, a method for predicting the recurrence of liver cancer 3 years after surgery comprises the following steps:

[0071] S101: Acquire prediction object related data, pre-process the prediction object related data, and obtain processed prediction object related data, wherein the prediction object related data includes preoperative multi-phase enhanced CT image information, pathological data, and liver-related clinical data of the prediction object;

[0072] The final extracted imaging features need to be normalized. The normalization formula is as follows:

[0073] Xn normalized=(Xn–Xmin) / (Xmax–Xmin);

[0074] Xn represents any numeric variable, Xn normalized represents the normalized value of the numeric variable Xn, Xmax represents the maximum value of the numeric variable, and Xmin represents the minimum value of the numeric variable;

[0075] Continuous variables in clinical and imaging characteristics need to be binarized, and the expression of the restricted cubic spline function is as follows:

[0076] RCS(X)=β0X+β1S1+…+βk-2Sk-2

[0077] Where Si is the cubic component falling in the i-th node.

[0078] Among them,

[0079] the clinical data include age, gender, hepatitis history, liver cirrhosis history, and the recurrence time after surgery of the patient;

[0080] the laboratory serum indicators include alpha-fetoprotein (ng / ml), alanine aminotransferase (U / L), aspartate aminotransferase (AST) (U / L), alkaline phosphatase (ALP) (U / L), gamma-glutamyltransferase (GGT) (U / L), lactate dehydrogenase (LDH) (U / L), total bilirubin (TB) (μmol / L), total bile acid (TBA) (μmol / L), white blood cell count (WBC) (10^9 / L), and platelet count (PLT) (10^9 / L). Calculate the evaluation indicators of liver function and liver cirrhosis: A / G, ALP / ALT, AST / ALP, modified albumin-bilirubin mALBI (albumin-bilirubin score), APRI index, FIB-4, DeRitis ratio (AST / ALT);

[0081] The formula for the APRI index is:

[0082] APRI = (AST (IU / L) / ULN) × 100 / (Platelet count (10^9 / L))

[0083] The formula for FIB-4 is:

[0084] FIB-4 = age (years) × AST (IU / L) / (Platelet count (10^9 / L) × ALT (IU / L)^1 / 2)

[0085] The formula for mALBI is:

[0086] ALBI = (log10 bilirubin × 0.66) + (albumin × -0.085)

[0087] Among them, the unit of bilirubin is μmol / L, the unit of albumin is g / L, mALBI grade 1: ALBI ≤ -2.60; mALBI grade 2a: -2.60 < ALBI ≤ -2.00; mALBI grade 2b: -2.00 < ALBI ≤ -1.39; mALBI grade 3: ALBI > -1.39;

[0088] The multi-phase enhanced CT imaging data include the maximum diameter of the tumor, morphology, liver capsule invasion; capsule, hypoenhancement in the arterial phase, low-density ring sign, intratumoral arterial visualization, intrahepatic metastasis, peripheral arterial phase hyperperfusion around the tumor, major vascular invasion, and blood supply situation (arterial, portal, dual blood supply or atypical).

[0089] S102: feature screening is performed on the processed prediction object related data to obtain indicators with significant correlation, and the indicators with significant correlation are input into a pre-established liver cancer recurrence prediction model 3 years after surgery to output a trained liver cancer recurrence prediction model 3 years after surgery;

[0090] A part of the collected data was selected as the training group. Here, the Second Affiliated Hospital of Anhui Medical University was used as the training center, and the First Affiliated Hospital of Anhui Medical University was used as the verification center of the model.

[0091] Afterwards, correlation analysis was performed in the training group to screen out the imaging omics feature indicators with strong robustness;

[0092] Afterwards, the selected indicators are further refined by lasso regression. The cost function of the lasso regression 5-fold cross validation is:

[0093]

[0094] is the mean square error term, m is the number of samples, and y i is the actual value of the i-th sample, h θ (x i ) is the model's predicted value for the i-th sample;

[0095] is the L1 regularization term, λ is the regularization parameter, θ j is the jth parameter of the model, and n is the total number of parameters;

[0096] In the 5-fold cross-validation process, the data set is divided into 5 subsets, 4 of which are used for training each time, and the remaining 1 subset is used for validation. In this way, the model can be trained and validated multiple times to evaluate the performance of the model on different data subsets. Finally, by comparing the model performance under different regularization parameters λ, the λ that minimizes the cost function is selected as the optimal parameter;

[0097] After that, lasso regression 5-fold cross validation was used to select non-zero coefficient variables as the input variables for the final modeling. The horizontal axis is log lambda (logλ), and the vertical axis is binomial deviation (Binomial Deviance) and coefficient value (Coefficience).

[0098] The Logistic regression algorithm was used to train the selected indicators and fit the microvascular invasion status of liver cancer; Logistic regression is a statistical model widely used in binary classification problems. It makes predictions by estimating the probability of an event. Compared with other models, Logistic regression has the advantages of simple model, easy interpretation, high computational efficiency, and is suitable for processing medium-sized data sets. Logistic regression determines the model parameters through maximum likelihood estimation so that the predicted probability is closest to the actual probability of occurrence. In order to improve the predictive performance and generalization ability of the model, the model can be regularized, such as L1 regularization for feature selection and L2 regularization for preventing overfitting. The regularization strength parameter is set to 0.1; in order to further optimize the model, the features can be standardized to eliminate the influence of different feature dimensions and make the model more stable; the steps of using the Logistic regression algorithm to construct a prediction model for liver cancer microvascular invasion based on multiple clinical data are as follows:

[0099] 1) Divide the data set into a training sample set and a test sample set;

[0100] 2) Use the Logistic regression method to establish a prediction model for liver cancer microvascular invasion and set model parameters, including regularization strength and whether to perform feature standardization;

[0101] 3) Input the training sample set into the prediction model of liver cancer microvascular infiltration to complete the training of the model;

[0102] 4) Input the test sample set into the liver cancer microvascular infiltration prediction model and output the relevant data of the liver cancer microvascular infiltration status;

[0103] 5) Receiver operating characteristic (ROC) curve was established based on the relevant data of microvascular infiltration status of liver cancer.

[0104] The pre-established model for predicting liver cancer recurrence 3 years after surgery is as follows:

[0105] The feature set of the joint model is X, which includes radiomics features and imaging features. The target variable is y (binary labels 0 and 1 for liver cancer microvascular invasion). When constructing the joint model (radiomic features + imaging features), the CatBoost algorithm is used and the hyperparameters are optimized through grid search. The model can be expressed as:

[0106]

[0107] Θ opt is the optimal hyperparameter combination obtained through grid search.

[0108] Specific:

[0109] Data grouping: used to select part of the data collected by a certain hospital as the training group, and part of the data collected by another hospital as the validation group;

[0110] Screening for robust imaging omics features: Using inter-group and intra-group consistency analysis, we screened out omics features with an ICC value greater than 0.95 for further construction of the logistic regression omics model;

[0111] Screening clinical and imaging feature indicators: used to include those with P values ​​less than 0.1 in the training group through univariate logistic regression analysis, and finally include the feature indicators with P values ​​less than 0.05 as meaningful feature indicators in the construction of Catboost machine learning prediction model;

[0112] Collinearity analysis was used to check whether the final included characteristic indicators had collinearity (Variance Inflation Factor < 2, Tollerence > 0.5 was considered to be non-collinear);

[0113] S103: The performance of the trained model for predicting liver cancer recurrence 3 years after surgery was optimized through cross-validation and independent validation sets to obtain the optimal model for predicting liver cancer recurrence 3 years after surgery. Based on the optimal model for predicting liver cancer recurrence 3 years after surgery, the prediction of liver cancer recurrence 3 years after surgery was achieved.

[0114] AUC values ​​were used to evaluate the effect and compare between models;

[0115] Although the intratumoral + peritumoral arterial phase radiomics model had the highest AUC value, the intratumoral arterial phase radiomics model had no difference in AUC value and had the advantages of being easier to implement and having fewer features;

[0116] Afterwards, the best radiomics prediction model was selected as the intratumoral omics feature data during the arterial phase, and the Radscore calculation method was as follows:

[0117] Radscore=log.sigma.1.mm.3D_firstorder_Variance×0.1+log.sigma.1.mm.3D_glszm_LowGrayLevelZoneEmphasis×-0.207+wavelet.HH-glcm_InverseVariance×-0.228+exponential_ngtdm_Contrast×-0.025+wavelet. LL_glrlm_RunVariance × 0.284 + wavelet.LH_gldm_LargeDependenceLowGrayLevelEmphasis ×-0.053 +squareroot_glszm_GrayLevelNonUniformity × 0.152 +firstorder_RootMeanSquared×-0.206+wavelet.HL_glrlm_RunEntropy×0.161-0.569

[0118] Wherein, firstorder_RootMeanSquared and log.sigma.1.mm.3D_firstorder_Variance are first-order statistical features, log.sigma.1.mm.3D_glszm_LowGrayLevelZoneEmphasis and squareroot_glszm_GrayLevelNonUniformity are grayscale run matrix features, wavelet.HH-glcm_InverseVariance is the grayscale co-occurrence matrix feature, wavelet.LH_gldm_LargeDependenceLowGrayLevelEmphasis is the grayscale dependency matrix feature, wavelet.LL_glrlm_RunVariance and wavelet.HL_glrlm_RunEntropy are grayscale run length matrix features, exponential_ngtdm_Contrast is the grayscale neighborhood difference matrix feature, wavelet.LL_glrlm_RunVariance, wavelet.HH-glcm_InverseVariance, wavelet.LH_gldm_LargeDependenceLowGrayLevelEmphasis, and wavelet.HL_glrlm_RunEntropy are wavelet transform features.

[0119] Among them, the AUC value is calculated based on the area under the ROC curve. The specific calculation process is as follows:

[0120] Calculate the True Positive Rate (TPR) and False Positive Rate (FPR): According to different thresholds, calculate the TPR and FPR under each threshold to obtain a series of points (FPR, TPR). The calculation formulas of TPR and FPR are as follows:

[0121] TPR (True Positive Rate): TPR = TP / (TP+FN)

[0122] FPR (false positive rate): FPR = FP / (FP+TN)

[0123] Among them, TP represents true positive, FP represents false positive, TN represents true negative, and FN represents false negative.

[0124] Draw the ROC curve: Connect these points to form the ROC curve.

[0125] Calculate area: Use the trapezoidal rule to calculate the area under the ROC curve. The specific formula is:

[0126]

[0127] Where i represents the i-th point on the ROC curve. Specifically, FPR i and TPR i They represent the false positive rate and true positive rate of the i-th point respectively. The n in the formula is the number of points on the ROC curve.

[0128] Specifically, the present invention is further described below by way of embodiments:

[0129] The optimal prediction model for 3-year recurrence of liver cancer after surgery is based on Catboopst's machine learning prediction model, and the machine learning model is interpreted globally and locally based on Shapley additive interpretation. A nomogram model is developed based on the optimal features selected by machine learning, specifically:

[0130] Confirm the best joint model: Based on the model verification and comparison results in step 6, confirm the superiority of the joint model (radiomic features + imaging features combined model), and further optimize the hyperparameters of the model based on the CatBoost algorithm to obtain the joint prediction model with the best performance;

[0131] Global and local feature interpretation based on Shapley additive interpretation (SHAP): Global interpretation: The SHAP value is used to calculate the importance of all model features to the prediction results, determine which features play a leading role in predicting microvascular invasion of liver cancer, and provide a scientific basis for feature selection; Local interpretation: For each sample, the SHAP value is used to provide personalized feature contribution analysis to explain why the model predicts a certain patient as positive or negative for microvascular invasion, thereby improving the clinical interpretability of the model;

[0132] like Figure 4 As shown in the figure, the nomogram model is constructed: the optimal features are selected: according to the global feature importance ranking of the SHAP value, the optimal features (radiology features + imaging features) are selected to construct the nomogram model; the total score is calculated: the selected features are modeled through the Cox proportional hazard regression model to generate the scores corresponding to the features, and the feature values ​​of each sample are mapped to the nomogram to calculate the total score; risk stratification is defined: based on the total score distribution of the recurrence group and the non-recurrence group, two cutoff points are selected to divide the patients into low-risk group, medium-risk group and high-risk group;

[0133] Nomogram model performance evaluation: Predictive ability evaluation: Figure 2 As shown, the ROC curve was used to calculate the AUC value of the nomogram model to predict 1-year, 2-year, and 3-year recurrence, and the 95% confidence interval was reported; Survival analysis verification: Figure 3 Kaplan-Meier survival curves were drawn to compare the survival differences between different risk groups, and the Log-rank test was performed to evaluate its statistical significance.

[0134] Figure 4 The nomogram shows an example of a patient: liver capsule invasion 0 = 3 points, intrahepatic metastasis = 32 points, Radcore-2.5 = 11 points, total score 46 points, belonging to the low-risk group; the patient's probability of recurrence within 1 year, 2 years, and 3 years is approximately 84%, 66%, and 56%, respectively; the predicted 1-year, 2-year, and 3-year AUC values ​​are 0.67 (95% CI 0.57-0.782), 0.78 (95% CI 0.69-0.88), and 0.86 (95% CI 0.78-0.94), respectively.

[0135] Embodiment 2: In the second aspect, as Figure 5 As shown, in order to achieve the above-mentioned purpose, the present invention discloses a system for predicting the recurrence of liver cancer 3 years after surgery, comprising:

[0136] The data processing module 11 is used to obtain the prediction object related data, pre-process the prediction object related data, and obtain the processed prediction object related data, wherein the prediction object related data includes the preoperative multi-phase enhanced CT image information, pathological data and liver-related clinical data of the prediction object;

[0137] The model training module 12 is used to perform feature screening on the processed prediction object related data to obtain indicators with significant correlation, input the indicators with significant correlation into a pre-established liver cancer recurrence prediction model 3 years after surgery, and output a trained liver cancer recurrence prediction model 3 years after surgery;

[0138] The recurrence prediction module 13 is used to optimize the performance of the trained liver cancer recurrence prediction model 3 years after surgery through cross-validation and independent validation sets, obtain the optimal liver cancer recurrence prediction model 3 years after surgery, and realize the recurrence prediction of liver cancer 3 years after surgery based on the optimal liver cancer recurrence prediction model 3 years after surgery.

[0139] Based on the same inventive concept, the present invention also provides a computer device, which includes: one or more processors, and a memory for storing one or more computer programs; the program includes program instructions, and the processor is used to execute the program instructions stored in the memory. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is used to implement one or more instructions, specifically for loading and executing one or more instructions in a computer storage medium to implement the above method.

[0140] It needs to be further explained that, based on the same inventive concept, the present invention also provides a computer storage medium, on which a computer program is stored, and the computer program is executed by a processor to execute the above method. The storage medium can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electrical, magnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0141] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0142] The above shows and describes the basic principles, main features and advantages of the present disclosure. Those skilled in the art should understand that the present disclosure is not limited by the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present disclosure. Without departing from the spirit and scope of the present disclosure, the present disclosure may have various changes and improvements, and these changes and improvements fall within the scope of the present disclosure to be protected.

Claims

1. A method for predicting liver cancer recurrence 3 years after surgery, characterized in that: The method comprises the following steps: Acquiring prediction object related data, preprocessing the prediction object related data to obtain processed prediction object related data, wherein the prediction object related data includes preoperative multi-phase enhanced CT image information, pathological data and liver-related clinical data of the prediction object; Performing feature screening on the processed prediction object related data to obtain indicators with significant correlation, inputting the indicators with significant correlation into a pre-established liver cancer recurrence prediction model 3 years after surgery, and outputting a trained liver cancer recurrence prediction model 3 years after surgery; The performance of the trained liver cancer recurrence prediction model 3 years after surgery was optimized through cross-validation and independent validation sets to obtain the optimal liver cancer recurrence prediction model 3 years after surgery. Based on the optimal liver cancer recurrence prediction model 3 years after surgery, the recurrence prediction of liver cancer 3 years after surgery was realized.

2. The method for predicting liver cancer recurrence 3 years after surgery according to claim 1, characterized in that: The preprocessing of the prediction object related data includes data standardization, missing value processing and image resolution consistency.

3. The method for predicting liver cancer recurrence 3 years after surgery according to claim 2, characterized in that: The preprocessing of the prediction object related data: The final extracted imaging features need to be normalized. The normalization formula is as follows: Xn normalized=(Xn–Xmin) / (Xmax–Xmin); Xn represents any numeric variable, Xn normalized represents the normalized value of the numeric variable Xn, Xmax represents the maximum value of the numeric variable, and Xmin represents the minimum value of the numeric variable; Continuous variables in clinical and imaging characteristics need to be binarized, and the expression of the restricted cubic spline function is as follows: RCS(X)=β0X+β1S1+…+β k-2 S k-2 Where Si is the cubic component falling in the i-th node.

4. The method for predicting liver cancer recurrence 3 years after surgery according to claim 3, characterized in that: Serum indicators: The APRI index formula is: APRI=(AST(IU / L) / ULN)×100 / (Platelet count(10^9 / L)) The FIB-4 formula is: FIB-4=age(years)×AST(IU / L) / (Platelet count(10^9 / L)×ALT(IU / L)^1 / 2) The mALBI formula is: ALBI=(log10 bilirubin×0.66)+(albumin×-0.085) In the formula, APRI, FIB-4 and ALBI are commonly used scoring systems in the hepatology department. APRI is the abbreviation of Aspartate Aminotransferase-to-Platelet Ratio Index, which is the aspartate aminotransferase to platelet ratio index; FIB-4 is the abbreviation of Fibrosis-4 Index, which is the fibrosis-4 index; ALBI is the abbreviation of Albumin-Bilirubin, which is the albumin-bilirubin index.

5. The method for predicting liver cancer recurrence 3 years after surgery according to claim 1, characterized in that: The process of feature screening of the processed prediction object related data is to further fine-screen the screened indicators through lasso regression. The cost function of the lasso regression 5-fold cross validation is: is the mean square error term, m is the number of samples, and y i is the actual value of the i-th sample, h θ (x i ) is the model's predicted value for the i-th sample; is the L1 regularization term, λ is the regularization parameter, θ j is the jth parameter of the model and n is the total number of parameters.

6. The method for predicting liver cancer recurrence 3 years after surgery according to claim 1, characterized in that: The process of inputting the significantly correlated indicators into the pre-established model for predicting liver cancer recurrence 3 years after surgery, and outputting the trained model for predicting liver cancer recurrence 3 years after surgery: The Logistic regression algorithm is used for construction, where: Divide the data set of indicators with significant correlation into a training sample set and a test sample set; Logistic regression method was used to establish a prediction model for microvascular invasion of liver cancer, and model parameters were set, including regularization strength and whether to perform feature standardization. The training sample set was input into the prediction model of liver cancer recurrence 3 years after surgery to complete the training; Input the test sample set into the prediction model of liver cancer recurrence 3 years after surgery, and output the relevant data of liver cancer recurrence status 3 years after surgery; The receiver operating characteristic (ROC) curve was established based on the data on the recurrence status of liver cancer 3 years after surgery. The pre-established model for predicting liver cancer recurrence 3 years after surgery is as follows: The feature set of the joint model is X, which includes radiomics features and imaging features. The target variable is y, where the binary labels of liver cancer microvascular invasion are 0 and 1. When constructing the joint model, the CatBoost algorithm is used and the hyperparameters are optimized through grid search. The model is expressed as: θ opt is the optimal hyperparameter combination obtained through grid search.

7. The method for predicting liver cancer recurrence 3 years after surgery according to claim 1, characterized in that: The performance of the trained liver cancer recurrence prediction model after surgery 3 years is optimized through cross-validation and independent validation sets, and the effect is evaluated by calculating the AUC value, wherein the calculation process of the AUC value is as follows: Calculate the true positive rate TPR and false positive rate FPR: According to different thresholds, calculate the TPR and FPR under each threshold to obtain a series of points (FPR, TPR). The calculation formulas of TPR and FPR are as follows: TPR: TPR = TP / (TP+FN) FPR: FPR = FP / (FP+TN) Among them, TP represents true positive, FP represents false positive, TN represents true negative, and FN represents false negative; Draw the ROC curve: connect the points to form the ROC curve; Calculate the area: Use the trapezoidal rule to calculate the area under the ROC curve. The formula is: Among them, i represents the i-th point on the ROC curve, FPR i and TPR i They represent the false positive rate and true positive rate of the i-th point respectively, and n is the number of points on the ROC curve.

8. A system for predicting liver cancer recurrence 3 years after surgery, characterized in that: include: A data processing module is used to obtain prediction object related data, pre-process the prediction object related data, and obtain processed prediction object related data, wherein the prediction object related data includes preoperative multi-phase enhanced CT image information, pathological data and liver-related clinical data of the prediction object; The model training module is used to perform feature screening on the processed prediction object related data to obtain indicators with significant correlation, input the indicators with significant correlation into the pre-established liver cancer recurrence prediction model 3 years after surgery, and output the trained liver cancer recurrence prediction model 3 years after surgery; The recurrence prediction module is used to optimize the performance of the trained liver cancer recurrence prediction model 3 years after surgery through cross-validation and independent validation sets, to obtain the optimal liver cancer recurrence prediction model 3 years after surgery, and to achieve liver cancer recurrence prediction 3 years after surgery based on the optimal liver cancer recurrence prediction model 3 years after surgery.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that: The memory stores a computer program that can be run on the processor. When the processor loads and executes the computer program, the method for predicting the recurrence of liver cancer 3 years after surgery according to any one of claims 1 to 7 is adopted.

10. A computer-readable storage medium having a computer program stored therein, characterized in that: When the computer program is loaded and executed by the processor, the method for predicting the recurrence of liver cancer 3 years after surgery according to any one of claims 1 to 7 is adopted.

Images