Method and device for predicting recurrence after HCC local treatment

Abstract

According to the method and device for predicting recurrence after HCC local treatment, multi-time-point peripheral blood lymphocyte subgroups are utilized, an integrated lymphocyte subgroup index is established through related factor screening and model optimization, and clinical information is assisted to establish an immune-clinical combined model to evaluate microwave ablation prognosis of liver cancer patients, so that liver cancer MWA postoperative recurrence and death risks are predicted in the early stage; the dynamic change of the lymphatic subgroup of the patient is longitudinally evaluated for multiple times through the verification queue to serve as input information to optimize the model, and guidance is provided for MWA preoperative planning and prognosis dynamic monitoring. The method comprises the following steps: (1) screening and constructing an accurate prediction model based on a circulating lymphatic subgroup; (2) explaining immune level change characteristics of the patient after ablation through the model, and performing prognosis risk assessment; and (3) verifying the efficacy of the post-treatment lymphatic subgroup dynamic monitoring and prediction model and re-optimizing the model.

Description

Technical Field

[0001] The present invention relates to the technical field of biomedicine, and in particular to a method for predicting recurrence after local treatment of HCC, and a device for predicting recurrence after local treatment of HCC. Background Art

[0002] Hepatocellular carcinoma (HCC) is the fifth most common malignant tumor, with a high incidence globally and a significant impact on patient survival. At present, a variety of technical means are used to treat liver cancer. Among them, microwave ablation therapy (MWA), because it can cause higher lethal temperatures in tumors, a larger necrosis area, a short treatment time, and a significant tumor-killing effect after immune induction, has been recommended by multiple international guidelines as a first-line radical treatment method for early HCC patients. Therefore, this minimally invasive technology with minimal trauma, repeatability, high efficiency, and immune protection has become one of the standard treatment methods for early liver cancer. However, even if MWA can completely inactivate tumors, recurrence after ablation remains a key challenge in the treatment of HCC. The recurrence rate within 5 years is as high as 50%-70%, and the mortality rate approaches 50%. However, due to the heterogeneity of liver cancer and individual differences among patients, the recurrence and death risks of HCC patients after MWA treatment vary greatly. Therefore, dynamic assessment of early prognostic risks after treatment is of great significance for detecting the prognostic status of patients and then assisting in early drug treatment.

[0003] A large number of previous studies have reported risk factors for recurrence after HCC treatment, mainly including tumor size, tumor number, AFP level, etc. However, due to individual differences among patients and tumor heterogeneity, it is difficult to accurately predict recurrence events after HCC ablation and changes in the patient's circulating immunity based on preoperative assessment at a single time point of conventional clinical factors. Moreover, the above clinical information is difficult to indicate the key molecules and potential mechanisms at the cellular level that lead to recurrence and poor prognosis. Therefore, how to accurately and dynamically evaluate the potential mechanisms and possible regulatory targets for recurrence and poor prognosis after microwave ablation in liver cancer patients is an important part of precision treatment of liver cancer. Summary of the Invention

[0004] To overcome the defects of the prior art, the technical problem to be solved by the present invention is to provide a method for predicting recurrence after local treatment of HCC, which uses peripheral blood lymphocyte subsets at multiple time points, establishes an integrated lymph subset index through relevant factor screening and model optimization, and combines clinical information to establish an immune-clinical joint model to evaluate the prognosis of liver cancer patients after microwave ablation, so as to early predict the recurrence and death risks of liver cancer patients after MWA, and further strengthen its longitudinal dynamic clinical verification. By longitudinally and repeatedly evaluating the dynamic changes of the patient's lymph subsets as input information to optimize the model, it provides guidance for preoperative planning and dynamic monitoring of the prognosis of MWA.

[0005] The technical solution of the present invention is: This method for predicting recurrence after local treatment of HCC includes the following steps:

[0006] (1) Screen and construct an accurate prediction model based on circulating lymphocyte subsets;

[0007] (2) Explain the characteristics of changes in immune levels after ablation of patients through the model and conduct prognostic risk assessment;

[0008] (3) Perform dynamic monitoring of lymphocyte subsets after treatment, verify the efficacy of the prediction model, and re-optimize the model.

[0009] The present invention utilizes peripheral blood lymphocyte subsets at multiple time points, establishes an integrated lymphocyte subset index through screening of relevant factors and model optimization, and assists in establishing an immune-clinical combined model to evaluate the prognosis of microwave ablation in liver cancer patients, so as to early predict the recurrence and death risks after MWA in liver cancer patients, and further strengthen its longitudinal dynamic clinical verification. By longitudinally and repeatedly evaluating the dynamic changes of lymphocyte subsets in patients in the verification cohort as input information to optimize the model, it provides guidance for preoperative planning and dynamic monitoring of prognosis of MWA.

[0010] A prediction device for recurrence after local treatment of HCC is also provided, which includes:

[0011] A construction module configured to screen and construct an accurate prediction model based on circulating lymphocyte subsets;

[0012] An interpretation module configured to explain the characteristics of changes in immune levels after ablation of patients through the model and conduct prognostic risk assessment;

[0013] A prediction module configured to perform dynamic monitoring of lymphocyte subsets after treatment, verify the efficacy of the prediction model, and re-optimize the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is a flowchart of a prediction method for recurrence after local treatment of HCC according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0015] Different from traditional treatment methods such as surgical resection, MWA can not only inactivate tumors in situ but also induce an immune activation effect in the tumor in situ and the circulatory system through multiple mechanisms. This immune activation effect can be induced by the production of a series of signaling molecules with immunomodulatory properties (such as cytokines, chemokines, metabolites, and tumor-specific antigens), and the balance between immune-activated cells and immune-suppressive cells determines the stability of the overall immune level in the circulation. Under normal circumstances, the immune system monitors and clears tumor cells to maintain the stability of the internal environment of the body. When the immune function is abnormal, it leads to disorders of cellular immunity and humoral immunity, thus creating conditions for the occurrence and development of tumors. Peripheral blood lymphocyte detection is a convenient and rapid way to evaluate the number and function of circulating lymphocyte subsets. By evaluating the number of multiple immune subsets, such as T cells mainly mediating cellular immune responses, which can directly kill target cells, respond to specific antigens, assist or inhibit B cells to produce antibodies, and produce cytokines, etc., playing a dominant role in the body's immune response; B cells mediate humoral immunity and can differentiate into plasma cells under antigen stimulation to synthesize and secrete antibodies; NK cells are related to innate immunity and participate in immune surveillance by recognizing and killing tumor cells, playing an auxiliary role in the body's immune response. Under normal circumstances, T, B, and NK cells maintain a certain number and ratio, and are stably regulated to maintain the normal immune function of the body. Therefore, by analyzing the circulating immune cell subsets, the immune status of the body can be dynamically understood.

[0016] Previously applied prediction models for HCC prognosis have great application inconvenience, limited predictive value, and cannot provide clues related to the potential mechanism of recurrence because the indicators involved are mostly based on conventional clinical tumor basic characteristic factors, with a wide variety and complexity, long collection time, and cumbersome statistics. In addition, it remains to be investigated whether the conventional prediction model for HCC can achieve the prediction accuracy and stability of HCC microwave ablation patients. Therefore, the core value of this patent is whether circulating peripheral lymphocyte subsets can be used to deeply explore the dynamic changes of immune cells before and after microwave ablation in HCC patients through the interaction and key changes at the immune cell level, and through data modeling, simplify the model tool, and provide model verification and model optimization by combining the dynamic follow-up data of lymph subset changes after treatment, so as to provide a more accurate and effective prediction and verification tool.

[0017] As Figure 1 shown, this prediction method for recurrence after local treatment of HCC includes the following steps:

[0018] (1) Screen and construct an accurate prediction model based on circulating lymph subsets;

[0019] (2) Interpret the characteristics of immune level changes in patients after ablation through the model and conduct prognostic risk assessment;

[0020] (3) Dynamic monitoring of lymphocyte subsets after treatment and validation of the efficacy of the prediction model and re-optimization of the model.

[0021] The present invention uses peripheral blood lymphocyte subsets at multiple time points, establishes an integrated lymphocyte subset index through screening of relevant factors and model optimization, and assists clinical information to establish an immune-clinical combined model to evaluate the prognosis of patients with liver cancer after microwave ablation, so as to early predict the recurrence and death risks after MWA for liver cancer patients, and further strengthen its longitudinal dynamic clinical verification. The dynamic changes of the patient's lymphocyte subsets are longitudinally evaluated multiple times in the validation cohort as input information to optimize the model, providing guidance for preoperative planning and dynamic monitoring of the prognosis of MWA.

[0022] Preferably, the step (1) includes the following sub-steps:

[0023] (1.1) Evaluating the potential prognostic value of circulating immune subsets: Applying restricted cubic splines to perform OS / RFS fitting curve analysis on 15 immune indexes, and finding the optimal

[0024] cut-off value, and analyzing the correlation between immune indexes and prognosis by using the Kaplan-Meier method and log-rank test;

[0025] (1.2) Multivariate screening process and integrated modeling process: After collinearity test, 5 immune characteristics with VIF variance inflation factors all less than 10 are screened out, and further included in LASSO-Cox regression for immune characteristic reduction and screening.

[0026] Preferably, in the step (1.1), the cut-off point selection strategy of restricted cubic splines: The basic principle of restricted cubic splines is to fit polynomials in segments, and a linear term and a spline term are obtained through the following formula,

[0027] Given x and k knots,a restricted cubic spline

[0028] can be defined by

[0029] y=α+x1β1+x2β2+…+x k-1 β k-1

[0030] where

[0031] x1=x

[0032]

[0033] for j=2,...,k-1

[0034]

[0035] Among them, y: the target variable, representing the output of the regression model; α: the intercept term, which is a constant in the model; β1, β2, …, βk - 1: the regression coefficients, which need to be determined by data fitting;

[0036] x1, x2, …, xk - 1: the independent variables after spline transformation, where: x1 = x: the first term is the original independent variable, x2, …, xk - 1: the non - linear transformation terms generated by the cubic spline basis function;

[0037] k: the total number of knots, used for piece - wise fitting; tj: the knot positions, j = 1, 2, …, k;

[0038] (u)+ = max(0, u): the positive part function, defined as taking u when u > 0, otherwise 0; (x - tj) 3 + :

[0039] A component of the cubic spline basis function, used to capture non - linear relationships;

[0040] Then, by weighted summing with the intercept respectively according to the parameters fitted by the model, the fitted curve is obtained. In this way, the best cut - off points are selected for 15 lymphoid subsets, and finally the best cut - off points of all factors are accurately distinguished for prognosis.

[0041] Preferably, in the step (1.2), the implementation strategy of LASSO regression: Lasso regression introduces the L1 regularization term and realizes sparsity by constraining the sum of the absolute values of the parameter vector, enabling Lasso to perform feature selection, estimating many parameters to be zero, thus simplifying the model. Lasso regression tends to select a few key features in high - dimensional data, improving the interpretability of the model.

[0042] The optimization objective of LASSO regression:

[0043]

[0044] Finally, the coefficients of each factor under the lambda.min value are selected, and thus two representative immune subset indicators - Log10CD8+ and Log10Treg are screened. Pearson’s test shows that Log10CD8+ is negatively correlated with Log10Treg, and the correlation coefficient is significant (P = 0.0004). Therefore, the two indicators are combined into a simple calculation formula Log10Treg / Log10CD8+. At the best cut - off, Log10Treg / Log10CD8+ shows a significant correlation with prognosis.

[0045] Preferably, in step (3), the dynamic detection process of lymphatic subsets: subsequent detection is carried out for at least 2 years to achieve long-term follow-up of lymphatic subsets and their change trends in ablated patients; finally, according to the predicted trajectories of changes in key subsets, combined risk early warnings are carried out at the highest value of neutrophils, the highest value of Tregs, the lowest value of cytotoxic CD8 subsets, and the highest value of the combined scoring index to achieve early intervention in the prognosis of patients.

[0046] Preferably, in step (3), immune checkpoint inhibitor drugs are used in combination according to the subset level to achieve dynamic assessment of immune function, risk point warning, and early intervention.

[0047] A prediction device for recurrence after local treatment of HCC is also provided, which includes:

[0048] A construction module configured to screen and construct an accurate prediction model based on circulating lymphatic subsets;

[0049] An interpretation module configured to interpret the characteristics of changes in immune levels after ablation of patients through the model and perform prognostic risk assessment;

[0050] A prediction module configured to perform dynamic monitoring of lymphatic subsets after treatment, verification of the efficacy of the prediction model, and re-optimization of the model.

[0051] Preferably, the construction module performs the following steps:

[0052] (1.1) Evaluate the potential prognostic value of circulating immune subsets: Apply restricted cubic splines to perform OS / RFS fitting curve analysis on 15 immune indicators, find the optimal cut-off value, and use the Kaplan-Meier method and log-rank test to analyze the correlation between immune indicators and prognosis;

[0053] (1.2) Multivariate screening process and integrated modeling process: After collinearity test, 5 immune characteristics with VIF variance inflation factors less than 10 are screened out, and further included in LASSO-Cox regression for immune feature reduction screening.

[0054] Preferably, in step (1.1), the cut-off point selection strategy of restricted cubic splines: The basic principle of restricted cubic splines is to fit polynomials in segments, and a linear term and a spline term are obtained through the following formula,

[0055] Given x and k knots,a restricted cubic spline

[0056] can be defined by

[0057] y=α+x1β1+x2β2+…+x k-1 βk-1

[0058] where

[0059] x1 = x

[0060]

[0061] for j = 2, ..., k - 1

[0062]

[0063] where y: the target variable, representing the output of the regression model; α: the intercept term, which is a constant in the model; β1, β2, …, βk - 1: the regression coefficients, which need to be determined by data fitting;

[0064] x1, x2, …, xk - 1: the independent variables after spline transformation, where: x1 = x: the first term is the original independent variable, x2, …, xk - 1: the non - linear transformation terms generated by the cubic spline basis functions;

[0065] k: the total number of knots, used for piece - wise fitting; tj: the knot positions, j = 1, 2, …, k;

[0066] (u)+ = max(0, u): the positive - part function, defined as taking u when u > 0, otherwise 0; (x - tj) 3 + :

[0067] a component of the cubic spline basis function, used to capture non - linear relationships;

[0068] Then, weighted summation is performed with the intercept respectively according to the parameters fitted by the model to obtain the fitted curve. By this method, the optimal cut - off points are selected for 15 lymphocyte subsets, and finally, the optimal cut - off points of all factors accurately distinguish the prognosis.

[0069] Preferably, in the step (1.2), the implementation strategy of LASSO regression: Lasso regression introduces the L1 regularization term, realizes sparsity by constraining the sum of the absolute values of the parameter vector, enables Lasso to perform feature selection, estimates many parameters to be zero, thus simplifying the model. Lasso regression selects a few key features in high - dimensional data.

[0070] The above are only the preferred embodiments of the present invention, and do not impose any form of limitation on the present invention. Any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A method for predicting recurrence after local treatment of HCC, characterized in that: It includes the following steps: (1) Screen and construct an accurate prediction model based on circulating lymphocyte subsets; (2) Explain the characteristics of the change in immune level after ablation of patients through the model and conduct prognostic risk assessment; (3) Dynamically monitor lymphocyte subsets after treatment, verify the efficacy of the prediction model, and re-optimize the model.

2. The prediction method for recurrence after local treatment of HCC according to claim 1, characterized in that: The step (1) includes the following sub-steps: (1.1) Evaluate the potential prognostic value of circulating immune subsets: Apply restricted cubic splines to perform OS / RFS fitting curve analysis on 15 immune indicators, find the optimal cut-off value, and use the Kaplan-Meier method and log-rank test to analyze the correlation between immune indicators and prognosis; (1.2) Multivariate screening process and integrated modeling process: After collinearity test, 5 immune features with VIF variance inflation factors less than 10 are screened out, and further included in LASSO-Cox regression for immune feature reduction screening.

3. The prediction method for recurrence after local treatment of HCC according to claim 2, characterized in that: In the step (1.1), the cut-off point selection strategy of restricted cubic splines: The basic principle of restricted cubic splines is to piecewise fit polynomials. A linear term and a spline term are obtained through the following formula, Given x and k knots, a restricted cubic spline can be defined by y = α + x1β1 + x2β2 + … + x k-1 β k-1 where x1 = x for j = 2,..., k - 1 where y: the target variable, representing the output of the regression model; α: the intercept term, which is a constant in the model; β1, β2,…, βk - 1: regression coefficients, which need to be determined by data fitting; x1, x2, …, xk-1: independent variables after spline transformation, where: x1 = x: the first term is the original independent variable, x2, …, xk-1: non-linear transformation terms generated by cubic spline basis functions; k: the total number of knots for piecewise fitting; tj: knot positions, j = 1, 2, …, k; (u)+ = max(0, u): positive part function, defined as taking u when u > 0, otherwise 0; (x - tj) 3 + : a component of the cubic spline basis function for capturing non-linear relationships; when weighted and summed with the intercept respectively according to the parameters fitted by the model, the fitted curve is obtained. By this method, the optimal cut-off points are selected for 15 lymphocyte subsets, and finally the optimal cut-off points of all factors are accurately distinguished for prognosis.

4. The prediction method for recurrence after local treatment of HCC according to claim 3, characterized in that: In the step (1.2), the implementation strategy of LASSO regression: Lasso regression introduces an L1 regularization term, realizes sparsity by constraining the sum of the absolute values of the parameter vector, enables Lasso to perform feature selection, estimates many parameters to be zero, thus simplifying the model, and Lasso regression selects a few key features in high-dimensional data.

5. The prediction method for recurrence after local treatment of HCC according to claim 4, characterized in that: In the step (3), the dynamic detection process of lymphocyte subsets: Follow-up detection is carried out for at least 2 years to realize long-term follow-up and change trend of lymphocyte subsets in ablated patients; Finally, according to the change prediction trajectory of key subsets, joint risk high warning is carried out at the highest value of neutrophils, the highest value of Treg, the lowest value of cytotoxic CD8 subsets, and the highest value of the combined scoring index to realize early intervention in patient prognosis.

6. The prediction method for recurrence after local treatment of HCC according to claim 4, characterized in that: In the step (3), immune checkpoint inhibitor drugs are used jointly according to the subset level to realize dynamic assessment of immune function, risk point warning, and early intervention.

7. Predictive device for recurrence after local treatment of HCC, characterized in that: It includes: A construction module configured to screen and construct an accurate prediction model based on circulating lymphocyte subsets; An interpretation module configured to explain the characteristics of the change in immune level after ablation of patients through the model and conduct prognostic risk assessment; A prediction module configured to perform dynamic monitoring of lymphocyte subsets after treatment, verify the efficacy of the prediction model, and re-optimize the model.

8. The prediction device for recurrence after local treatment of HCC according to claim 7, characterized in that: The construction module performs the following steps: (1.1) Evaluating the potential prognostic value of circulating immune subsets: Applying restricted cubic splines to perform OS / RFS fitting curve analysis on 15 immune indicators, and finding the optimal cut-off value. Using the Kaplan-Meier method and log-rank test to analyze the correlation between immune indicators and prognosis; (1.2) Multivariate screening process and integrated modeling process: After collinearity test, 5 immune features with VIF variance inflation factor less than 10 were selected, and further included in LASSO-Cox regression for immune feature reduction screening.

9. The prediction device for recurrence after local treatment of HCC according to claim 8, characterized in that: In the step (1.1), the selection strategy of the breakpoints of the restricted cubic spline: The basic principle of the restricted cubic spline is to fit polynomials piecewise, and a linear term and a spline term are obtained through the following formula, Given x and k knots, a restricted cubic spline can be defined by y = α + x1β1 + x2β2 + … + x k-1 β k-1 where x1 = x for j = 2,..., k - 1 Among them, y: the target variable, representing the output of the regression model; α: the intercept term, which is a constant in the model; β1, β2, …, βk-1: the regression coefficients, which need to be determined by data fitting; x1, x2, …, xk-1: the independent variables after spline transformation, where: x1 = x: the first term is the original independent variable, x2, …, xk-1: the non-linear transformation terms generated by the cubic spline basis function; k: the total number of knots, used for piecewise fitting; tj: the knot positions, j = 1, 2, …, k; (u)+ = max(0, u): the positive part function, defined as taking u when u > 0, otherwise 0; (x - tj) 3 + : a component of the cubic spline basis function, used to capture non-linear relationships; and then weighted summation is performed with the intercept respectively according to the parameters fitted by the model to obtain the fitted curve. By this method, the optimal cut-off points are selected for 15 lymphoid subsets, and finally the optimal cut-off points of all factors are accurately distinguished for prognosis.

10. The prediction device for recurrence after local treatment of HCC according to claim 9, wherein: In the step (1.2), the implementation strategy of LASSO regression: Lasso regression introduces an L1 regularization term to achieve sparsity by constraining the sum of the absolute values of the parameter vector, enabling Lasso to perform feature selection, estimating many parameters to be zero, thus simplifying the model. Lasso regression selects a few key features in high-dimensional data.

Images