Prediction model, construction method and prediction system for liver cancer TACE postoperative active focus risk degree layering

Abstract

The invention discloses a prediction model, a construction method and a prediction system for liver cancer TACE postoperative active lesion risk degree stratification, and belongs to the technical field of molecular biology research and bioinformatics. The method comprises the following steps: firstly, acquiring basic data established by a model, then taking the survival degree of the active lesion as an outcome index in a training set, analyzing and screening independent prediction factors of the survival degree of the active lesion by utilizing single-factor and multi-factor logistic regression, establishing a column graph model, acquiring corresponding coefficients of the independent prediction factors, and finally, calculating the survival degree of the active lesion according to the corresponding coefficients. And constructing an active lesion risk degree total score RSV score of the liver cancer TACE postoperative active lesion risk degree layering prediction model. According to the model constructed by the invention, the survival degree of the liver cancer TACE postoperative active focus can be evaluated, and multi-dimensional verification proves that the model has relatively high reliability.

Description

Technical Field

[0001] The present invention relates to a prediction model, construction method and prediction system for the risk stratification of active foci after transcatheter arterial chemoembolization (TACE) for liver cancer, belonging to the technical field of bioinformatics. Background Art

[0002] Hepatocellular carcinoma (HCC) is the most common type of liver cancer, accounting for 90% of all liver cancer patients. Most HCCs are in the middle and advanced stages at the time of diagnosis, and transcatheter arterial chemoembolization (TACE) is the preferred treatment for intermediate-stage HCC. However, due to the wide heterogeneity of patients and tumors, there are usually residual active foci after TACE. Currently, for the imaging evaluation after TACE, such as the mRECIST criteria or the EASL criteria focus on the measurement of the size of enhanced foci, or the LR-TRA criteria focus on the diagnosis of active foci. However, different active foci often have different prognostic outcomes in clinical practice, indicating that there are different invasiveness or risks for active foci. Therefore, performing risk stratification of active foci has high guiding significance for guiding the subsequent treatment decisions of clinical patients and predicting the prognosis of patients.

[0003] In previous studies, in order to provide clinical guidance and assistance to TACE patients, many researchers focused on the prediction of TACE treatment response and developed many prediction models for TACE treatment response using a series of preoperative clinical laboratory indicators, imaging semantic signs or high-throughput radiomics features. However, there are more or less certain problems in the model generalization and guiding clinical practice. There is currently no research on the active foci after TACE. Previous studies have low guiding value for the active foci after TACE, and predicting the postoperative situation using preoperative data usually brings low credibility and interpretability. Summary of the Invention

[0004] The present invention provides a prediction model, construction method and prediction system for the risk stratification of active foci after TACE for liver cancer. By analyzing the multi-sequence and multi-angle MRI imaging signs of active foci after TACE, the present invention constructs a prediction model for the risk stratification of active foci and conducts multi-dimensional verification of the model in terms of prognosis and pathological results, realizing the assessment of the survival and risk of active foci, and solving the blank in previous research on active foci and the problem of low credibility of preoperative prediction of postoperative situation.

[0005] The technical solution of the present invention is as follows:

[0006] The construction method of the prediction model for the risk stratification of active foci after TACE for liver cancer of the present invention includes the following steps:

[0007] S1. Obtain MR imaging signs of active foci after TACE through postoperative follow-up MRI of liver cancer patients, and divide the active foci into high- and low-survival active foci according to the evaluation results of the second follow-up MRI;

[0008] S2. Taking the high or low survival of the active foci as the outcome index, analyze the MR imaging signs of the active foci in S1, screen the independent predictors of the survival of the active foci, and construct a prediction model for the risk stratification of active foci after TACE for liver cancer, the total risk score of the active foci RSV score.

[0009] Furthermore, in S1, the MR imaging signs of the active foci include: mild to moderate T1 hypointensity, mild to moderate T2 hyperintensity, T2 peritumoral hyperintensity, diffusion restriction, intratumoral contrast enhancement, peritumoral contrast enhancement, non-peripheral "washout", peritumoral hyperenhancement, peritumoral hypoenhancement in the portal venous phase, irregular shape, internal location, heterogeneity of the active foci, non-smooth tumor margin, size, number of active foci, and morphology.

[0010] Furthermore, in S1, the evaluation scheme for the survival of the active foci is: based on the evaluation of the second follow-up MRI images of the patients by LR-TRA v2024, define that the survival of the active foci with the evaluation result still being viable is high, and the survival of the active foci with the evaluation result turning into non-viable or equivocal is low.

[0011] Furthermore, in S2, transform the MR imaging signs into 15 binary variables and 1 four-category variable.

[0012] Furthermore, in S2, use univariate and multivariate logistic regression analysis to screen the independent predictors of the survival of the active foci, and finally screen out 5 signs, including mild to moderate T2 hyperintensity, T2 peritumoral hyperintensity, diffusion restriction, irregular shape, and heterogeneity of the active foci.

[0013] Furthermore, construct a prediction model for the risk stratification of active foci after TACE for liver cancer, the total risk score of the active foci RSV score = 8 × mild to moderate T2 hyperintensity + 10 × T2 peritumoral hyperintensity + 9 × diffusion restriction + 7 × irregular shape + 7 × heterogeneity of the active foci.

[0014] The present invention also provides a prediction model constructed by the above construction method.

[0015] Furthermore, the present invention provides a prediction system for the risk stratification of active foci after TACE for liver cancer, including:

[0016] Data acquisition module: It is used to acquire and analyze the MR imaging signs of the first follow-up after TACE for liver cancer patients, and transform them into 5 binary variables, including: mild to moderate T2 hyperintensity, peritumoral T2 hyperintensity, diffusion restriction, irregular shape, and heterogeneity of active foci;

[0017] Model scoring module: It uses the data of mild to moderate T2 hyperintensity, peritumoral T2 hyperintensity, diffusion restriction, irregular shape, and heterogeneity of active foci obtained by the data acquisition module as input data, provides them to the above prediction model, and obtains the RSVscore; specifically: RSV score = 8 × mild to moderate T2 hyperintensity + 10 × peritumoral T2 hyperintensity + 9 × diffusion restriction + 7 × irregular shape + 7 × heterogeneity of active foci.

[0018] Result output module: It is used to obtain the RSV score in the model scoring module and output the prediction result, compare the RSVscore with the threshold, if the risk value is greater than or equal to the threshold, the prediction result is output as high risk, and if the risk value is less than the threshold, the prediction result is output as low risk.

[0019] Furthermore, the threshold is set to 25, and it is defined that the total score of the active focus risk degree ≥ 25 is high risk, and the total score of the active focus risk degree < 25 is low risk.

[0020] Compared with the prior art, the present invention has the following advantages:

[0021] The present invention combines multi-sequence and multi-angle MRI imaging signs of active foci after TACE, uses univariate and multivariate logistic regression analysis to screen independent predictive factors for the survival degree of active foci and establishes a nomogram model, and conducts multi-dimensional verification of the model in prognosis and pathological results. Compared with the low-risk group, the high-risk group separated by the prediction model of the present invention has a shorter progression-free survival (PFS) and overall survival (OS). In addition, through pathological correlation analysis, it is found that compared with the low-risk group, the high-risk group separated by the model has a higher correlation with microvascular invasion and liver capsule invasion. The present invention focuses on the active foci and their imaging signs after TACE, realizes the evaluation of the survival degree and risk degree of active foci, and the credibility and reliability of the model are higher, solving the problems of subsequent treatment decision-making for clinical patients and predicting the prognosis of patients. Description of the drawings

[0022] Figure 1 It is the scoring model for predicting the risk degree stratification of active foci of the present invention.

[0023] Figure 2 It is the ROC curve for predicting the risk degree stratification of active foci of the present invention, where A is the training data set and B is the external test data set.

[0024] Figure 3This is the decision curve and calibration curve graph for predicting the risk stratification of active foci in the present invention. Among them, A - B is the decision curve (DCA) graph; C - D is the calibration curve graph.

[0025] Figure 4 This is the K - M curve of PFS and OS of the risk stratification model for predicting active foci in the present invention in the external test set. Among them, A is PFS, and B is OS at 2 years.

[0026] Figure 5 This is the correlation graph between the risk stratification model for predicting active foci in the present invention and pathology. Among them, A is the correlation between high - and low - risk active foci and the grading of microvascular invasion, and B is the correlation between high - and low - risk active foci and the invasion of the liver capsule. Detailed implementation manners

[0027] The following combines specific embodiments to clearly and completely describe the technical solutions of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present invention.

[0028] The test methods used in the following embodiments are all conventional methods without special instructions; the materials, reagents, etc. used are all reagents and materials that can be obtained from commercial channels without special instructions.

[0029] Example 1

[0030] 1. Case selection

[0031] Patient inclusion criteria: Clinically and radiologically diagnosed with hepatocellular carcinoma; undergoing TACE treatment for the first time; performing MRI plain scan + enhanced follow - up examinations at 1 month and 6 months after the operation; the MRI images at 1 month are evaluated as viable lesions according to LR - TRA (LI - RADS Treatment Response Algorithm) v2024.

[0032] 2. Acquisition and analysis of data

[0033] Collect and analyze the clinical data, MRI image signs, PFS, overall survival OS, and pathological report data of liver cancer patients.

[0034] (1) The clinical data of the patients can be obtained by consulting the electronic medical records, including gender, age, history of diabetes, history of hypertension, history of hepatitis, whether they received secondary treatment during the follow-up period, tumor markers, and clinical routine laboratory test indicators. And all variables were transformed into binary variables according to the normal critical values or medians. Clinical routine laboratory test indicators include Hepatitis B, Hepatitis C, Leukocyte (×10 9 / L), Hemoglobin (g / L), Platelet (×10 9 / L), Total bilirubin (μmol / L), Direct bilirubin (μmol / L), Albumin (g / L), Alanine aminotransferase (U / L), Aspartate aminotransferase (U / L), Alkaline phosphatase (U / L), Serum creatinine (μmol / L), Prothrombin time (s), International normalized ratio, Alpha-fetoprotein (μg / L).

[0035] (2) Analysis of the imaging signs characteristics of the viable foci: Analyze the MRI imaging signs of the patients 1 month after TACE, and transform them into 15 binary variables and 1 four-category variable. A total of 16 imaging signs were included in the analysis, including: mild to moderate T1 hypointensity (yes / no), mild to moderate T2 hyperintensity (yes / no), T2 peritumoral hyperintensity (yes / no), diffusion restriction (yes / no), intratumoral contrast enhancement (yes / no), peritumoral contrast enhancement (yes / no), non-peripheral "washout" (yes / no), peritumoral high enhancement (yes / no), peritumoral low enhancement in the portal venous phase (yes / no), irregular shape (yes / no), internal location (yes / no), heterogeneity of the viable foci (yes / no), non-smooth tumor margin (yes / no), size (>3 cm / ≤3 cm), number of viable foci (single / multiple); morphology (encapsulated nodular type, single nodular type, fused multinodular type, infiltrative type). Except for the morphology variable which is a four-category variable, all the other variables were transformed into binary variables.

[0036] 3. Analysis of the viability of the viable foci

[0037] Evaluate the MRI images of liver cancer patients at 1 month and 6 months after TACE based on the LR-TRA v2024 standard to obtain the viability of viable foci. Specifically: Evaluate the MRI images of liver cancer patients at 1 month after TACE according to LR-TRA v2024, and select the foci with the evaluation result of viable as viable foci; then evaluate the MRI images of liver cancer patients at 6 months after TACE according to LR-TRA v2024, and define that the viability of viable foci with the evaluation result still being viable at 6 months is high, and the viability of viable foci with the evaluation result turning into non-viable or equivocal at 6 months is low.

[0038] 4. Screen independent predictors of the viability of viable foci

[0039] Using the above collected data as the training set data, with the viability of viable foci as the outcome index, use univariate and multivariate logistic regression analysis to screen independent predictors of the viability of viable foci. As shown in Table 1, finally 5 signs are screened out, including mild to moderate T2 hyperintensity, peritumoral T2 hyperintensity, diffusion restriction, irregular shape, and viable focus heterogeneity.

[0040] Table 1 Independent predictors of the viability of viable foci screened by univariate and multivariate logistic regression

[0041]

[0042]

[0043] Note: CI, Confidence interval; OR, Odds Ratio.

[0044] 5. Construct a nomogram model

[0045] Construct a nomogram model according to the results of multivariate logistic regression analysis. Substitute the 5 independent predictors obtained in S3 into the nomogram model to obtain the corresponding coefficients of each independent predictor, and construct a prediction model for the risk stratification of viable foci after liver cancer TACE, the total risk score of viable foci RSV score. The total risk score of viable foci RSV score is the sum of the 5 independent predictors weighted by the corresponding coefficients. Specifically: RSV score = 8 × mild to moderate T2 hyperintensity + 10 × peritumoral T2 hyperintensity + 9 × diffusion restriction + 7 × irregular shape + 7 × viable focus heterogeneity.

[0046] Stratify the risk of viable foci according to the total risk score of viable foci RSV score. High risk is defined as a total score ≥ 25, and low risk is defined as < 25, as Figure 1 shown.

[0047] In this embodiment, the R 3.6.1 software was used to perform graphical visualization analysis on the model, including ROC, decision curve, and calibration curve, as Figures 2-3 shown.

[0048] In this embodiment, 203 active foci were listed as the training set and 102 active foci in a multi-center clinical trial dataset were used as the test set. Patient inclusion criteria: Clinically and radiologically diagnosed with hepatocellular carcinoma; underwent TACE treatment for the first time; underwent MRI plain scan + enhanced follow-up examinations at 1 month and 6 months after the operation; the 1-month imaging was evaluated as viable lesions according to LR-TRA v2024. In addition, survival was verified in the test set, including PFS and OS, as Figure 4 shown.

[0049] In the implementation case of this case, another 120 active foci were used as an independent pathological verification set. Patient inclusion criteria: The patient underwent hepatectomy after the first TACE treatment; MRI plain scan + enhanced examination was performed after TACE and before hepatectomy; the MR imaging was evaluated as viable lesions according to LR-TRAv2024. Patient exclusion criteria: Underwent TACE treatment two or more times before hepatectomy; pathological results confirmed the absence of active lesions or non-HCC. The RSV score of the active foci was calculated in this dataset. A total score of ≥25 was considered high risk, and <25 was considered low risk. The correlation between high- and low-risk active foci and pathological signs and immunohistochemical indicators was analyzed. As Figure 5 shown, high- and low-risk active foci were significantly correlated with the grade of microvascular invasion and capsular invasion of the liver.

[0050] In summary, the present invention utilizes the molecular biological information of the imaging signs of active foci, integrates multi-sequence and multi-angle active focus imaging data, constructs a risk stratification model for active foci after TACE for hepatocellular carcinoma, realizes the evaluation of the viability and risk of active foci, and provides a more reliable tool for subsequent clinical diagnosis and treatment.

[0051] In addition, it should be understood that although this specification is described according to embodiments, not every embodiment only contains an independent technical solution. This narrative manner of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for constructing a prediction model for the risk stratification of active foci after TACE for liver cancer, characterized in that, It includes the following steps: S1. Obtain the MR imaging signs of active lesions after TACE through postoperative follow-up MRI of liver cancer patients, and divide the active lesions into high- and low-viability active lesions according to the evaluation results of the second follow-up MRI; S2. Taking the viability level of the active lesions as the outcome index, analyze the MR imaging signs of the active lesions in S1, screen the independent predictors of the viability of the active lesions, and construct a prediction model for the risk stratification of active lesions after TACE in liver cancer, namely the total RSV score of the risk of active lesions.

2. The construction method according to claim 1, characterized in that, In S1, the MR imaging signs of the active lesions include: mild to moderate T1 hypointensity, mild to moderate T2 hyperintensity, peritumoral T2 hyperintensity, diffusion restriction, intratumoral contrast enhancement, peritumoral contrast enhancement, non-peripheral "washout", peritumoral hyperenhancement, peritumoral hypoenhancement in the portal venous phase, irregular shape, internal location, heterogeneity of the active lesions, non-smooth tumor margin, size, number of active lesions, and morphology.

3. The construction method according to claim 1, characterized in that, In S1, the evaluation scheme for the viability of the active lesions is as follows: According to the LR-TRA v2024, evaluate the second follow-up MRI images of the patients. Define that the viability of the active lesions with the evaluation result still being viable is high, and the viability of the active lesions with the evaluation result changing to non-viable or equivocal is low.

4. The construction method according to claim 1, characterized in that In S2, transform the MR imaging signs into 15 binary variables and 1 four-category variable.

5. The construction method according to claim 1, characterized in that In S2, use univariate and multivariate logistic regression analyses to screen the independent predictors of the viability of the active lesions. Finally, 5 signs are selected, including mild to moderate T2 hyperintensity, peritumoral T2 hyperintensity, diffusion restriction, irregular shape, and heterogeneity of the active lesions.

6. A prediction model constructed by the construction method according to any one of claims 1 to 5.

7. The prediction model according to claim 6, wherein The total RSV score of the risk of active lesions = 8 × mild to moderate T2 hyperintensity + 10 × peritumoral T2 hyperintensity + 9 × diffusion restriction + 7 × irregular shape + 7 × heterogeneity of the active lesions.

8. A prediction system for risk stratification of active foci after liver cancer TACE, characterized in that, It includes: A data acquisition module: used to obtain and analyze the MR imaging signs of the first follow-up after TACE in liver cancer patients, and transform them into 5 binary variables, including: mild to moderate T2 hyperintensity, peritumoral T2 hyperintensity, diffusion restriction, irregular shape, and heterogeneity of the active lesions; A model scoring module: using the data of mild to moderate T2 hyperintensity, peritumoral T2 hyperintensity, diffusion restriction, irregular shape, and heterogeneity of the active lesions obtained by the data acquisition module as input data, and providing them to the above prediction model to obtain the RSV score; specifically: RSV score = 8 × mild to moderate T2 hyperintensity + 10 × peritumoral T2 hyperintensity + 9 × diffusion restriction + 7 × irregular shape + 7 × heterogeneity of the active lesions; A result output module, used to obtain the RSV score in the model scoring module and output the prediction result, compare the RSV score with the threshold. If the risk value is greater than or equal to the threshold, output the prediction result as high risk; if the risk value is less than the threshold, output the prediction result as low risk.

9. The prediction system according to claim 8, wherein Set the threshold to 25, and define that the total RSV score of the risk of active lesions ≥ 25 is high risk, and the total RSV score of the risk of active lesions < 25 is low risk.

Images