Clinical new auxiliary curative effect prediction system for liver cancer and medium

Abstract

The invention discloses a liver cancer clinical new auxiliary curative effect prediction system and medium, and the prediction system comprises an input unit which is used for inputting the value of a diagnostic serological detection index of a patient; the prediction unit is used for generating a prediction result of whether the new adjuvant therapy is invalid to the patient or not based on the prediction model and the diagnostic serological detection indexes; and the output unit is used for outputting the prediction result. According to the present invention, the diagnosis serological detection index of the patient can be conveniently and rapidly obtained based on the detection results of the blood biochemical detection, the specific protein detection and the like during the liver cancer treatment process so as to predict whether the new adjuvant therapy of the tiranibizumab combined with the lenvatinib is invalid to the patient based on the diagnosis serological detection index; therefore, the delay of the operation process of a part of patients with invalid new adjuvant therapy is avoided, and the method has wide application value clinically.

Description

Technical Field

[0001] This invention relates to the field of neoadjuvant therapy for tumors, specifically to a clinical neoadjuvant therapy efficacy prediction system and medium for liver cancer. Background Technology

[0002] Neoadjuvant therapy refers to systemic or localized anti-tumor treatment administered before primary treatments aimed at curing liver cancer, such as surgical resection or liver transplantation. However, while the value of neoadjuvant therapy in liver cancer is recognized, it also carries the risk of delaying potentially curative surgery. For patients who are otherwise eligible for surgery, delays or even the inability to undergo surgery due to their insensitivity to neoadjuvant therapy pose a significant clinical concern.

[0003] Therefore, some scholars have attempted to predict the radiographic or pathological response to perioperative immunotherapy using biomarkers. However, related studies mainly focus on the correlation analysis between treatment efficacy and biomarkers in resected tumor samples. This approach often makes it difficult to develop personalized treatment plans and is not conducive to perioperative decision-making. Among existing technologies, the research by Xia et al. showed that dendritic cell infiltration in the tumor microenvironment is a predictive biomarker for the perioperative response to caberizumab combined with apatinib in patients with resectable hepatocellular carcinoma. However, such biomarkers rely on invasive detection methods and need to be independent of routine liver cancer examination procedures; the detection methods for some biomarkers are quite complex. Summary of the Invention

[0004] One objective of this invention is to provide a clinical neoadjuvant efficacy prediction system for liver cancer. Based on a trained TabPFN model, this system can easily, non-invasively, and efficiently screen patients who are unresponsive to tisraizumab combined with lenvatinib in the preoperative stage after inputting preoperative hematological indicators. This avoids delaying surgery for these patients due to neoadjuvant therapy and effectively improves the treatment outcome for liver cancer patients.

[0005] This invention is achieved through the following technical solution:

[0006] A neoadjuvant therapy prediction system for liver cancer includes:

[0007] The acquisition unit is used to acquire several serological test indicators of several liver cancer patients before neoadjuvant therapy.

[0008] The training set construction unit is used to randomly combine at least two of the several serological test indicators to obtain several serological test indicator groups. Based on the serological test indicator groups and the corresponding neoadjuvant therapy efficacy for patients, several training sets are constructed. The neoadjuvant therapy refers to the treatment method of using tislelizumab combined with lenvatinib before surgery.

[0009] The training unit is used to train the model using the plurality of training sets to obtain a plurality of trained models, test the trained models to obtain test results, determine the prediction model based on the test results, and determine the test serological detection indicators in the training set corresponding to the prediction model as the diagnostic serological detection indicators to be input.

[0010] The input unit is used to input the values ​​of the patient's diagnostic serological test indicators;

[0011] The prediction unit is used to generate a prediction result on whether neoadjuvant therapy is ineffective for the patient based on the prediction model and the diagnostic serological test indicators.

[0012] The output unit is used to output the prediction result.

[0013] In this technical solution, the acquisition unit of the prediction system is used to acquire several serological test indicators for liver cancer patients before neoadjuvant therapy. The serological test indicators are preferentially selected based on statistical significance and their ability to contribute practically to subsequent model training.

[0014] In this technical solution, after obtaining the serological indicators for testing, several sets of serological indicator groups are obtained through several random combinations. Each set of serological indicator groups includes at least two serological indicators. A training set is constructed based on each set of serological indicator groups and the corresponding efficacy of neoadjuvant therapy for patients, resulting in several training sets. Each training set includes the neoadjuvant therapy efficacy results for several patients, as well as at least two serological indicators for these patients.

[0015] In this technical solution, several training sets are constructed to train the model, resulting in several trained models. The trained models can then be validated using a test set. In a preferred embodiment, the test set uses retrospective cohort data as an independent internal validation set, which includes patient information, serological test indicators for each patient, and the effectiveness of neoadjuvant therapy. Finally, the trained models are evaluated by comparing the accuracy (ACC), area under the ROC curve (AUC), and mean precision (AP) of each trained model during validation, and the model with the best performance is determined as the predictive model.

[0016] Once the prediction model is determined, it can be used in the prediction unit. The various serological indicators in the test serological indicator group corresponding to the prediction model will also serve as the values ​​of the diagnostic serological indicators to be input. For example, when the model trained on C-reactive protein, total bilirubin, and gamma-glutamyl transferase is determined as the prediction model, these three indicators will be used as diagnostic serological indicators.

[0017] After obtaining the prediction model, the effectiveness of neoadjuvant therapy for patients can be predicted based on the input values ​​of diagnostic serological test indicators of the patients to be tested and the prediction model. The output prediction result is either that neoadjuvant therapy is ineffective for the patients or that neoadjuvant therapy is effective for the patients.

[0018] This predictive system can conveniently and quickly obtain the patient's diagnostic serological indicators based on the results of blood biochemistry tests and specific protein tests during liver cancer treatment. Based on these diagnostic serological indicators, it can predict whether neoadjuvant therapy with tisraizumab combined with lenvatinib will be ineffective for the patient, thereby avoiding delays in the surgical process for some patients who have not responded to neoadjuvant therapy. It has broad clinical application value.

[0019] Furthermore, the diagnostic serological test indicators consist of lactate dehydrogenase isoenzyme, gamma-glutamyl transferase, uric acid, gamma globulin, total bilirubin, and C-reactive protein.

[0020] In this technical solution, testing revealed that the predictive model trained using a set of serological test indicators consisting of six indicators—lactate dehydrogenase isoenzyme (LDH-ISO), gamma-glutamyl transferase (γ-GT), uric acid (UA), gamma globulin (γ-G), total bilirubin (TB), and C-reactive protein (CRP-C)—exhibited high predictive ability on the test set, with an AUC of 0.917, an ACC of 0.778, and an AP of 0.966. Therefore, this technical solution uses these six indicators as diagnostic serological test indicators. The results for all six indicators can be obtained directly from one or more test reports in the liver cancer screening process, without the need for invasive testing methods or tests separate from the routine liver cancer screening procedure.

[0021] In a preferred embodiment of the present invention, after obtaining the test serological indicators corresponding to the prediction model, the training unit uses linear discriminant analysis to screen one or more test serological indicators corresponding to the prediction model as diagnostic serological indicators to be input, based on the test serological indicators corresponding to the prediction model and the corresponding neoadjuvant therapy effectiveness of the patients, and constructs a linear discriminant formula based on the diagnostic serological indicators.

[0022] The prediction unit is used to generate a prediction result on whether neoadjuvant therapy is ineffective for the patient based on the diagnostic serological test indicators and the linear discriminant.

[0023] In this technical solution, after determining the test serological indicators corresponding to the prediction model, they are not directly used as the diagnostic serological indicators to be input. Instead, linear discriminant analysis is used to process the test serological indicators corresponding to the prediction model.

[0024] Specifically, linear discriminant analysis was used to analyze the synergistic effects and mutual influences among various serological test indicators, and several linear discriminant formulas were constructed. Based on the test results of the test set, the linear discriminant formula with the best performance was selected as the final linear discriminant formula used for prediction. At the same time, its corresponding serological test indicator was used as the diagnostic serological test indicator. This approach aims to reduce the number of diagnostic serological test indicators without significantly affecting the prediction results, thereby further improving the efficiency of prediction.

[0025] Furthermore, the diagnostic serological test indicators consist of lactate dehydrogenase isoenzyme, uric acid, and γ-globulin.

[0026] In this technical solution, linear discriminant analysis revealed that although γ-glutamyltransferase has the second-highest weight in the prediction model, lactate dehydrogenase isoenzyme, uric acid, and γ-globulin showed stronger interactions in the optimal linear discriminant. The linear discriminant constructed based on these three indicators achieved an AUC of 0.933 on the test set and an AUC of 0.913 on the training set.

[0027] Furthermore, in the linear discriminant expression, the discriminant score... Where G is the value of γ-globulin, UA is the value of uric acid, and LDH_ISO is the value of lactate dehydrogenase isoenzyme;

[0028] When the discrimination score D is greater than 0.3819, neoadjuvant therapy is ineffective for the patient;

[0029] Neoadjuvant therapy is effective for patients when the discrimination score D is less than or equal to 0.3819.

[0030] In this technical protocol, gamma globulin values ​​are dimensionless percentages, uric acid values ​​are in μmol / L, and lactate dehydrogenase isoenzyme values ​​are in U / L. Inconsistent units do not affect the discriminant score D. When using this protocol, input the patient's lactate dehydrogenase isoenzyme, uric acid, and gamma globulin values ​​to calculate the discriminant score D. If the discriminant score D is greater than the cutoff value of 0.3819, neoadjuvant therapy is considered ineffective, and surgery should be performed as soon as possible; otherwise, neoadjuvant therapy with tislelizumab combined with lenvatinib is recommended.

[0031] Furthermore, the acquisition unit is used to acquire several raw serological test indicators, process the several raw serological test indicators using at least one evaluation method among the F-statistic method, mutual information method, and random forest method, and screen out the several test serological test indicators.

[0032] In this technical solution, when determining the serological test indicators, in order to avoid missing important indicators, all or most of the original serological test indicators that are statistically significant can be systematically collected from liver cancer patients before the start of neoadjuvant therapy.

[0033] Furthermore, the training unit uses the TabPFN model. In medical research, especially in the study of specific new therapies, the sample size is often limited. In this technical solution, to address the problem of limited sample size, the TabPFN model, which has high generalization ability in small sample scenarios, is selected as the training model. This effectively addresses the overfitting risk and class imbalance problems commonly encountered in small sample training, ensuring that the model maintains robust and reliable performance even with limited data.

[0034] Another object of the present invention is to provide a storage medium based on a predictive model, comprising a stored computer program, wherein, when the computer program is executed, the device in which the storage medium is located performs the following steps:

[0035] The diagnostic serological indicators of the patient are obtained, which consist of lactate dehydrogenase isoenzyme, gamma-glutamyl transferase, uric acid, gamma globulin, total bilirubin, and C-reactive protein.

[0036] Based on the diagnostic serological test indicators and prediction model, a prediction result of whether neoadjuvant therapy is ineffective for the patient is generated and the prediction result is output. The neoadjuvant therapy refers to the treatment method of using tislelizumab combined with lenvatinib before surgery.

[0037] The method for constructing the prediction model includes the following steps:

[0038] To obtain several serological test indicators from several liver cancer patients before neoadjuvant therapy;

[0039] At least two of the serological test indicators are randomly combined to obtain several serological test indicator groups. Based on the serological test indicator groups and the corresponding neoadjuvant therapy efficacy in patients, several training sets are constructed.

[0040] The model is trained using the aforementioned training sets to obtain several trained models. The trained models are then tested to obtain test results. Based on the test results, the prediction model is determined. The test serological detection indicators in the training set corresponding to the prediction model are used as diagnostic serological detection indicators.

[0041] Another object of the present invention is to provide a linear discriminant-based storage medium, comprising a stored computer program, wherein, when the computer program is executed, the device containing the storage medium controls the following steps to be performed:

[0042] The diagnostic serological indicators of the patient are obtained, which consist of lactate dehydrogenase isoenzyme, uric acid, and gamma globulin;

[0043] Based on the diagnostic serological test indicators and linear discriminant formula, a predictive result is generated as to whether neoadjuvant therapy is ineffective for the patient. Neoadjuvant therapy refers to the treatment method of using tisraizumab combined with lenvatinib before surgery.

[0044] The method for constructing the linear discriminant includes the following steps:

[0045] To obtain several serological test indicators from several liver cancer patients before neoadjuvant therapy;

[0046] At least two of the serological test indicators are randomly combined to obtain several serological test indicator groups. Based on the serological test indicator groups and the corresponding neoadjuvant therapy efficacy in patients, several training sets are constructed.

[0047] Several training sets are used to train models, resulting in several trained models. These trained models are then tested to obtain test results. Based on the test results, a prediction model is determined. Based on the test serological indicators corresponding to the prediction model and the corresponding neoadjuvant therapy effectiveness in patients, linear discriminant analysis is used to select one or more test serological indicators corresponding to the prediction model as diagnostic serological indicators to be input. A linear discriminant expression based on these diagnostic serological indicators is then constructed, wherein the discriminant score in the linear discriminant expression... Where G is the value of γ-globulin, UA is the value of uric acid, and LDH_ISO is the value of lactate dehydrogenase isoenzyme; when the discriminant score D is greater than 0.3819, neoadjuvant therapy is ineffective for the patient; when the discriminant score D is less than or equal to 0.3819, neoadjuvant therapy is effective for the patient.

[0048] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0049] 1. This invention can conveniently and quickly obtain the diagnostic serological test indicators of patients based on the results of blood biochemistry tests and specific protein tests during the treatment of liver cancer. Based on the diagnostic serological test indicators, it can predict whether neoadjuvant therapy of tisraizumab combined with lenvatinib is ineffective for the patient, thereby avoiding delays in the surgical process of some patients who have not responded to neoadjuvant therapy. It has a wide range of clinical application value.

[0050] 2. This invention can use linear discriminant analysis to further screen out fewer diagnostic serological indicators, thereby reducing the number of diagnostic serological indicators without significantly affecting the prediction results, and further improving the prediction efficiency. Attached Figure Description

[0051] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:

[0052] Figure 1 The AUC, ACC, and AP of the prediction model in a specific embodiment of the present invention are shown in the training and test sets.

[0053] Figure 2 This illustrates the combined ranking of six diagnostic serological test indicators in a specific embodiment of the present invention based on F-statistic, MI value, and random forest importance.

[0054] Figure 3 This illustrates the weight ranking of six diagnostic serological test indicators in the prediction model in a specific embodiment of the present invention;

[0055] Figure 4 The ROC curves of the linear discriminant constructed from three diagnostic serological test indicators in a specific embodiment of the present invention are shown in the test set and training set. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0057] All raw materials used in this invention are not particularly limited in their source; they can be purchased commercially or prepared using conventional methods well-known to those skilled in the art. The purity of all raw materials used in this invention is not particularly limited; however, analytical grade or conventional purity requirements in the analytical field are preferred. All raw materials used in this invention have common designations and abbreviations in the art, and each designation and abbreviation is clearly defined within its relevant application. Those skilled in the art can obtain them commercially available or prepare them using conventional methods based on the designation, abbreviation, and corresponding application.

[0058]

Example 1

[0059] A neoadjuvant therapy prediction system for liver cancer includes:

[0060] The acquisition unit is used to acquire several serological test indicators of several liver cancer patients before neoadjuvant therapy.

[0061] The training set construction unit is used to randomly combine at least two of the several serological test indicators to obtain several serological test indicator groups. Based on the serological test indicator groups and the corresponding neoadjuvant therapy efficacy for patients, several training sets are constructed. The neoadjuvant therapy refers to the treatment method of using tislelizumab combined with lenvatinib before surgery.

[0062] The training unit is used to train the model using the plurality of training sets to obtain a plurality of trained models, test the trained models to obtain test results, determine the prediction model based on the test results, and determine the test serological detection indicators in the training set corresponding to the prediction model as the diagnostic serological detection indicators to be input.

[0063] The input unit is used to input the values ​​of the patient's diagnostic serological test indicators;

[0064] The prediction unit is used to generate a prediction result on whether neoadjuvant therapy is ineffective for the patient based on the prediction model and the diagnostic serological test indicators.

[0065] The output unit is used to output the prediction result.

[0066] In some embodiments, the number of serological indicators tested is no less than 50 to avoid missing important indicators.

[0067] In some embodiments, the number of random combinations is no less than 1000 to systematically explore the predictive performance of different test serological indicator groups.

[0068] In a preferred embodiment, the test set uses retrospective cohort data from Tianjin Cancer Hospital as an independent internal validation set, which includes patient information, as well as each patient's test serological indicators and neoadjuvant therapy efficacy.

[0069] In some embodiments, the acquisition unit is used to acquire several raw serological test indicators, and processes these raw serological test indicators using at least one evaluation method selected from the F-statistic method, mutual information method, and random forest method to screen out several test serological test indicators. For example, in some preferred embodiments, 119 raw serological test indicators are screened. Then, test serological test indicators are obtained by further screening using one or more existing methods such as the F-statistic method, mutual information method, and random forest method.

[0070] In some embodiments, the training unit uses the TabPFN model. In one or more embodiments, other models suitable for small sample sizes may also be used as the training model. In some embodiments, as the number of samples available for the training set increases, other non-small sample models may also be used.

[0071]

Example 2

[0072] Based on Example 1, after obtaining the test serological indicators corresponding to the prediction model, the training unit uses linear discriminant analysis to screen one or more test serological indicators corresponding to the prediction model as diagnostic serological indicators to be input, based on the test serological indicators corresponding to the prediction model and the corresponding neoadjuvant therapy effectiveness of the patient, and constructs a linear discriminant expression based on the diagnostic serological indicators; the prediction unit is used to generate a prediction result on whether neoadjuvant therapy is ineffective for the patient based on the diagnostic serological indicators and the linear discriminant expression.

[0073] In some preferred embodiments, the diagnostic serological test indicators consist of lactate dehydrogenase isoenzyme, uric acid, and γ-globulin.

[0074]

Example 3

[0075] Based on the foregoing embodiments, this embodiment provides a more preferred prediction system.

[0076] Specifically, we first systematically acquired 119 original serological test indicators from liver cancer patients before the start of neoadjuvant therapy. These indicators were all routine clinical tests, ensuring the accessibility and generalizability of the data. Subsequently, we used three different indicator evaluation methods—the F-statistic method, the mutual information method, and the random forest method—to select the top 50 core indicators with the highest predictive value from the 119 original serological test indicators. These were then used as test serological test indicators, ensuring that the selected features were both statistically significant and made practical contributions to the model.

[0077] After identifying 50 serological indicators for testing, 1000 sets of serological indicator groups were obtained through 1000 random combinations. A prospective clinical study cohort (trial registration number: NCT04834986) registered in the US Clinical Trials Database by the inventors' team was used to construct 1000 training sets. Each training set included patient information, the therapeutic efficacy of tislelizumab combined with lenvatinib, and the values ​​of the patients' serological indicators. Next, the TabPFN model was trained using different training sets, resulting in 1000 trained models. The predictive performance of different serological indicator groups in the TabPFN model was then systematically explored using a test set. The test set used retrospective cohort data from Tianjin Cancer Hospital as an independent internal validation set to test the model's generalization ability and practical application value in real-world scenarios.

[0078] Based on the test results, the final prediction model maintained a high level of predictive ability, such as... Figure 1 As shown, the prediction model has an AUC of 0.917, an ACC of 0.778, an AP of 0.966, and a specificity of 100%, reflecting its good generalization ability and clinical applicability. The diagnostic serological indicators corresponding to this prediction model consist of lactate dehydrogenase isoenzyme, gamma-glutamyl transferase, uric acid, gamma globulin, total bilirubin, and C-reactive protein. Figure 2 The results visually demonstrate the combined ranking of the six finalized diagnostic serological indicators based on F-statistic, MI value, and random forest importance. It can be seen that while the importance ranking of these six diagnostic serological indicators is not high, they exhibit excellent predictive ability when combined with the TabPFN model, enabling the determination of whether patients should receive tisraizumab combined with lenvatinib before surgery.

[0079] To reduce the number of diagnostic serological indicators without significantly affecting the prediction results, the weights of six diagnostic serological indicators in the prediction model were analyzed. For example... Figure 3As shown, the importance ranking is as follows: lactate dehydrogenase isoenzyme, γ-glutamyl transferase, uric acid, γ-globulin, total bilirubin, and C-reactive protein. Further, the synergistic relationship among the six diagnostic serological indicators was analyzed using linear discriminant analysis. Specifically, training and test sets were constructed using prospective and retrospective cohort data, with randomized seeding iteratively updated to cover multiple partitioning scenarios. Forward sequence feature selection (SFS) combined with linear discriminant analysis (LDA) was used as the base classifier, and 3-fold cross-validation was performed using ROC-AUC as the scoring criterion to select the optimal feature subset. For the selected feature subset, the LDA model was trained, and ROC-AUC was calculated on the test set to evaluate the model's discriminative ability, and the Mann-Whitney U test was used. Ultimately, it was found that although lactate dehydrogenase isoenzyme, uric acid, and γ-globulin were not among the top three weights in the prediction model, they showed a stronger interaction relationship, and a linear discriminant expression was constructed based on these three indicators:

[0080]

[0081] Where G is the value of γ-globulin, UA is the value of uric acid, and LDH_ISO is the value of lactate dehydrogenase isoenzyme;

[0082] When the discriminant score D is greater than 0.3819, neoadjuvant therapy is ineffective for the patient; when the discriminant score D is less than or equal to 0.3819, neoadjuvant therapy is effective for the patient.

[0083] like Figure 4 As shown, the linear discriminant constructed based on lactate dehydrogenase isoenzyme, uric acid, and γ-globulin achieved an AUC of 0.913 in the training set and an AUC of 0.933 in the test set, indicating that it can easily, non-invasively, and efficiently screen patients who are unresponsive to tislelizumab combined with lenvatinib before surgery.

[0084]

Example 4

[0085] Based on the above embodiments, a storage medium based on a prediction model includes a stored computer program, wherein, when the computer program is executed, the device where the storage medium is located is controlled to perform the following steps:

[0086] The diagnostic serological indicators of the patient are obtained, which consist of lactate dehydrogenase isoenzyme, gamma-glutamyl transferase, uric acid, gamma globulin, total bilirubin, and C-reactive protein.

[0087] Based on the diagnostic serological test indicators and prediction model, a prediction result of whether neoadjuvant therapy is ineffective for the patient is generated and the prediction result is output. The neoadjuvant therapy refers to the treatment method of using tislelizumab combined with lenvatinib before surgery.

[0088] The method for constructing the prediction model includes the following steps:

[0089] To obtain several serological test indicators from several liver cancer patients before neoadjuvant therapy;

[0090] At least two of the serological test indicators are randomly combined to obtain several serological test indicator groups. Based on the serological test indicator groups and the corresponding neoadjuvant therapy efficacy in patients, several training sets are constructed.

[0091] The model is trained using the aforementioned training sets to obtain several trained models. The trained models are then tested to obtain test results. Based on the test results, the prediction model is determined. The test serological detection indicators in the training set corresponding to the prediction model are used as diagnostic serological detection indicators.

[0092] In this embodiment, the storage medium is a computer-readable storage medium. If the prediction steps are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0093]

Example 5

[0094] Based on the above embodiments, a linear discriminant-based storage medium includes a stored computer program, wherein, when the computer program is executed, the device containing the storage medium is controlled to perform the following steps:

[0095] The diagnostic serological indicators of the patient are obtained, which consist of lactate dehydrogenase isoenzyme, uric acid, and gamma globulin;

[0096] Based on the diagnostic serological test indicators and linear discriminant formula, a predictive result is generated as to whether neoadjuvant therapy is ineffective for the patient. Neoadjuvant therapy refers to the treatment method of using tisraizumab combined with lenvatinib before surgery.

[0097] The method for constructing the linear discriminant includes the following steps:

[0098] To obtain several serological test indicators from several liver cancer patients before neoadjuvant therapy;

[0099] At least two of the serological test indicators are randomly combined to obtain several serological test indicator groups. Based on the serological test indicator groups and the corresponding neoadjuvant therapy efficacy in patients, several training sets are constructed.

[0100] Several training sets are used to train models, resulting in several trained models. These trained models are then tested to obtain test results. Based on the test results, a prediction model is determined. Based on the test serological indicators corresponding to the prediction model and the corresponding neoadjuvant therapy effectiveness in patients, linear discriminant analysis is used to select one or more test serological indicators corresponding to the prediction model as diagnostic serological indicators to be input. A linear discriminant expression based on these diagnostic serological indicators is then constructed. In this linear discriminant expression, the discriminant score... Where G is the value of γ-globulin, UA is the value of uric acid, and LDH_ISO is the value of lactate dehydrogenase isoenzyme; when the discriminant score D is greater than 0.3819, neoadjuvant therapy is ineffective for the patient; when the discriminant score D is less than or equal to 0.3819, neoadjuvant therapy is effective for the patient.

[0101] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A clinical neoadjuvant therapy efficacy prediction system for liver cancer, characterized in that, include: The acquisition unit is used to acquire several serological test indicators of several liver cancer patients before neoadjuvant therapy. The training set construction unit is used to randomly combine at least two of the several serological test indicators to obtain several serological test indicator groups. Based on the serological test indicator groups and the corresponding neoadjuvant therapy efficacy for patients, several training sets are constructed. The neoadjuvant therapy refers to the treatment method of using tislelizumab combined with lenvatinib before surgery. The training unit is used to train the model using the plurality of training sets to obtain a plurality of trained models, test the trained models to obtain test results, determine the prediction model based on the test results, and determine the test serological detection indicators in the training set corresponding to the prediction model as the diagnostic serological detection indicators to be input. The input unit is used to input the values ​​of the patient's diagnostic serological test indicators; The prediction unit is used to generate a prediction result on whether neoadjuvant therapy is ineffective for the patient based on the prediction model and the diagnostic serological test indicators. The output unit is used to output the prediction result.

2. The clinical neoadjuvant therapy efficacy prediction system for liver cancer according to claim 1, characterized in that, The diagnostic serological indicators consist of lactate dehydrogenase isoenzyme, gamma-glutamyl transferase, uric acid, gamma globulin, total bilirubin, and C-reactive protein.

3. The clinical neoadjuvant therapy efficacy prediction system for liver cancer according to claim 1, characterized in that, After obtaining the test serological indicators corresponding to the prediction model, the training unit uses linear discriminant analysis to screen one or more test serological indicators corresponding to the prediction model as diagnostic serological indicators to be input, based on the test serological indicators corresponding to the prediction model and the corresponding neoadjuvant treatment effectiveness of the patients, and constructs a linear discriminant formula based on the diagnostic serological indicators. The prediction unit is used to generate a prediction result on whether neoadjuvant therapy is ineffective for the patient based on the diagnostic serological test indicators and the linear discriminant.

4. The clinical neoadjuvant therapy efficacy prediction system for liver cancer according to claim 3, characterized in that, The diagnostic serological indicators consist of lactate dehydrogenase isoenzyme, uric acid, and gamma globulin.

5. The clinical neoadjuvant therapy efficacy prediction system for liver cancer according to claim 3, characterized in that, In the linear discriminant expression, the discriminant score Where G is the value of γ-globulin, UA is the value of uric acid, and LDH_ISO is the value of lactate dehydrogenase isoenzyme; When the discrimination score D is greater than 0.3819, neoadjuvant therapy is ineffective for the patient; Neoadjuvant therapy is effective for patients when the discrimination score D is less than or equal to 0.3819.

6. The clinical neoadjuvant therapy efficacy prediction system for liver cancer according to claim 1, characterized in that, The acquisition unit is used to acquire several raw serological test indicators, and to process the raw serological test indicators using at least one evaluation method among the F-statistic method, mutual information method, and random forest method, and to screen out the several test serological test indicators.

7. The clinical neoadjuvant therapy efficacy prediction system for liver cancer according to claim 1, characterized in that, The training unit uses the TabPFN model.

8. A storage medium, characterized in that, Includes a stored computer program, wherein, when the computer program is executed, the device containing the storage medium controls the following steps to be performed: The diagnostic serological indicators of the patient are obtained, which consist of lactate dehydrogenase isoenzyme, gamma-glutamyl transferase, uric acid, gamma globulin, total bilirubin, and C-reactive protein. Based on the diagnostic serological test indicators and prediction model, a prediction result of whether neoadjuvant therapy is ineffective for the patient is generated and the prediction result is output. The neoadjuvant therapy refers to the treatment method of using tislelizumab combined with lenvatinib before surgery. The method for constructing the prediction model includes the following steps: To obtain several serological test indicators from several liver cancer patients before neoadjuvant therapy; At least two of the serological test indicators are randomly combined to obtain several serological test indicator groups. Based on the serological test indicator groups and the corresponding neoadjuvant therapy efficacy in patients, several training sets are constructed. The model is trained using the aforementioned training sets to obtain several trained models. The trained models are then tested to obtain test results. Based on the test results, the prediction model is determined. The test serological detection indicators in the training set corresponding to the prediction model are used as diagnostic serological detection indicators.

9. A storage medium, characterized in that, Includes a stored computer program, wherein, when the computer program is executed, the device containing the storage medium controls the following steps to be performed: The diagnostic serological indicators of the patient are obtained, which consist of lactate dehydrogenase isoenzyme, uric acid, and gamma globulin; Based on the diagnostic serological test indicators and linear discriminant formula, a predictive result is generated to determine whether neoadjuvant therapy is ineffective for the patient. Neoadjuvant therapy refers to the treatment method of using tisraizumab combined with lenvatinib before surgery. The method for constructing the linear discriminant includes the following steps: To obtain several serological test indicators from several liver cancer patients before neoadjuvant therapy; At least two of the serological test indicators are randomly combined to obtain several serological test indicator groups. Based on the serological test indicator groups and the corresponding neoadjuvant therapy efficacy in patients, several training sets are constructed. The model is trained using the aforementioned training sets to obtain several trained models. The trained models are then tested to obtain test results. Based on the test results, the prediction model is determined. Based on the test serological indicators corresponding to the prediction model and the corresponding neoadjuvant therapy effectiveness for the patients, linear discriminant analysis is used to screen one or more test serological indicators corresponding to the prediction model as diagnostic serological indicators to be input, and a linear discriminant expression based on the diagnostic serological indicators is constructed.

10. A storage medium according to claim 9, characterized in that, In the linear discriminant expression, the discriminant score Where G is the value of γ-globulin, UA is the value of uric acid, and LDH_ISO is the value of lactate dehydrogenase isoenzyme; When the discrimination score D is greater than 0.3819, neoadjuvant therapy is ineffective for the patient; Neoadjuvant therapy is effective for patients when the discrimination score D is less than or equal to 0.3819.

Images