Hepatocellular carcinoma tumor micro-necrosis prediction method and system based on radiomics

Abstract

The invention belongs to the technical field of cancer cell detection, and discloses a hepatocellular carcinoma tumor micronecrosis prediction method and system based on radiomics, and the method comprises the steps: obtaining the abdomen enhanced CT of a hepatocellular carcinoma patient before treatment, and drawing and circling a tumor region; the method comprises the following steps: extracting radiomics characteristics, screening radiomics key characteristics highly related to tumor micro-necrosis by using an LASSO logistic regression mode, and further evaluating the state of hepatocellular carcinoma tumor micro-necrosis by training and determining an optimal hyper-parameter and a prediction model. According to the image omics-based prediction method, the prediction accuracy of hepatocellular carcinoma tumor micro-necrosis can be improved by analyzing and processing the three-dimensional tumor image; meanwhile, through accurate prediction of hepatocellular carcinoma tumor micro-necrosis, a personalized treatment scheme can be formulated for the patient, and the treatment effect and life quality of the patient can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of cancer cell detection, and relates to a method and system for predicting hepatocellular carcinoma tumor micronecrosis based on imaging omics. Background Art

[0002] Hepatocellular carcinoma (HCC) is highly malignant. Although there are a variety of treatments for HCC, their efficacy varies and the overall prognosis remains poor. An important reason for the different prognosis of HCC is that HCC has high heterogeneity both inside and outside the tumor. Tumor micronecrosis is an important biological behavior marker of HCC, which reflects the heterogeneity within the tumor to a certain extent and can cause poor prognosis in HCC patients. It is of great significance to stratify HCC patients according to the status of tumor micronecrosis and adopt personalized treatment. Existing studies have used H&E stained sections of surgically resected HCC pathological tissue to evaluate tumor micronecrosis. However, some HCC patients cannot undergo surgery, so it is difficult to evaluate the micronecrosis status, which is a limiting factor that makes the existing evaluation system difficult to apply to all HCC patients. Summary of the invention

[0003] The purpose of the present invention is to solve the problem in the prior art that some HCC patients cannot undergo surgery, resulting in difficulty in evaluating the micronecrosis status, and to provide a method and system for predicting micronecrosis of hepatocellular carcinoma tumors based on imaging omics.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions:

[0005] The prediction method of tumor micronecrosis in hepatocellular carcinoma based on radiomics includes:

[0006] Obtain preoperative abdominal enhanced CT data of HCC patients, outline the tumor area in the CT image, and obtain a three-dimensional tumor image;

[0007] Extract radiomic features from tumor images; use H&E stained sections of surgically resected hepatocellular carcinoma pathological tissue to assess tumor micronecrosis as the standard, analyze and calculate the association between radiomic features and tumor micronecrosis, and screen out key radiomic features;

[0008] The screened key features are linearly combined and weighted by their non-zero LASSO coefficients to construct a radiomics signature that encapsulates the comprehensive tumor image features as independent predictors;

[0009] A logistic regression prediction model was constructed to score the imaging information of each patient based on the predictive factors and the prediction model as the prediction value of tumor micronecrosis.

[0010] Determine the relationship between the obtained predicted value and the preset threshold, perform preoperative non-invasive prediction, and then evaluate the prognosis level and guide the treatment plan.

[0011] A further improvement of the present invention is:

[0012] Furthermore, the preoperative abdominal enhanced CT data of HCC patients were obtained, and the tumor area in the CT image was outlined to obtain a three-dimensional tumor image. Specifically, the DICOM format of the preoperative abdominal enhanced CT data was converted into the nifti format, and the CT data after format conversion was input into the ITK-SNAP software, and the tumor area in the CT image was manually outlined.

[0013] Furthermore, radiomic features were extracted from tumor images. Specifically, radiomic features were extracted from the outlined nifit format files, and quantitative image features were divided into four groups: (a) first-order features calculated directly from intensity values, including percentile, energy, and entropy; (b) shape and size features, quantifying the 3D shape and size of each tumor, including diameter, volume, and sphericity; (c) texture features; and (d) filter-based features.

[0014] Furthermore, the tumor micronecrosis assessed by H&E staining of surgically resected HCC pathological tissue was used as the standard to analyze and calculate the association between imaging genomics features and tumor micronecrosis, and screen out the key features of imaging genomics, specifically:

[0015] Obtain high-quality H&E-stained sections from surgically resected HCC pathological tissues and determine whether there is tumor micronecrosis in a specific section, and then determine a binary classification label, 1 for the presence of tumor micronecrosis and 0 for the absence; a specific section is 1 section of the tumor center plus 4 sections of the tumor periphery;

[0016] Match the extracted radiomics feature dataset with the binary classification label dataset to ensure that each feature has a corresponding label;

[0017] Based on statistical tests, features that were not significantly associated with tumor micronecrosis were screened out to obtain key features of imaging omics.

[0018] Furthermore, it is determined whether there is tumor micronecrosis in a specific slice, and then the binary classification label is determined, specifically:

[0019] The necrosis score (SNS) of the slice is determined according to the characteristics and scope of its pathological manifestations. The grading and scoring criteria of SNS are as follows:

[0020] SNS=0, necrotic area ≤5%; SNS=1, 5%<necrotic area ≤20%; SNS=2, 20%<necrotic area ≤50%; SNS=3, necrotic area>50%.

[0021] According to the slice necrosis score, the liver cancer necrosis score Nscore was calculated. The grading criteria of Nscore were as follows: Nscore = 0, all slices SNS = 0; Nscore = 1, ≥1 slice SNS = 1 and ≤1 slice SNS = 2; Nscore = 2, ≥2 slices SNS = 2 or only 1 slice SNS = 3; Nscore = 3, ≥2 slices SNS = 3;

[0022] The slices were divided into two categories according to the Nscore score: Nscore>0 indicated the presence of tumor micronecrosis, with the label 1; Nscore=0 indicated the absence of tumor micronecrosis, with the label 0.

[0023] Further, statistical tests were performed, specifically: the stability of each feature was evaluated by the intraclass correlation coefficient (ICC), and features with a 95% confidence interval of 0.8 or higher were designated as robust features; features with variance close to zero were removed after ICC analysis; robust features were ranked based on the minimum redundancy maximum correlation (mRMR) algorithm, and the first 30 features of multivariate logistic regression modeling were promoted using the least absolute shrinkage and selection operator LASSO; the coefficient of each feature in the LASSO model was checked, and features with non-zero coefficients were retained from the first 30 features, ultimately resulting in 12 features.

[0024] Furthermore, the selected key features are linearly combined and weighted by their non-zero LASSO coefficients to construct a radiomics label, which encapsulates the comprehensive tumor image features into independent predictors, specifically:

[0025] Based on the k-fold cross-validation method, the regularization parameter λ in the LASSO regression is determined, and then the optimal LASSO regression model is obtained, in which the feature corresponding to the non-zero coefficient is the key feature;

[0026] Multiply the values ​​of the selected key features by their respective LASSO coefficients, i.e., non-zero coefficients, and then add these products to obtain the result of the linear combination;

[0027] The result of the linear combination is:

[0028]

[0029] Among them, X i is the value of the i-th key feature, β i is the LASSO coefficient corresponding to the feature, and n is the number of key features; the result of the above linear combination is encapsulated as an independent predictor, namely the radiomics label.

[0030] Furthermore, a logistic regression prediction model is constructed, specifically:

[0031] Based on the key features of radiomics, a logistic regression prediction model was constructed;

[0032] The obtained three-dimensional tumor images are divided into a development set, an internal test set, and an external validation set;

[0033] The logistic regression prediction model was trained in the development set using the five-fold cross-validation method. The average performance index of each hyperparameter combination on the five-fold was recorded, and the hyperparameter combination with the best average performance index was selected as the hyperparameter of the final model.

[0034] The Logistic regression model was retrained on the entire development set through the selected optimal hyperparameter combination; the internal test set was used as input into the optimized logistic regression prediction model, and the optimal model was obtained by evaluating the AUC performance of the model; the external validation set was input into the optimal logistic regression prediction model tested by the test set, and the accuracy, precision, recall, F1 score, and AUC-ROC performance indicators were selected to evaluate the performance of the model.

[0035] The prediction system of tumor micronecrosis in hepatocellular carcinoma based on radiomics includes:

[0036] An acquisition module, wherein the acquisition module acquires preoperative abdominal enhanced CT data of HCC patients, and outlines the tumor area in the CT image to obtain a three-dimensional tumor image;

[0037] A screening module, wherein the screening module extracts radiomic features from tumor images; and uses H&E stained sections of surgically resected hepatocellular carcinoma pathological tissue to evaluate tumor micronecrosis as a standard, analyzes and calculates the correlation between radiomic features and tumor micronecrosis, and screens out key radiomic features;

[0038] A first construction module, wherein the first construction module linearly combines the screened key features and weights them by their non-zero LASSO coefficients to construct a radiomics label, which encapsulates the comprehensive tumor image features as independent predictors;

[0039] A second building module, wherein the second building module constructs a logistic regression prediction model, and scores the imaging omics of each patient based on the predictive factors and the prediction model as a prediction value of tumor micronecrosis;

[0040] The judgment module judges the relationship between the obtained prediction value and the preset threshold value, performs preoperative non-invasive prediction, and further evaluates the prognosis level and guides the treatment plan.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] The present invention obtains abdominal enhanced CT scans of hepatocellular carcinoma patients before treatment, outlines and circles the tumor area; extracts radiomics features, uses LASSO logistic regression to screen radiomics key features that are highly correlated with tumor micronecrosis, and trains and determines the optimal hyperparameters and prediction models to further evaluate the state of hepatocellular carcinoma tumor micronecrosis. The radiomics-based prediction method of the present invention can improve the accuracy of predicting hepatocellular carcinoma tumor micronecrosis by analyzing and processing three-dimensional tumor images; at the same time, through accurate prediction of hepatocellular carcinoma tumor micronecrosis, a personalized treatment plan can be formulated for the patient, which helps to improve the patient's treatment effect and quality of life. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0044] Figure 1 Schematic diagram of the process of the method for predicting tumor micronecrosis in hepatocellular carcinoma based on imaging omics of the present invention;

[0045] Figure 2 Schematic diagram of the structure of the hepatocellular carcinoma tumor micronecrosis prediction system based on imaging omics of the present invention;

[0046] Figure 3 The figure is an overall flow chart of the method for predicting tumor micronecrosis in hepatocellular carcinoma based on radiomics of the present invention;

[0047] Figure 4 (a) is a schematic diagram of the AUC curve of the prediction model in the development set;

[0048] Figure 4 (b) is a schematic diagram of the AUC curve of the prediction model in the internal test set;

[0049] Figure 4 (c) is a schematic diagram of the AUC curve of the prediction model in the external validation set;

[0050] Figure 5 (a) Schematic diagram of the overall survival curve of patients with different tumor micronecrosis status confirmed by pathology;

[0051] Figure 5 (b) Schematic diagram of the overall survival curve of patients with different tumor micronecrosis status predicted by imaging;

[0052] Figure 5(c) Schematic diagram of the recurrence time curve of patients with different tumor micronecrosis status confirmed by pathology;

[0053] Figure 5 (d) Schematic diagram of the recurrence time curve of patients with different tumor micronecrosis status predicted by imaging. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0055] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0057] In the description of the embodiments of the present invention, it should be noted that if the terms "upper", "lower", "horizontal", "inner", etc. indicate an orientation or positional relationship based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the product of the invention is usually placed when in use, it is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0058] In addition, if the term "horizontal" appears, it does not mean that the component must be absolutely horizontal, but can be slightly tilted. For example, "horizontal" only means that its direction is more horizontal than "vertical", which does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0059] In the description of the embodiments of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal connection of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0060] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0061] See also Figure 1 The present invention discloses a method for predicting tumor micronecrosis in hepatocellular carcinoma based on radiomics, comprising:

[0062] S101: Obtain preoperative abdominal enhanced CT data of HCC patients, outline the tumor area in the CT image, and obtain a three-dimensional tumor image;

[0063] The DICOM format of the preoperative abdominal enhanced CT data was converted into the nifti format, and the converted CT data were input into the ITK-SNAP software to manually outline the tumor area in the CT image.

[0064] S102: Extract radiomic features from tumor images; use the H&E stained sections of surgically resected hepatocellular carcinoma pathological tissue to evaluate tumor micronecrosis as the standard, analyze and calculate the correlation between radiomic features and tumor micronecrosis, and screen out key radiomic features;

[0065] Extract radiomic features from tumor images. Specifically, extract radiomic features from the outlined nifit format files. Quantitative image features are divided into four groups: (a) first-order features calculated directly from intensity values, including percentile, energy, and entropy; (b) shape and size features, quantifying the 3D shape and size of each tumor, including diameter, volume, and sphericity; (c) texture features; and (d) filter-based features.

[0066] Taking the tumor micronecrosis assessed by H&E staining sections of surgically resected hepatocellular carcinoma pathological tissue as the standard, the association between imaging omics features and tumor micronecrosis was analyzed and calculated, and the key imaging omics features were screened out, specifically:

[0067] Obtaining high-quality H&E-stained sections from surgically resected HCC pathological tissues, and determining whether tumor micronecrosis exists in a specific section, and then determining a binary classification label, where 1 indicates the presence of tumor micronecrosis and 0 indicates the absence of tumor micronecrosis; the specific section is selected from the obtained high-quality H&E-stained sections;

[0068] Match the extracted radiomics feature dataset with the binary classification label dataset to ensure that each feature has a corresponding label;

[0069] Based on statistical tests, features that were not significantly associated with tumor micronecrosis were screened out to obtain key features of imaging omics.

[0070] Determine whether tumor micronecrosis exists in a specific slice, and then determine the binary classification label, specifically:

[0071] The specific slices consisted of 1 slice of the tumor center plus 4 slices of the tumor surrounding areas;

[0072] The necrosis score (SNS) of the slice is determined according to the characteristics and scope of its pathological manifestations. The grading and scoring criteria of SNS are as follows:

[0073] SNS=0, necrotic area ≤5%; SNS=1, 5%<necrotic area ≤20%; SNS=2, 20%<necrotic area ≤50%; SNS=3, necrotic area>50%.

[0074] According to the slice necrosis score, the liver cancer necrosis score Nscore was calculated. The grading criteria of Nscore were as follows: Nscore = 0, all slices SNS = 0; Nscore = 1, ≥1 slice SNS = 1 and ≤1 slice SNS = 2; Nscore = 2, ≥2 slices SNS = 2 or only 1 slice SNS = 3; Nscore = 3, ≥2 slices SNS = 3;

[0075] The slices were divided into two categories according to the Nscore score: Nscore>0 indicated the presence of tumor micronecrosis, with the label 1; Nscore=0 indicated the absence of tumor micronecrosis, with the label 0.

[0076] The statistical test is specifically as follows: the stability of each feature is evaluated by the intraclass correlation coefficient ICC, and the features with a 95% confidence interval of 0.8 or higher are designated as robust features; after ICC analysis, the variance features close to zero are removed; the robust features are ranked based on the minimum redundancy maximum correlation (mRMR) algorithm, and the first 30 features of multivariate logistic regression modeling are promoted using the minimum absolute shrinkage and selection operator LASSO; the coefficient of each feature in the LASSO model is checked, and the features with non-zero coefficients are retained from the first 30 features, and finally 12 features are obtained.

[0077] S103: Linearly combine the selected key features and weight them by their non-zero LASSO coefficients to construct a radiomics signature that encapsulates the comprehensive tumor image features as independent predictors;

[0078] Based on the k-fold cross-validation method, the regularization parameter λ in the LASSO regression is determined, and then the optimal LASSO regression model is obtained, in which the feature corresponding to the non-zero coefficient is the key feature;

[0079] Multiply the values ​​of the selected key features by their respective LASSO coefficients, i.e., non-zero coefficients, and then add these products to obtain the result of the linear combination;

[0080] The result of the linear combination is:

[0081]

[0082] Among them, X i is the value of the i-th key feature, β i is the LASSO coefficient corresponding to the feature, and n is the number of key features; the result of the above linear combination is encapsulated as an independent predictor, namely the radiomics label.

[0083] S104: Construct a logistic regression prediction model and score each patient's imaging omics based on the predictors and prediction model as the predictive value of tumor micronecrosis;

[0084] Construct a logistic regression prediction model, specifically:

[0085] Based on the key features of radiomics, a logistic regression prediction model was constructed;

[0086] The obtained three-dimensional tumor images are divided into a development set, an internal test set, and an external validation set;

[0087] The logistic regression prediction model was trained in the development set using the five-fold cross-validation method. The average performance index of each hyperparameter combination on the five-fold was recorded, and the hyperparameter combination with the best average performance index was selected as the hyperparameter of the final model.

[0088] The Logistic regression model was retrained on the entire development set through the selected optimal hyperparameter combination; the internal test set was used as input into the optimized logistic regression prediction model, and the optimal model was obtained by evaluating the AUC performance of the model; the external validation set was input into the optimal logistic regression prediction model tested by the test set, and the accuracy, precision, recall, F1 score, and AUC-ROC performance indicators were selected to evaluate the performance of the model.

[0089] S105: Determine the relationship between the obtained prediction value and the preset threshold value, perform preoperative non-invasive prediction, and then evaluate the prognosis level and guide the treatment plan.

[0090] By judging the relationship between the obtained prediction value and the preset threshold, HCC patients were divided into a high-risk group with micronecrosis and a low-risk group with micronecrosis, and then the prognosis level was evaluated and the treatment plan was guided.

[0091] See also Figure 2 The present invention discloses a hepatocellular carcinoma tumor micronecrosis prediction system based on radiomics, comprising:

[0092] An acquisition module, wherein the acquisition module acquires preoperative abdominal enhanced CT data of HCC patients, and outlines the tumor area in the CT image to obtain a three-dimensional tumor image;

[0093] A screening module, wherein the screening module extracts radiomic features from tumor images; and uses H&E stained sections of surgically resected hepatocellular carcinoma pathological tissue to evaluate tumor micronecrosis as a standard, analyzes and calculates the correlation between radiomic features and tumor micronecrosis, and screens out key radiomic features;

[0094] A first construction module, wherein the first construction module linearly combines the screened key features and weights them by their non-zero LASSO coefficients to construct a radiomics label, which encapsulates the comprehensive tumor image features as independent predictors;

[0095] A second building module, wherein the second building module constructs a logistic regression prediction model, and scores the imaging omics of each patient based on the predictive factors and the prediction model as a prediction value of tumor micronecrosis;

[0096] The judgment module judges the relationship between the obtained prediction value and the preset threshold value, performs preoperative non-invasive prediction, and further evaluates the prognosis level and guides the treatment plan.

[0097] Example:

[0098] This study collected 951 abdominal enhanced CT image data from the First Affiliated Hospital and its branch hospitals of a university medical school from 2014 to 2022; the inclusion criteria were as follows: (a) pathological diagnosis of hepatocellular carcinoma; (b) R0 resection of liver cancer or liver transplantation; (c) CECT images within one month before surgery. Patients with the following characteristics were excluded: (a) with other malignant tumors; (b) preoperative anti-tumor treatment; (c) incomplete clinical data or poor image quality. The collected image data were processed as follows:

[0099] The image data DICOM format was converted into nifti format, and the tumor area in the CT image was manually outlined using ITK-SNAP software.

[0100] Radiomic features were extracted from the outlined nifit format files, and quantitative image features were divided into four groups: (a) first-order features calculated directly from intensity values ​​(18), including percentile, energy, entropy, etc.; (b) shape and size features (14), quantifying the 3D shape and size of each tumor, including diameter, volume, sphericity, etc.; (c) texture features (75); (d) filter-based features (744). A total of 1702 radiomic features were extracted.

[0101] The data is divided into three parts: development set, internal test set and external validation set. The flowchart can be seen Figure 3 Taking the tumor micronecrosis assessed by H&E staining sections of surgically resected hepatocellular carcinoma pathological tissue as the standard, the association between radiomics features and tumor micronecrosis was analyzed and calculated, and the key features of radiomics were screened out by LASSO logistic regression. Finally, 12 key features of radiomics were screened out, which are:

[0102] Obtain high-quality H&E-stained sections from surgically resected HCC pathological tissues and determine whether tumor micronecrosis exists in specific sections, and then determine a binary classification label, where 1 indicates the presence of tumor micronecrosis and 0 indicates the absence;

[0103] Match the extracted radiomics feature dataset with the binary classification label dataset to ensure that each feature has a corresponding label;

[0104] Based on statistical tests, features that were not significantly associated with tumor micronecrosis were screened out to obtain key features of imaging omics.

[0105] The 12 key features of radiomics are: rad_V_wavelet_HHH_glszm_GrayLevelVariance, rad_A_wavelet_HHH_glszm_HighGrayLevelZoneEmphasis, rad_V_wavelet_HHH_glszm_SizeZoneNonUniformityNormalized , rad_V_original_firstorder_Median, rad_V_wavelet_HLL_ngtdm_Contrast, rad_A_original_firstorder_90Percentile, rad_V_wavelet_LLL_firstorder_Uniformity, rad_V_wavelet_LLL_firstorder_10Percentile , rad_V_wavelet_LLL_glszm_SmallAreaLowGrayLevelEmphasis , rad_V_wavelet_LLL_glcm_Correlation , rad_A_wavelet_HHH_glrlm_GrayLevelNonUniformityNormalized, rad_A_wavelet_LHL_glszm_SmallAreaLowGrayLevelEmphasis.

[0106] Based on the key features of radiomics, a logistic regression prediction model was constructed;

[0107] The obtained three-dimensional tumor images are divided into a development set, an internal test set, and an external validation set;

[0108] The logistic regression prediction model was trained in the development set using the five-fold cross-validation method. The average performance index of each hyperparameter combination on the five-fold was recorded, and the hyperparameter combination with the best average performance index was selected as the hyperparameter of the final model.

[0109] The Logistic regression model was retrained on the entire development set through the selected optimal hyperparameter combination; the internal test set was used as input into the optimized logistic regression prediction model, and the optimal model was obtained by evaluating the AUC performance of the model; the external validation set was input into the optimal logistic regression prediction model tested by the test set, and the accuracy, precision, recall, F1 score, and AUC-ROC performance indicators were selected to evaluate the performance of the model.

[0110] See also Figure 4 (a) Figure 4 (b) and Figure 4 (c), the AUC of the radiomics score was 0.83 (95% CI, 0.80-0.86) in the model development set, 0.80 (95% CI, 0.73-0.86) in the internal test set, and 0.76 (95% CI, 0.65-0.87) in the external validation set. Figure 4 (a) Figure 4 (b) and Figure 4 In (c), RadClinical is radiomics + clinical information; Radiomic is radiomics; Clinical is clinical information.

[0111] See also Figure 5 (a) Figure 5 (b) Figure 5 (c) and Figure 5 (d) In the patient-level analysis, the radiomics score of each patient was calculated, with the threshold value of 0.43 calculated according to the Yoden index method as the cutoff value. Patients with a score greater than or equal to 0.43 were predicted to be positive for micronecrosis, and patients with a score less than 0.43 were predicted to be negative. The Kaplan-Meier curve showed that there was a significant difference in OS between the two groups (P<0.05), and it was close to the actual grouping curve determined by pathological criteria.

[0112] In summary, the imaging omics model can accurately predict tumor micronecrosis in hepatocellular carcinoma and has certain clinical value.

[0113] The terminal device provided in an embodiment of the present invention. The terminal device of this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps in the above-mentioned method embodiments are implemented. Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiments are implemented.

[0114] The computer program may be divided into one or more modules / units, and the one or more modules / units are stored in the memory and executed by the processor to accomplish the present invention.

[0115] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0116] The processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0117] The memory may be used to store the computer program and / or module, and the processor implements various functions of the terminal device by running or executing the computer program and / or module stored in the memory and calling the data stored in the memory.

[0118] If the module / unit integrated in the terminal device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0119] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for predicting tumor micronecrosis in hepatocellular carcinoma based on radiomics, characterized in that: include: Obtain preoperative abdominal enhanced CT data of HCC patients, outline the tumor area in the CT image, and obtain a three-dimensional tumor image; Extract radiomic features from tumor images; use H&E stained sections of surgically resected hepatocellular carcinoma pathological tissue to assess tumor micronecrosis as the standard, analyze and calculate the association between radiomic features and tumor micronecrosis, and screen out key radiomic features; The screened key features are linearly combined and weighted by their non-zero LASSO coefficients to construct a radiomics signature that encapsulates the comprehensive tumor image features as independent predictors; A logistic regression prediction model was constructed to score the imaging information of each patient based on the predictive factors and the prediction model as the prediction value of tumor micronecrosis. Determine the relationship between the obtained predicted value and the preset threshold, perform preoperative non-invasive prediction, and then evaluate the prognosis level and guide the treatment plan.

2. The method for predicting tumor micronecrosis in hepatocellular carcinoma based on radiomics according to claim 1, characterized in that: The method of obtaining preoperative abdominal enhanced CT data of HCC patients, outlining the tumor area in the CT image, and obtaining a three-dimensional tumor image specifically includes: converting the DICOM format of the preoperative abdominal enhanced CT data into the nifti format, inputting the format-converted CT data into the ITK-SNAP software, and manually outlining the tumor area in the CT image.

3. The method for predicting tumor micronecrosis in hepatocellular carcinoma based on radiomics according to claim 2, characterized in that: The method of extracting radiomic features from tumor images is as follows: extracting radiomic features from the outlined nifit format file, and the quantitative image features are divided into four groups: (a) first-order features calculated directly from intensity values, including percentile, energy, and entropy; (b) shape and size features, quantifying the 3D shape and size of each tumor, including diameter, volume, and sphericity; (c) texture features; (d) Filter-based features.

4. The method for predicting tumor micronecrosis in hepatocellular carcinoma based on radiomics according to claim 3, characterized in that: The tumor micronecrosis evaluated by H&E staining sections of surgically resected hepatocellular carcinoma pathological tissue was used as the standard to analyze and calculate the association between imaging omics features and tumor micronecrosis, and screen out key imaging omics features, specifically: Obtain high-quality H&E-stained sections from surgically resected HCC pathological tissues and determine whether there is tumor micronecrosis in a specific section, and then determine a binary classification label, 1 for the presence of tumor micronecrosis and 0 for the absence; a specific section is 1 section of the tumor center plus 4 sections of the tumor periphery; Match the extracted radiomics feature dataset with the binary classification label dataset to ensure that each feature has a corresponding label; Based on statistical tests, features that were not significantly associated with tumor micronecrosis were screened out to obtain key features of imaging omics.

5. The method for predicting tumor micronecrosis in hepatocellular carcinoma based on radiomics according to claim 4, characterized in that: The determination of whether there is tumor micronecrosis in a specific slice and then determining a binary classification label is specifically as follows: The necrosis score (SNS) of the slice is determined according to the characteristics and scope of its pathological manifestations. The grading and scoring criteria of SNS are as follows: SNS=0, necrotic area ≤5%; SNS=1, 5%<necrotic area ≤20%; SNS=2, 20%<necrotic area ≤50%; SNS=3, necrotic area>50%. According to the slice necrosis score, the liver cancer necrosis score Nscore was calculated. The grading criteria of Nscore were as follows: Nscore = 0, all slices SNS = 0; Nscore = 1, ≥1 slice SNS = 1 and ≤1 slice SNS = 2; Nscore = 2, ≥2 slices SNS = 2 or only 1 slice SNS = 3; Nscore = 3, ≥2 slices SNS = 3; The slices were divided into two categories according to the Nscore score: Nscore>0 indicated the presence of tumor micronecrosis, with the label 1; Nscore=0 indicated the absence of tumor micronecrosis, with the label 0.

6. The method for predicting tumor micronecrosis in hepatocellular carcinoma based on radiomics according to claim 5, characterized in that: The statistical test is specifically as follows: the stability of each feature is evaluated by the intraclass correlation coefficient (ICC), and the features with a 95% confidence interval of 0.8 or higher are designated as robust features; After ICC analysis, features with variance close to zero were removed; robust features were ranked based on the minimum redundancy maximum relevance (mRMR) algorithm, and the least absolute shrinkage and selection operator (LASSO) was used to advance the first 30 features for multivariate logistic regression modeling; the coefficient of each feature in the LASSO model was checked, and features with non-zero coefficients were retained from the first 30 features, ultimately resulting in 12 features.

7. The method for predicting tumor micronecrosis in hepatocellular carcinoma based on radiomics according to claim 6, characterized in that: The screened key features are linearly combined and weighted by their non-zero LASSO coefficients to construct an imaging omics label, which encapsulates the comprehensive tumor image features as independent predictors, specifically: Based on the k-fold cross-validation method, the regularization parameter λ in the LASSO regression is determined, and then the optimal LASSO regression model is obtained, in which the feature corresponding to the non-zero coefficient is the key feature; Multiply the values ​​of the selected key features by their respective LASSO coefficients, i.e., non-zero coefficients, and then add these products to obtain the result of the linear combination; The result of the linear combination is: Among them, X i is the value of the i-th key feature, β i is the LASSO coefficient corresponding to the feature, and n is the number of key features; the results of the above linear combination are encapsulated as an independent predictor, namely the radiomics label.

8. The method for predicting tumor micronecrosis in hepatocellular carcinoma based on radiomics according to claim 7, characterized in that: The logistic regression prediction model is constructed as follows: Based on the key features of radiomics, a logistic regression prediction model was constructed; The obtained three-dimensional tumor images are divided into a development set, an internal test set, and an external validation set; The logistic regression prediction model was trained in the development set using the five-fold cross-validation method. The average performance index of each hyperparameter combination on the five-fold was recorded, and the hyperparameter combination with the best average performance index was selected as the hyperparameter of the final model. Retrain the Logistic regression model on the entire development set using the selected optimal hyperparameter combination; The internal test set was input into the optimized logistic regression prediction model, and the optimal model was obtained by evaluating the AUC performance of the model. The external validation set was input into the optimal logistic regression prediction model tested by the test set, and the accuracy, precision, recall, F1 score, and AUC-ROC performance indicators were selected to evaluate the performance of the model.

9. A radiomics-based prediction system for hepatocellular carcinoma tumor micronecrosis, characterized in that: include: An acquisition module, wherein the acquisition module acquires preoperative abdominal enhanced CT data of HCC patients, and outlines the tumor area in the CT image to obtain a three-dimensional tumor image; A screening module, wherein the screening module extracts radiomic features from tumor images; and uses H&E stained sections of surgically resected hepatocellular carcinoma pathological tissue to evaluate tumor micronecrosis as a standard, analyzes and calculates the correlation between radiomic features and tumor micronecrosis, and screens out key radiomic features; A first construction module, wherein the first construction module linearly combines the screened key features and weights them by their non-zero LASSO coefficients to construct a radiomics label, which encapsulates the comprehensive tumor image features as independent predictors; A second building module, wherein the second building module constructs a logistic regression prediction model, and scores the imaging omics of each patient based on the predictive factors and the prediction model as a prediction value of tumor micronecrosis; The judgment module judges the relationship between the obtained prediction value and the preset threshold value, performs preoperative non-invasive prediction, and further evaluates the prognosis level and guides the treatment plan.

Images