Hepatocellular carcinoma microvessel invasion prediction method, device and electronic equipment

Abstract

The application discloses a hepatocellular carcinoma microvessel invasion prediction method and device and electronic equipment, wherein the method comprises the following steps: obtaining preoperative image, clinical indicators and corresponding lesion pathological results of a treated patient; training a pre-constructed prediction model according to the preoperative image, clinical indicators and corresponding lesion pathological results of the treated patient; when the prediction model meets a preset standard, determining that the prediction model is a microvessel invasion prediction model; obtaining preoperative image and clinical indicators of a patient to be predicted; inputting the preoperative image and clinical indicators of the patient to be predicted into the microvessel invasion prediction model, and outputting a classification result, so as to predict whether the patient to be tested has hepatocellular carcinoma microvessel invasion. Through the method, the actual clinical scene can be adapted, more accurate classification prediction results can be obtained, and reference and convenience are provided for clinical actual diagnosis, surgical plan and prognosis evaluation and the like.

Description

Technical Field

[0001] One or more embodiments of the present invention relate to the field of medical diagnostic technology, and in particular to a method, device and electronic device for predicting microvascular invasion in hepatocellular carcinoma. Background Technology

[0002] Hepatocellular carcinoma (HCC) is one of the most common primary malignant liver tumors, and its incidence is increasing worldwide. With advancements in medical technology, the treatment of liver cancer has made significant progress, with surgical resection and liver transplantation remaining the most effective treatment methods. However, a crucial factor influencing long-term survival after HCC surgery is the presence of microvascular invasion (MVI). MVI is an independent prognostic factor for HCC patients, associated with postoperative recurrence and long-term survival. Currently, preoperative prediction of MVI in clinical practice relies primarily on imaging examinations, such as CT images and magnetic resonance imaging (MRI), to assess macroscopic tumor characteristics and serum biomarkers. However, these methods cannot accurately determine the presence of MVI in a patient. Summary of the Invention

[0003] To address the problems existing in the prior art, this invention provides a method, device, and electronic device for predicting microvascular invasion in hepatocellular carcinoma (HCC). Based on a deep learning algorithm, it uses arterial and venous phase images from enhanced abdominal CT images, along with some conventional clinical indicators, to classify and predict microvascular invasion (MVI). This provides reference and convenience for clinical diagnosis, surgical planning, and prognostic assessment, reducing postoperative recurrence rates and improving the overall prognosis of patients.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] In a first aspect, the present invention provides a method for predicting microvascular invasion in hepatocellular carcinoma, the method comprising the following steps:

[0006] Obtain preoperative imaging images, clinical indicators, and corresponding pathological results of lesions from treated patients;

[0007] Based on the preoperative images, clinical indicators, and corresponding pathological results of the lesions of the treated patients, the pre-constructed prediction model is trained. When the prediction model meets the preset criteria, the prediction model is determined to be a microvascular invasion prediction model.

[0008] Obtain preoperative imaging images and clinical indicators of the patient to be predicted;

[0009] The preoperative images and clinical indicators of the patient to be predicted are input into the microvascular invasion prediction model, and the classification results are output to predict whether the patient has microvascular invasion of hepatocellular carcinoma.

[0010] In one possible implementation, the image is an enhanced CT image.

[0011] In one possible implementation, the clinical indicators include: patient age, sex, liver function biochemical indicators, and tumor volume and maximum diameter associated with the lesion.

[0012] In one possible implementation, the microvascular invasion prediction model consists of a first single-input model and a second single-input model based on imaging images, and a first dual-input model and a second dual-input model based on imaging images and clinical indicators. The microvascular invasion prediction model uses an ensemble method to determine a classification result based on the first classification result of the first single-input model and the second classification result of the second single-input model, and the third classification result of the first dual-input model and the fourth classification result of the second dual-input model, in order to predict whether the patient under test has microvascular invasion of hepatocellular carcinoma.

[0013] In one possible implementation, the method further includes: inputting the preoperative images and clinical indicators of the patient to be predicted into the microvascular invasion prediction model to generate a model activation map corresponding to the image of the lesion area, for interpreting the classification results.

[0014] In a second aspect, the present invention provides a device for predicting microvascular invasion in hepatocellular carcinoma, the device comprising:

[0015] The first acquisition module is used to acquire preoperative imaging images, clinical indicators, and corresponding pathological results of lesions of treated patients.

[0016] The model building module is used to train a pre-built prediction model based on the preoperative images, clinical indicators and corresponding pathological results of the treated patients. When the prediction model meets the preset standards, the prediction model is determined to be a microvascular invasion prediction model.

[0017] The second acquisition module is used to acquire preoperative imaging images and clinical indicators of the patient to be predicted.

[0018] The prediction module is used to input the preoperative images and clinical indicators of the patient to be predicted into the microvascular invasion prediction model and output the classification results to predict whether the patient has microvascular invasion of hepatocellular carcinoma.

[0019] In one possible implementation, the microvascular invasion prediction model consists of a first single-input model and a second single-input model based on imaging images, and a first dual-input model and a second dual-input model based on imaging images and clinical indicators. The microvascular invasion prediction model uses an ensemble method to determine a classification result based on the first classification result of the first single-input model and the second classification result of the second single-input model, and the third classification result of the first dual-input model and the fourth classification result of the second dual-input model, in order to predict whether the patient under test has microvascular invasion of hepatocellular carcinoma.

[0020] In one possible implementation, the prediction model is further used to input the preoperative images and clinical indicators of the patient to be predicted into the microvascular invasion prediction model to generate a model activation map corresponding to the image of the lesion area, which is used to interpret the classification results.

[0021] Thirdly, an electronic device is provided, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus, and the processor includes a graphics processor and a central processing unit.

[0022] Memory, used to store computer programs;

[0023] When a processor executes a program stored in memory, it implements the steps of the method for predicting microvascular invasion in hepatocellular carcinoma as described in any embodiment of the first aspect.

[0024] Fourthly, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the hepatocellular carcinoma microvascular invasion prediction method as described in any embodiment of the first aspect.

[0025] The technical solutions provided in the embodiments of the present invention have the following advantages compared with the prior art:

[0026] This invention provides a method for predicting microvascular invasion in hepatocellular carcinoma. The method involves acquiring preoperative imaging images, clinical indicators, and corresponding pathological results of lesions from treated patients; training a pre-constructed prediction model based on these data; and determining the prediction model as a microvascular invasion prediction model when it meets preset criteria. The method also involves acquiring preoperative imaging images and clinical indicators from the patient to be predicted; inputting these data into the microvascular invasion prediction model; and outputting classification results to predict whether the patient has microvascular invasion in hepatocellular carcinoma. This approach adapts to real-world clinical scenarios, yielding more accurate classification prediction results and providing reference and convenience for clinical diagnosis, surgical planning, and prognostic assessment. Attached Figure Description

[0027] Figure 1 This is a schematic flowchart of a method for predicting microvascular invasion in hepatocellular carcinoma provided in an embodiment of the present invention;

[0028] Figure 2 This is a schematic diagram of the prediction model;

[0029] Figure 3 This is a schematic diagram of a single-input model;

[0030] Figure 4 This is a schematic diagram of a dual-input model;

[0031] Figure 5 This is a schematic diagram of the LinBnDrop layer structure;

[0032] Figure 6 This is a schematic diagram of the model activation graph;

[0033] Figure 7 ROC curve for diagnosing microvascular invasion in a hepatocellular carcinoma microvascular invasion model;

[0034] Figure 8 This is a schematic diagram of the structure of a hepatocellular carcinoma microvascular invasion prediction device provided in an embodiment of the present invention;

[0035] Figure 9 This is a schematic diagram of an electronic device structure provided in an embodiment of the present invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] To facilitate understanding of the embodiments of the present invention, further explanations and descriptions will be provided below with reference to the accompanying drawings and specific embodiments. These embodiments do not constitute a limitation on the embodiments of the present invention.

[0038] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element. Additionally, in the description of this invention, "a plurality of" means two or more, unless otherwise expressly specified.

[0039] To address the technical problems mentioned in the background section, this invention provides a method for predicting microvascular invasion (MVI) in hepatocellular carcinoma. This method utilizes preoperative enhanced CT images, clinical indicators, and corresponding pathological results of lesions from hepatocellular carcinoma patients in actual clinical practice to establish a 3D convolutional neural network model based on deep learning algorithms to predict the MVI status of hepatocellular carcinoma. See details below. Figure 1 As shown, Figure 1 This is a schematic flowchart of a method for predicting microvascular invasion in hepatocellular carcinoma provided in an embodiment of the present invention, as shown below. Figure 1 As shown, the prediction method includes the following steps:

[0040] Step 110: Obtain preoperative imaging images, clinical indicators, and corresponding pathological results of lesions from treated patients.

[0041] Specifically, the preoperative imaging images are enhanced CT scans, specifically arterial phase CT images and venous phase CT images of the liver cancer lesions. The clinical indicators consist of data composed of 10 clinical characteristic indicators, including the patient's age, gender, liver function biochemical indicators, and tumor volume and maximum diameter related to the lesion. The liver function biochemical indicators include ascites, PT, ALB, TB, AST, and AFP.

[0042] Step 120: Based on the preoperative images, clinical indicators and corresponding pathological results of the lesions of the treated patients, the pre-constructed prediction model is trained. When the prediction model meets the preset standards, the prediction model is determined to be a microvascular invasion prediction model.

[0043] The pre-built prediction model in this application is a hybrid classification model, consisting of two 3D deep learning classification models: a single-input model based on image data and a dual-input model based on image data and clinical indicators. Therefore, the single-input model is also called the Image Only Model, with input data being hepatocellular carcinoma (LC) VOI data, specifically venous or arterial CT scan data. The dual-input model is also called the Image + Clinical Indicator (CI) Model, with input data being LC VOI data and clinical indicators (ClinicalIndex). In other words, the pre-built prediction model in this application is a multi-model combination structure, adaptable to real-world clinical scenarios, and utilizes multimodal data to obtain more accurate MVI classification prediction results.

[0044] Specifically, Figure 2 This is a schematic diagram of the prediction model, such as... Figure 2 As shown, the single-input model in this application includes a first single-input model (Image Only Model (Venous)) and a second single-input model (Image Only Model (Arterial)). The first single-input model takes a CT image of a liver cancer lesion in the venous phase as input and outputs a first classification result. The second single-input model takes a CT image of a liver cancer lesion in the arterial phase as input and outputs a second classification result. The dual-input model includes a first dual-input model (Image+CI Model (Venous)) and a second dual-input model (Image+CI Model (Arterial)). The first dual-input model takes CT images and clinical data of hepatocellular carcinoma lesions in the venous phase as input and outputs a third classification result. The second dual-input model takes CT images and clinical data of hepatocellular carcinoma lesions in the arterial phase as input and outputs a fourth classification result. Finally, the hepatocellular carcinoma microvascular invasion model uses an ensemble method (Ensembling) to determine the final classification result based on the first and second classification results of the single-input model and the third and fourth classification results of the dual-input model. This final classification result is used to predict whether there is microvascular invasion of hepatocellular carcinoma. Specifically, the classification results of the four models are averaged to obtain the final classification prediction result.

[0045] Figure 3 This is a schematic diagram of the structure of a single-input model, as shown below. Figure 3As shown, the single-input model consists of an image feature extraction module and a classification head module. The image feature extraction module is primarily built upon a 3D residual network structure (ResNet3D 18Bady). Preferably, this application uses 18 layers, i.e., R3D-18, to complete the 3D image feature extraction function. It should be noted that other 3D residual network structures can also be selected according to requirements; no limitation is made here. The classification head module is built according to the classification category, which is determined by the classification target. In this application, the classification target is to classify whether hepatocellular carcinoma has microvascular invasion. Therefore, the classification target is selected as 2, i.e., MVI+ (positive microvascular invasion) and MVI- (negative microvascular invasion). That is to say, this module completes the binary classification task based on the features extracted by the aforementioned image feature extraction module. The input data of the single-input model has been described above and will not be repeated here.

[0046] Figure 4 This is a schematic diagram of the structure of a dual-input model, as shown below. Figure 4 As shown, the dual-input model includes two input modules and one output model. The two input modules take image data and tabular clinical data as inputs, respectively. The image data, as previously described, consists of CT arterial or venous phase images, while the tabular clinical indicator data refers to the aforementioned clinical indicators, which are composed of 10 clinical feature indicators, as detailed in step 110. Therefore, the dual-input model includes two data processing modules: a 3D image data processing module and a structured data processing module. The 3D image data processing module processes the input image data, specifically the liver cancer cell data. Its basic structure is the same as the single-input model, built using a 3D convolutional neural network, specifically a 3D deep residual network. The difference lies in the number of features in the classifier head: c = 192, to extract 3D image features, whereas in the single-input model c = 2. The structured data processing module handles tabular clinical indicators, i.e., it processes the input clinical indicator data. This module divides the structured data into categorical data and continuous data for separate processing. Categorical data processing primarily involves embedding matrix mapping followed by a dropout layer, while continuous data processing mainly utilizes 1D batch normalization. The processed categorical and continuous data are then integrated through fully connected layers, with a feature count of 64 (c=64), completing the structured data processing. Based on the above description and... Figure 4 As can be seen, the output layers mainly consist of linear layers, batch normalization layers, and dropout layers. They integrate the 192 image features obtained from the image data processing module and the 64 data features obtained from the structured data processing module to finally obtain the classification result. The output module of the dual-input model also outputs two types of results: MVI+ (positive microvascular invasion) and MVI- (negative microvascular invasion). Furthermore, the specific structure of the LinBnDrop layer in the dual-input model is as follows... Figure 5 As shown.

[0047] In addition, this invention also generates a model activation map corresponding to the lesion area based on the patient's preoperative imaging images, clinical indicators, and microvascular invasion prediction model, specifically as follows: Figure 6 As shown, the highlighted areas in the image correspond to the actual lesion areas, used to interpret the classification results. The model activation map is an activation heatmap obtained by weighted summation of feature maps generated by convolutional layers of different categories. This activation heatmap can be used to interpret the model's classification results. Specifically, the categorical (CAT) and continuous (CONT) values ​​are processed separately and integrated into a classification prediction model based on fully connected layers. Subsequently, it is integrated with image features extracted by a 3D convolutional neural network (CNN) architecture to establish an interpretable binary classification model that can simultaneously display the classification structure and model activation map.

[0048] The preoperative imaging images, clinical indicators, and corresponding pathological results of lesions obtained from treated patients are divided into training and validation sets. The pre-constructed prediction model is trained using the training set and validated using the validation set. Specifically, for example... Figure 6 As shown, the area under the curve (AUC) is 0.82, indicating that the prediction model has high resolution. Therefore, it can be determined that the prediction model is a microvascular invasion model for hepatocellular carcinoma.

[0049] It should be noted that the model performance in this application is primarily evaluated from a classification perspective. In addition to the above... Figure 7 In addition to the evaluation methods mentioned above, the performance of this classification model can also be evaluated from different aspects using precision, recall, and F1-score, which will not be elaborated here.

[0050] Step 130: Obtain preoperative imaging images and clinical indicators of the patient to be predicted.

[0051] Step 140: Input the preoperative images and clinical indicators of the patient to be predicted into the microvascular invasion prediction model, and output the classification results to predict whether the patient has microvascular invasion of hepatocellular carcinoma.

[0052] In addition, it can generate model activation maps corresponding to the image of the lesion area to interpret the classification results.

[0053] This invention provides a method for predicting microvascular invasion (MVI) in hepatocellular carcinoma (HCC). The method involves acquiring preoperative imaging images, clinical indicators, and corresponding pathological results of lesions from treated patients; training a pre-constructed prediction model based on these data; and determining the model as a microvascular invasion prediction model when it meets preset criteria. The method also involves acquiring preoperative imaging images and clinical indicators of the patient to be predicted; inputting these data into the microvascular invasion prediction model; and outputting classification results to predict whether the patient has microvascular invasion in HCC. Microvascular invasion is a significant factor affecting the survival rate of HCC patients and is an important biological characteristic of HCC. Clinically, besides pathological detection of MVI, the prediction method provided in this invention obtains highly accurate MVI classification preoperatively using enhanced CT arteriovenous phase images and some routine clinical indicators. This method is adaptable to real-world clinical scenarios and will help in developing reasonable surgical and treatment plans, reducing postoperative recurrence rates, and improving the overall prognosis of patients.

[0054] The above describes the embodiments of the hepatocellular carcinoma microvascular invasion prediction method provided by the present invention. Other embodiments provided by the present invention will be described below.

[0055] Figure 8 This is a schematic diagram of the structure of a hepatocellular carcinoma microvascular invasion prediction device provided in an embodiment of the present invention, as shown below. Figure 8 As shown, the device includes a first acquisition module 801, a model building module 802, a second acquisition module 803, and a prediction module 804.

[0056] The first acquisition module 801 is used to acquire preoperative imaging images, clinical indicators and corresponding pathological results of lesions of treated patients.

[0057] The model building module 802 is used to train the pre-built prediction model based on the preoperative images, clinical indicators and corresponding pathological results of the lesions of the treated patients. When the prediction model meets the preset standards, the prediction model is determined to be a microvascular invasion prediction model.

[0058] The second acquisition module 803 is used to acquire preoperative imaging images and clinical indicators of the patient to be predicted.

[0059] The prediction module 804 is used to input the preoperative images and clinical indicators of the patient to be predicted into the microvascular invasion prediction model and output the classification results to predict whether the patient has microvascular invasion of hepatocellular carcinoma.

[0060] In one example, the microvascular invasion prediction model consists of a first single-input model and a second single-input model based on imaging images, and a first dual-input model and a second dual-input model based on imaging images and clinical indicators. The microvascular invasion prediction model uses an ensemble method to determine the classification result based on the first classification result of the first single-input model and the second single-input model, as well as the third classification result of the first dual-input model and the fourth classification result of the second dual-input model, in order to predict whether the patient under test has microvascular invasion of hepatocellular carcinoma.

[0061] In one example, prediction model 804 is also used to input preoperative images and clinical indicators of the patient to be predicted into the microvascular invasion prediction model to generate a model activation map corresponding to the image of the lesion area, which is used to interpret the classification results.

[0062] The functions performed by each component in the hepatocellular carcinoma microvascular invasion prediction device provided in this embodiment have been described in detail in any of the above method embodiments, and therefore will not be repeated here.

[0063] This invention provides a device for predicting microvascular invasion in hepatocellular carcinoma. It acquires preoperative imaging images, clinical indicators, and corresponding pathological results of lesions from treated patients. Based on these images, a pre-constructed prediction model is trained. When the prediction model meets preset standards, it is determined to be a microvascular invasion prediction model. The device also acquires preoperative imaging images and clinical indicators from patients to be predicted. These images and indicators are input into the microvascular invasion prediction model, which outputs classification results to predict whether microvascular invasion of hepatocellular carcinoma exists in the patient. This method adapts to real-world clinical scenarios, yielding more accurate classification prediction results and providing reference and convenience for clinical diagnosis, surgical planning, and prognostic assessment.

[0064] like Figure 9 As shown, an embodiment of the present invention provides an electronic device, including a processor 111, a communication interface 112, a memory 113, and a communication bus 114. The processor 111, the communication interface 112, and the memory 113 communicate with each other through the communication bus 114. The processor 111 includes a graphics processor and a central processing unit.

[0065] Memory 113 is used to store computer programs;

[0066] In one embodiment of the present invention, when the processor 111 executes the program stored in the memory 113, it implements the steps of the hepatocellular carcinoma microvascular invasion prediction method provided in any of the foregoing method embodiments.

[0067] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the hepatocellular carcinoma microvascular invasion prediction method provided in any of the foregoing method embodiments.

[0068] Compared with the prior art, the technical solution provided in this application has the following advantages:

[0069] 1. The prediction model in this application consists of two 3D deep learning classification models: a 3D deep convolutional neural network model based on individual CT images and a hybrid classification model based on CT images and conventional clinical indicators. This multi-model combination structure adapts to real-world clinical scenarios, utilizing multimodal data to obtain more accurate MVI classification prediction results, and simultaneously generating a comparison with the original CT scan of the lesion area. Figure 1 A corresponding model activation region map provides interpretable information for clinical practice.

[0070] 2. We used preoperative enhanced CT image data of hepatocellular carcinoma patients, clinical indicators and corresponding pathological results of lesions to establish two 3D convolutional neural network models based on deep learning algorithms to predict the MVI status of hepatocellular carcinoma. We verified its feasibility and effectiveness with actual clinical data. The comprehensive classification efficiency (AUC) of the system can reach 0.82.

[0071] 3. We used arterial and venous phase images of enhanced abdominal CT scans of patients, along with routine clinical indicators, to predict whether microvascular invasion exists in patients with hepatocellular carcinoma. The proposed MVI classification prediction model system is expected to provide reference and convenience for clinical diagnosis, surgical planning, prognostic assessment, and other related work.

[0072] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0073] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0074] The above are merely specific embodiments of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for predicting microvessel invasion of hepatocellular carcinoma, characterized by, The method includes the following steps: Obtain preoperative imaging images, clinical indicators, and corresponding pathological results of lesions from treated patients; Based on the preoperative images, clinical indicators, and corresponding pathological results of the lesions of the treated patients, the pre-constructed prediction model is trained. When the prediction model meets the preset criteria, the prediction model is determined to be a microvascular invasion prediction model. Obtain preoperative imaging images and clinical indicators of the patient to be predicted; The preoperative images and clinical indicators of the patient to be predicted are input into the microvascular invasion prediction model, and the classification results are output to predict whether the patient has microvascular invasion of hepatocellular carcinoma. The microvascular invasion prediction model consists of a first single-input model and a second single-input model based on imaging images, and a first dual-input model and a second dual-input model based on imaging images and clinical indicators. The microvascular invasion prediction model uses an ensemble method to determine the classification result based on the first classification result of the first single-input model and the second classification result of the second single-input model, as well as the third classification result of the first dual-input model and the fourth classification result of the second dual-input model, in order to predict whether the patient under test has microvascular invasion of hepatocellular carcinoma. The first single-input model inputs CT images of liver cancer lesions in the venous phase, and the second single-input model inputs CT images of liver cancer lesions in the arterial phase; the first dual-input model inputs CT images of liver cancer lesions in the venous phase and clinical data, and the second dual-input model inputs CT images of liver cancer lesions in the arterial phase and clinical data. The first single-input model and the second single-input model are composed of an image feature extraction module and a classification head module. The image feature extraction module is built based on a 3D residual network structure. The first dual-input model and the second dual-input model each include a 3D image data processing module, a structured data processing module, and an output layer. The 3D image data processing module processes the input image data. The structured data processing module processes the input clinical indicators. The 3D image data processing module includes a 3D deep residual network and a classification head. The structured data processing module includes an embedding matrix and a Dropout layer for processing categorical data in the clinical indicators. The structured data processing module also includes 1D batch normalization for processing continuous data in the clinical indicators. The structured data processing module further includes a fully connected layer for integrating the processed categorical data and continuous data. The output layer includes a linear layer, batch normalization, and a Dropout layer for integrating the data features output by the 3D image data processing module and the structured data processing module to obtain classification results.

2. The method of claim 1, wherein, The image is an enhanced CT image.

3. The method of claim 1, wherein, The clinical indicators include: patient age, gender, liver function biochemical indicators, and tumor volume and maximum diameter related to the lesion.

4. The method of claim 1, wherein, The method further includes: inputting the preoperative images and clinical indicators of the patient to be predicted into the microvascular invasion prediction model to generate a model activation map corresponding to the image of the lesion area, which is used to interpret the classification results.

5. A hepatocellular carcinoma microvessel invasion prediction device characterized by comprising: The device includes: The first acquisition module is used to acquire preoperative imaging images, clinical indicators, and corresponding pathological results of lesions of treated patients. The model building module is used to train a pre-built prediction model based on the preoperative images, clinical indicators and corresponding pathological results of the treated patients. When the prediction model meets the preset standards, the prediction model is determined to be a microvascular invasion prediction model. The second acquisition module is used to acquire preoperative imaging images and clinical indicators of the patient to be predicted. The prediction module is used to input the preoperative images and clinical indicators of the patient to be predicted into the microvascular invasion prediction model and output the classification results to predict whether the patient to be predicted has microvascular invasion of hepatocellular carcinoma. The microvascular invasion prediction model consists of a first single-input model and a second single-input model based on imaging images, and a first dual-input model and a second dual-input model based on imaging images and clinical indicators. The microvascular invasion prediction model uses an ensemble method to determine the classification result based on the first classification result of the first single-input model, the second classification result of the second single-input model, the third classification result of the first dual-input model, and the fourth classification result of the second dual-input model, in order to predict whether the patient under test has microvascular invasion of hepatocellular carcinoma. The first single-input model inputs CT images of liver cancer lesions in the venous phase, and the second single-input model inputs CT images of liver cancer lesions in the arterial phase; the first dual-input model inputs CT images of liver cancer lesions in the venous phase and clinical data, and the second dual-input model inputs CT images of liver cancer lesions in the arterial phase and clinical data. The first single-input model and the second single-input model are composed of an image feature extraction module and a classification head module. The image feature extraction module is built based on a 3D residual network structure. The first dual-input model and the second dual-input model each include a 3D image data processing module, a structured data processing module, and an output layer. The 3D image data processing module processes the input image data. The structured data processing module processes the input clinical indicators. The 3D image data processing module includes a 3D deep residual network and a classification head. The structured data processing module includes an embedding matrix and a Dropout layer for processing categorical data in the clinical indicators. The structured data processing module also includes 1D batch normalization for processing continuous data in the clinical indicators. The structured data processing module further includes a fully connected layer for integrating the processed categorical data and continuous data. The output layer includes a linear layer, batch normalization, and a Dropout layer for integrating the data features output by the 3D image data processing module and the structured data processing module to obtain classification results.

6. The apparatus of claim 5, wherein, The prediction model is also used to input the preoperative images and clinical indicators of the patient to be predicted into the microvascular invasion prediction model to generate a model activation map corresponding to the image of the lesion area, which is used to interpret the classification results.

7. An electronic device, comprising: It includes a processor, a communication interface, a memory, and a communication bus. The processor, communication interface, and memory communicate with each other through the communication bus. The processor includes a graphics processor and a central processing unit. Memory, used to store computer programs; When executing a program stored in memory, the processor implements the steps of the method for predicting microvascular invasion in hepatocellular carcinoma as described in any one of claims 1-4.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method for predicting microvascular invasion of hepatocellular carcinoma as described in any one of claims 1-4.

Images