Indocyanine green metabolism prediction method and device based on preoperative clinical indexes and medium
By constructing ICG-T and ICG-R15 prediction models based on preoperative clinical indicators, the problem of ICG testing being unavailable in primary healthcare institutions has been solved, achieving greater accuracy and accessibility in liver function assessment and lowering the assessment threshold, especially for patients with poor liver function reserve.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-16
- Publication Date
- 2026-03-20
AI Technical Summary
Existing indocyanine green (ICG) clearance test equipment is expensive and difficult to make available in primary healthcare institutions. Some patients are allergic to ICG reagents, making it impossible to perform ICG testing directly, which increases the risk of liver resection surgery.
By acquiring the preoperative clinical characteristics and ICG test indicators of patients with primary liver cancer, Spearman correlation coefficient and wrap-around feature screening method were used to determine key feature data, and ICG-T and ICG-R15 prediction models were constructed. The ICG-T prediction value was used as an intermediate variable to improve the accuracy of the ICG-R15 prediction value.
In primary healthcare institutions lacking ICG testing equipment, accurate preoperative liver function assessment has been achieved, lowering the assessment threshold and improving the accuracy of ICG-R15 predictive values, especially for patients with poor liver function reserves.
Smart Images
Figure CN121709256A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of liver function evaluation, in particular to an indocyanine green metabolism prediction method based on preoperative clinical indicators, a device and a medium. BACKGROUND
[0002] Hepatocellular carcinoma is a common malignant tumor worldwide, and liver resection is the main cure for it. However, hepatocellular carcinoma patients are often accompanied by complications such as cirrhosis and viral hepatitis, which significantly increases the risk of postoperative liver failure, so accurate preoperative assessment of liver function reserve is crucial to reduce the risk of surgery.
[0003] Currently, the indocyanine green (ICG) clearance test is the standard for clinically evaluating liver function reserve. The indocyanine green excretion test detects the dynamic clearance rate of indocyanine green in the plasma by intravenous injection of the dye, reflecting the uptake and excretion capacity of the liver and effective blood flow. Clinical applications include reserve function evaluation of cirrhosis patients, risk prediction of liver resection, and postoperative monitoring of liver transplantation, etc. Its core indicator, 15-minute retention rate (ICG-R15), can provide a quantitative standard for surgical indications. However, the existing ICG clearance test equipment is expensive and difficult for primary medical institutions to popularize, and some patients are allergic to ICG reagents, which poses a risk of adverse reactions, so in many scenarios, direct indocyanine green (ICG) detection cannot be performed. SUMMARY
[0004] The purpose of the present application is to provide an indocyanine green metabolism prediction method based on preoperative clinical indicators, a device, equipment, a medium and a product, which can solve the problem that "in many scenarios, direct indocyanine green (ICG) detection cannot be performed".
[0005] To achieve the above-mentioned purpose, the present application provides the following solutions: In a first aspect, the present application provides an indocyanine green metabolism prediction method based on preoperative clinical indicators, comprising: Obtaining preoperative clinical characteristics and ICG detection indicators of a primary liver cancer patient, the preoperative clinical characteristics including a plurality of continuous characteristics and a plurality of discrete characteristics; Determining clinical characteristic data based on a plurality of the continuous characteristics, a plurality of the discrete characteristics and the ICG detection indicators; Obtaining an ICG-T predicted value based on the clinical characteristic data and a pre-constructed ICG-T prediction model; Determining an ICG-R15 predicted value based on the ICG-T predicted value, the clinical characteristic data and a pre-constructed ICG-R15 prediction model.
[0006] In an embodiment, the step of determining the clinical feature data based on the plurality of continuous features, the plurality of discrete features and the ICG detection index specifically comprises: analyzing the correlation between each of the continuous features and the ICG detection index to determine first key feature data; determining second key feature data based on the first key feature data and the discrete features, and determining the first key feature data and the second key feature data as the clinical feature data.
[0007] In an embodiment, the step of analyzing the correlation between each of the continuous features and the ICG detection index to determine first key feature data specifically comprises: using Spearman correlation coefficient to analyze the correlation between the continuous features and the ICG detection index; calculating the correlation coefficient of each of the continuous features and the ICG detection index; sorting all of the continuous features according to the size of the correlation coefficient, and taking the first preset number of continuous features as the first key feature data.
[0008] In an embodiment, the step of determining second key feature data based on the first key feature data and the discrete features specifically comprises: based on the first key feature data, verifying the contribution of the plurality of discrete features by a wrapping feature screening method to determine second key feature data.
[0009] In an embodiment, the step of pre-constructing the ICG-T prediction model specifically comprises: obtaining a first training data set; dividing the first training data set into a high-value sample subset and a low-value sample subset; identifying extreme samples from the high-value sample subset; sampling the remaining samples in the high-value sample subset to obtain a first sample set; sampling the samples in the low-value sample subset to obtain a second sample set; constructing a training set data using all of the extreme samples, the first sample set and the second sample set; training a preset machine learning model using the training set data to obtain an ICG-T prediction model.
[0010] In an embodiment, the step of pre-constructing the ICG-R15 prediction model specifically comprises: obtaining a second training data set; Train the preset machine learning model based on the second training data set and a balanced root mean square loss function to obtain an ICG-R15 prediction model.
[0011] In an embodiment, the continuous features are direct bilirubin, alpha-fetoprotein, albumin level, aspartate aminotransferase, coagulation function indicators, prothrombin time, blood sodium concentration, and glutamic-pyruvic transaminase.
[0012] In an embodiment, the discrete features are hepatitis B virus surface antigen.
[0013] In a second aspect, the present application also provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above method.
[0014] In a third aspect, the present application also provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executable on a processor to implement the above method.
[0015] According to the specific embodiments provided in the present application, the following technical effects are disclosed: The present application provides an indocyanine green metabolism prediction method based on preoperative clinical indicators, which obtains preoperative clinical characteristics and ICG detection indicators of a primary liver cancer patient, wherein the preoperative clinical characteristics include a plurality of continuous features and a plurality of discrete features; determines clinical characteristic data based on the plurality of continuous features, the plurality of discrete features, and the ICG detection indicators; in the present application, preoperative clinical characteristics that are easy to obtain and have high popularity are selected, and preoperative clinical characteristics that have significant correlation with the ICG detection indicators are selected as the clinical characteristic data, thereby ensuring the effectiveness and representativeness of the feature data input into the model.
[0016] The present application obtains an ICG-T prediction value based on the clinical characteristic data and a pre-constructed ICG-T prediction model; and determines an ICG-R15 prediction value based on the ICG-T prediction value, the clinical characteristic data, and a pre-constructed ICG-R15 prediction model. The present application first predicts an ICG-T prediction value by using the ICG-T prediction model, and then uses the ICG-T prediction value as an intermediate variable to predict an ICG-R15 prediction value together with the clinical characteristic data, which can improve the prediction accuracy of the ICG-R15 prediction value; at the same time, the present application can predict the two key liver function reserve parameters of the ICG-T prediction value and the ICG-R15 prediction value through the clinical characteristic data, the ICG-T prediction model, and the ICG-R15 prediction model, thereby reducing the evaluation threshold and enabling primary medical institutions without ICG detection equipment to also perform accurate preoperative liver function evaluation. BRIEF DESCRIPTION OF DRAWINGS
[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings described below only constitute some of the embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.
[0018] Figure 1 A method flow chart of a preoperative clinical index-based indocyanine green metabolism prediction method according to an embodiment of the present application; Figure 2 A flowchart of a preoperative clinical index-based indocyanine green metabolism prediction method according to an embodiment of the present application; Figure 3 A real data distribution diagram of an ICG-T index of a preoperative clinical index-based indocyanine green metabolism prediction method according to an embodiment of the present application; Figure 4 A real data distribution diagram of an ICG-R15 index of a preoperative clinical index-based indocyanine green metabolism prediction method according to an embodiment of the present application; Figure 5 A flowchart of constructing training set data of a preoperative clinical index-based indocyanine green metabolism prediction method according to an embodiment of the present application; Figure 6 A structure diagram of an MLP network of a preoperative clinical index-based indocyanine green metabolism prediction method according to an embodiment of the present application; Figure 7 A structure diagram of a computer device according to an embodiment of the present application. DETAILED DESCRIPTION
[0019] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments only constitute some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.
[0020] The above purposes, features and advantages of the present application can be more obvious and easy to understand. The present application will be described in further detail below with reference to the drawings and specific embodiments.
[0021] Referring to Figure 1 The present application provides a preoperative clinical index-based indocyanine green metabolism prediction method, comprising the following steps: Step 100: obtaining preoperative clinical characteristics and ICG detection indexes of a primary liver cancer patient, wherein the preoperative clinical characteristics include a plurality of continuous characteristics and a plurality of discrete characteristics; Step 200: determining clinical feature data based on the plurality of continuous features, the plurality of discrete features, and the ICG detection index; Step 300: obtaining an ICG-T prediction value based on the clinical feature data and the pre-constructed ICG-T prediction model; Step 400: determining an ICG-R15 prediction value based on the ICG-T prediction value, the clinical feature data, and the pre-constructed ICG-R15 prediction model.
[0022] In step 100, the preoperative clinical features refer to a series of physiological, biochemical, and pathological indexes obtained through routine examination before the patient receives a liver resection surgery. The preoperative clinical feature examination is a routine examination item, which is easy to obtain and closely related to the liver function status, thereby providing a key and reliable data basis for subsequent model construction.
[0023] Preferably, in the data collection stage, information of a plurality of patients diagnosed as primary liver cancer is collected, and 23 preoperative clinical features of each patient are counted, including 18 continuous features and 5 discrete features.
[0024] It should be noted that the continuous features refer to indexes that can change continuously in a numerical range, such as direct bilirubin, alpha-fetoprotein, and albumin level, and other laboratory test values; and the discrete features refer to indexes that exhibit limited categories or states, such as negative or positive state of hepatitis B virus surface antigen.
[0025] In the embodiments of the present application, in the data collection stage, the ICG detection index of the patient with primary liver cancer is obtained, the ICG detection index is ICG-R15, and the ICG-R15 refers to the indocyanine green 15-minute retention rate, which is a key index for clinically evaluating liver reserve function and predicting the risk of liver function failure after liver resection. In other embodiments, the ICG detection index can also be ICG-R10 or ICG-T, etc.
[0026] In step 200, specifically, the step of determining the clinical feature data based on the plurality of continuous features, the plurality of discrete features, and the ICG detection index specifically includes: analyzing the correlation between each continuous feature and the ICG detection index to determine first key feature data; determining second key feature data based on the first key feature data and the discrete features, and determining the first key feature data and the second key feature data as the clinical feature data.
[0027] More specifically, the step of analyzing the correlation between each continuous feature and the ICG detection index to determine the first key feature data specifically includes: using Spearman correlation coefficient (Rho) to analyze the correlation between the continuous feature and the ICG detection index; calculating the correlation coefficient of each continuous feature and the ICG detection index; according to the size of the correlation coefficient, all continuous features are ranked in high and low, and the first preset number of continuous features are taken as the first key feature data.
[0028] In this step, for the numerical continuous feature, Spearman correlation coefficient is used to analyze the correlation between the conventional preoperative clinical index and ICG-R15, and the calculation formula of the Spearman correlation coefficient is as follows: Wherein, is the numerical value of the Spearman correlation coefficient, and the calculation result is between-1 and +1, which is used to measure the strength and direction of the monotonic relationship between two variables; is the sample size; is the rank difference of the i-th pair of data on the variable and .
[0029] It should be noted that the rank difference is calculated as follows: assuming , the values in are arranged in ascending order, and each numerical value is assigned a corresponding rank, then and The rank difference value of is 2. The calculation formula uses the sum of the square of the rank difference to measure the correlation between two variables, if the rank difference of two variables always increases or decreases at the same time, they are positively correlated, and the closer the correlation coefficient is to 1; if the rank difference of one variable increases and the rank difference of the other variable decreases, they are negatively correlated, and the closer the correlation coefficient is to-1.
[0030] In order to exclude the problem of self-correlation between continuous features, the Spearman correlation coefficient between continuous features can also be calculated, if the Rho between two continuous features is greater than 0.9, it is considered that the two are highly self-correlated features, at this time, the Spearman correlation coefficient of the two with the ICG-R15 index is calculated respectively, and the feature with higher numerical value is selected, and the other feature is excluded. The present application calculates the Spearman correlation coefficient between continuous features, if two features are highly correlated, the feature with higher correlation with ICG-R15 index is retained, so as to avoid information redundancy.
[0031] Further, the correlation coefficients of all continuous features and the ICG-R15 indicator are calculated respectively, and are sorted from large to small, and the top pre-set number of continuous features are screened out, preferably, the top 8 continuous features are screened out as the first key feature data.
[0032] More specifically, the step of determining the second key feature data based on the first key feature data and the discrete feature includes: based on the first key feature data, verifying the contribution of the plurality of discrete features by the package feature screening method to determine the second key feature data.
[0033] In this step, for the discrete feature, the package feature screening method is used for screening. The specific process is: on the basis of the 8 selected continuous features, each remaining discrete feature is sequentially added to the feature set (i.e. the first key feature data) as a candidate feature, and after adding one discrete feature each time, the decision tree model is used and the model performance is evaluated by cross-validation. The final experimental results show that after adding the discrete feature of hepatitis B virus surface antigen, the prediction performance of the model reaches the best, so it is determined as the second key feature data.
[0034] In the present application, 8 continuous features and 1 discrete feature are selected as analysis indicators. The continuous features are direct bilirubin (DB0), alpha-fetoprotein (AFP), albumin level (ALB), aspartate aminotransferase (AST), coagulation function indicator (INR), prothrombin time (PT), blood sodium concentration (Na+) and glutamic-pyruvic transaminase (ALT), and the discrete feature is hepatitis B virus surface antigen (HBsAg).
[0035] Referring to Figure 2 The present application proposes a two-stage prediction model, which first inputs the screened 9 clinical feature data into the ICG-T prediction model to predict the ICG-T predicted value, and then uses the ICG-T predicted value as an intermediate variable to jointly predict the ICG-R15 predicted value with the clinical feature data.
[0036] In the process of statistical analysis, it is observed that the ICG-T indicator of the patient is highly related to the ICG-R15 indicator. The ICG-T indicator refers to the time when the concentration of ICG in the body decays to half after intravenous injection in the indocyanine green clearance test. The smaller the T value, the shorter the time for ICG to be metabolized to half after injection into the patient's body, and the stronger the patient's liver metabolic function, which is also a direct indicator reflecting the patient's liver reserve function. There is no direct calculation relationship between ICG-T and ICG-R15 value, but it is highly close to linear relationship, and the Pearson correlation coefficient of the two reaches 0.89, that is, if the value of ICG-T is obtained as the input of the ICG-R15 prediction model, the prediction effect of the ICG-R15 predicted value can be improved.
[0037] Since the data distribution range of the ICG-T index is smaller than that of the ICG-R15 index, the accuracy of the prediction of the ICG-T by the machine learning model is generally higher, and the error is smaller. If the ICG-T is first predicted by the machine learning model, and then the ICG-T is used as an intermediate variable to predict the ICG-R15 together with the clinical feature data, the prediction accuracy of the predicted value of the ICG-R15 can be improved.
[0038] Referring to Figures 3 to 4 , it can be observed from the data distribution of the ICG-R15 and the ICG-T that both of them exhibit obvious long-tail distribution characteristics. Therefore, the present application designs corresponding optimization strategies for the ICG-T prediction model and the ICG-R15 prediction model, respectively, as follows: In step 300, the step of pre-constructing the ICG-T prediction model, specifically comprising: obtaining a first training data set; dividing the first training data set into a high-value sample subset and a low-value sample subset; identifying extreme samples from the high-value sample subset; sampling the remaining samples in the high-value sample subset to obtain a first sample set; sampling the samples in the low-value sample subset to obtain a second sample set; constructing a training set data using all the extreme samples, the first sample set and the second sample set; and training a preset machine learning model using the training set data to obtain the ICG-T prediction model. The training set data includes clinical feature data of multiple primary liver cancer patients and corresponding ICG-T true values, which are used to train the ICG-T prediction model.
[0039] Among them, the first training data set is divided into a high-value sample subset and a low-value sample subset according to the median of the ICG-T values of all samples.
[0040] For example, referring to Figure 5 In the construction process of the ICG-T prediction model, a training set sampling strategy is adopted, and the specific steps are as follows: all samples are divided into two parts according to the median of the ICG-T values. In the part greater than the median, the samples with T>7 are determined as extreme data, a total of 88 examples, all of which are included in the training set data. The remaining samples in the high-value sample subset are sampled, that is, the part with T≤7 is supplemented with sampling, so that the sampling amount in the high-value sample subset accounts for 35% of the total sample amount. In the part less than the median, a total of 35% of the total sample amount is also randomly sampled. Finally, the training set data accounts for 70% of the total samples, and the remaining 30% is used as a test set. This sampling strategy helps to reduce the influence of extreme values on model training.
[0041] Step 400, the step of pre-constructing the ICG-R15 prediction model, specifically includes: obtaining a second training dataset; training a pre-defined machine learning model based on the second training dataset and a balanced root mean square loss function to obtain the ICG-R15 prediction model. The second training dataset includes clinical characteristic data, ICG-T predicted values, and corresponding ICG-R15 true values of multiple patients with primary liver cancer, used to train the ICG-R15 prediction model.
[0042] In the construction of the ICG-R15 prediction model, a balanced root mean square function is introduced as the loss function, and its calculation formula is as follows: In the formula, The true ICG-R15 value of the sample. The ICG-R15 predicted by the model. This represents the true ICG-R15 values for all samples within the batch. The temperature parameter is set to 4.
[0043] It should be noted that both the machine learning models in the ICG-T and ICG-R15 prediction models use MLP networks. See [link / reference] Figure 6 The MLP network consists of three linear fully connected layers and two dropout layers. The number of nodes in the fully connected layers are 64, 32, and 1, respectively. The layers with 64 and 32 nodes are hidden layers, and the layer with 1 node is the output layer.
[0044] This application provides a reliable method for assessing liver function reserve using readily available clinical indicators with high accuracy. Furthermore, by incorporating a long-tail data optimization strategy, this method can more accurately predict ICG-R15 values in patients with poor liver reserve, thereby assessing whether a patient is eligible for liver resection and the extent of resection. This is of great significance for the preoperative assessment of patients with severe liver cancer.
[0045] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 7As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and databases. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media to run. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection.
[0046] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0047] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.
[0048] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.
[0049] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.
[0050] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.
[0051] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).
[0052] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0053] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0054] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.
Claims
1. A method for predicting indocyanine green metabolism based on preoperative clinical indicators, characterized in that, include: The preoperative clinical characteristics and ICG test indicators of patients with primary liver cancer were obtained. The preoperative clinical characteristics included multiple continuous features and multiple discrete features. Clinical feature data are determined based on multiple continuous features, multiple discrete features, and ICG detection indicators; ICG-T predicted values are obtained based on the clinical characteristic data and the pre-built ICG-T prediction model; The ICG-R15 predicted value is determined based on the ICG-T predicted value, the clinical characteristic data, and the pre-constructed ICG-R15 prediction model.
2. The indocyanine green metabolism prediction method based on preoperative clinical indicators according to claim 1, characterized in that, The step of determining clinical feature data based on multiple continuous features, multiple discrete features, and the ICG detection index specifically includes: Analyze the correlation between each of the continuous features and the ICG detection indicators to determine the first key feature data; Based on the first key feature data and the discrete features, the second key feature data is determined, and the first key feature data and the second key feature data are determined as clinical feature data.
3. The indocyanine green metabolism prediction method based on preoperative clinical indicators according to claim 2, characterized in that, The step of analyzing the correlation between each of the continuous features and the ICG detection index to determine the first key feature data specifically includes: The correlation between the continuous features and the ICG detection indicators was analyzed using Spearman correlation coefficient. Calculate the correlation coefficient between each of the continuous features and the ICG detection index; Based on the magnitude of the correlation coefficient, all the continuous features are sorted from high to low, and the first preset number of continuous features are taken as the first key feature data.
4. The indocyanine green metabolism prediction method based on preoperative clinical indicators according to claim 3, characterized in that, The step of determining the second key feature data based on the first key feature data and the discrete features specifically includes: Based on the first key feature data, the contribution of multiple discrete features is verified by a wrap-around feature screening method to determine the second key feature data.
5. The indocyanine green metabolism prediction method based on preoperative clinical indicators according to claim 1, characterized in that, The steps for pre-constructing the ICG-T prediction model specifically include: Obtain the first training dataset; The first training dataset is divided into a high-value sample subset and a low-value sample subset; Extreme samples are identified from the high-value sample subset; The remaining samples in the high-value sample subset are sampled to obtain the first sample set; Samples from the low-value sample subset are sampled to obtain a second sample set; The training set data is constructed using all the extreme samples, the first sample set, and the second sample set; The training set data is used to train a preset machine learning model to obtain an ICG-T prediction model.
6. The method for predicting indocyanine green metabolism based on preoperative clinical indicators according to claim 1, characterized in that, The steps for pre-constructing the ICG-R15 prediction model specifically include: Obtain the second training dataset; The preset machine learning model is trained based on the second training dataset and a balanced root mean square loss function to obtain the ICG-R15 prediction model.
7. The method for predicting indocyanine green metabolism based on preoperative clinical indicators according to claim 1, characterized in that, The continuous characteristics are direct bilirubin, alpha-fetoprotein, albumin levels, aspartate aminotransferase, coagulation function indicators, prothrombin time, serum sodium concentration, and alanine aminotransferase.
8. The method for predicting indocyanine green metabolism based on preoperative clinical indicators according to claim 1, characterized in that, The discrete characteristic is hepatitis B virus surface antigen.
9. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the indocyanine green metabolism prediction method based on preoperative clinical indicators as described in any one of claims 1-8.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the indocyanine green metabolism prediction method based on preoperative clinical indicators as described in any one of claims 1-8.