Method and device for predicting postoperative nausea and vomiting in patients

Abstract

The present invention relates to the field of medical technology, and discloses a method and device for predicting postoperative nausea and vomiting of a patient, the method comprising: obtaining brain imaging data and clinical data of the patient; preprocessing and feature extraction of the brain imaging data and clinical data respectively to obtain brain structural features and clinical features; screening the brain structural features to obtain target screening brain structural features; constructing an imaging feature model based on patient labels and target screening brain structural features; screening clinical features to obtain target screening clinical features; constructing a clinical feature model based on patient labels and target screening clinical features; determining a logistic regression model based on the imaging feature model and the clinical feature model; determining an imaging-clinical nomogram based on the logistic regression model; and obtaining a prediction result based on the imaging-clinical nomogram. The present invention improves the accuracy of predicting postoperative nausea and vomiting of patients, and uses a machine learning model to automate prediction, reduce manual judgment, and improve efficiency.

Description

Technical Field

[0001] The present disclosure relates to the field of medical technology, and in particular to a method and device for predicting postoperative nausea and vomiting of patients. Background Art

[0002] The average incidence of postoperative nausea and vomiting (PONV) is 20%-40%, and the incidence of nausea and vomiting in high-risk patients can even reach 80%. In addition to causing severe discomfort to patients, PONV may also cause other medical complications, such as wound dehiscence, bleeding, electrolyte imbalance, and aspiration pneumonia, leading to delayed recovery and prolonged hospitalization for patients after lung surgery. Domestic and foreign studies have shown that the risk factors for PONV are: female, history of motion sickness, history of postoperative nausea and vomiting, non-smoking, and age <50 years old. These risk factors point the way to the occurrence of postoperative nausea and vomiting. Nevertheless, because these factors are relatively rough and have limited predictive efficacy, and due to the development of perioperative medicine and the need for patient rehabilitation quality, it is crucial to determine the cause and predisposing factors of PONV, prevention techniques, early diagnosis, and management strategies. At present, there is a lack of effective methods for predicting postoperative nausea and vomiting in patients.

[0003] In view of this, this application is filed. Summary of the invention

[0004] In order to solve the above technical problems or at least partially solve the above technical problems, the embodiments of the present disclosure provide a method and device for predicting postoperative nausea and vomiting of patients, which utilizes ultrasound imaging data to screen important features of vomiting, thereby improving the accuracy of predicting postoperative nausea and vomiting of patients, and uses machine learning models for automated prediction, thereby reducing manual judgment and improving efficiency.

[0005] In a first aspect, an embodiment of the present disclosure provides a method for predicting postoperative nausea and vomiting of a patient, the method comprising:

[0006] Obtain brain imaging data and clinical data of patients;

[0007] Preprocessing and feature extraction are performed on the brain imaging data and the clinical data respectively to obtain brain structural features and clinical features;

[0008] Using a first preset screening method to screen the brain structure feature to obtain a target screening brain structure feature;

[0009] Based on the patient label and the target, screening brain structure features builds an imaging feature model;

[0010] The clinical characteristics are screened using a second preset screening method to obtain target screening clinical characteristics;

[0011] Constructing a clinical feature model based on the patient label and the target screening clinical features;

[0012] Determining a logistic regression model based on the image feature model and the clinical feature model;

[0013] Based on the logistic regression model, determining an imaging-clinical nomogram;

[0014] Based on the imaging-clinical nomogram, the prediction results of nausea and vomiting were obtained.

[0015] In a second aspect, the present disclosure also provides a device for predicting postoperative nausea and vomiting of a patient, the device comprising:

[0016] A first acquisition module is used to acquire brain imaging data and clinical data of the patient;

[0017] A second acquisition module is used to preprocess and extract features from the brain image data and the clinical data to obtain brain structure features and clinical features;

[0018] A third acquisition module is used to screen the brain structure feature using a first preset screening method to obtain a target screening brain structure feature;

[0019] A first building module is used to build an image feature model based on the patient label and the target screened brain structure features;

[0020] A fourth acquisition module is used to screen the clinical characteristics using the second preset screening method to obtain target screening clinical characteristics;

[0021] A second building module is used to build a clinical feature model based on the patient label and the target screening clinical features;

[0022] A first determination module is used to determine a logistic regression model based on the image feature model and the clinical feature model;

[0023] A second determination module is used to determine an imaging-clinical nomogram based on the logistic regression model;

[0024] The fifth acquisition module is used to obtain nausea and vomiting prediction results based on the imaging-clinical nomogram.

[0025] In a third aspect, an embodiment of the present disclosure further provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the method for predicting postoperative nausea and vomiting of patients as described above.

[0026] In a fourth aspect, an embodiment of the present disclosure further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for predicting postoperative nausea and vomiting of patients as described above.

[0027] The disclosed embodiments provide a method for predicting postoperative nausea and vomiting of patients, which comprises the following steps: obtaining brain imaging data and clinical data of the patient; preprocessing and extracting features of the brain imaging data and clinical data respectively to obtain brain structural features and clinical features; then using a first preset screening method to screen the brain structural features to obtain target screening brain structural features; constructing an imaging feature model based on patient labels and target screening brain structural features; using a second preset screening method to screen clinical features to obtain target screening clinical features; constructing a clinical feature model based on patient labels and target screening clinical features; determining a logistic regression model based on the imaging feature model and the clinical feature model; determining an imaging-clinical nomogram based on the logistic regression model; and obtaining a nausea and vomiting prediction result based on the imaging-clinical nomogram, improving the accuracy of data through preprocessing, and then using a screening algorithm to screen important features of vomiting, thereby improving the accuracy of predicting postoperative nausea and vomiting of patients, and using a machine learning model for automated prediction to reduce manual judgment and improve efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and the originals and elements are not necessarily drawn to scale.

[0029] Figure 1 This is a flow chart of a method for predicting postoperative nausea and vomiting in a patient according to an embodiment of the present disclosure;

[0030] Figure 2 is a schematic diagram of the structure of an imaging-clinical nomogram in an embodiment of the present disclosure;

[0031] Figure 3 This is a schematic diagram of the structure of a device for predicting postoperative nausea and vomiting of patients in an embodiment of the present disclosure;

[0032] Figure 4 It is a structural schematic diagram of an electronic device in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0033] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, which are instead provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0034] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0035] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0036] Preoperative brain structural magnetic resonance imaging (MRI) is a non-invasive imaging technology that can provide high-resolution brain structural images, which is crucial for identifying brain structural features related to postoperative nausea and vomiting. Through the patient's preoperative imaging data, brain images are segmented and features are extracted to form researchable data. The extracted data are then statistically analyzed and combined with clinical data to form intuitive result charts. The constructed machine learning prediction model is then used to predict the probability of postoperative nausea and vomiting in patients.

[0037] In response to the above problems, the embodiments of the present disclosure provide a method for predicting postoperative nausea and vomiting of patients, which uses ultrasound image data to screen important features of image nausea and vomiting, thereby improving the accuracy of predicting postoperative nausea and vomiting of patients, and uses model-automated prediction to reduce manual judgment and improve detection efficiency.

[0038] Figure 1 : is a flow chart of a method for predicting postoperative nausea and vomiting of a patient in an embodiment of the present disclosure. The method can be executed by a device for predicting postoperative nausea and vomiting of a patient, which can be implemented in software and / or hardware, and can be configured in an electronic device. Figure 1 As shown, the method may specifically include the following steps:

[0039] S110. Obtain brain imaging data and clinical data of the patient.

[0040] Understandably, brain imaging data can be preoperative brain structural magnetic resonance imaging (MRI) data. Before obtaining MRI data, a brain sMRI scan is required to obtain high-resolution brain imaging data through professional magnetic resonance imaging equipment. Appropriate imaging parameters need to be set before the magnetic resonance imaging equipment scans, and the patient must remain still during the scan. Among them, imaging parameters include:

[0041] 1. Resolution: usually required to be less than 1 mm in order to clearly display the fine structure of brain tissue;

[0042] 2. Scanning sequence: Commonly used sequences include T1-weighted imaging (T1WI), T2-weighted imaging (T2WI) and proton density-weighted imaging (PDWI). T1WI can clearly show the contrast between gray matter and white matter, while T2WI and PDWI help to show the water content of brain tissue;

[0043] 3. Number of imaging layers: Usually, continuous cross-sectional images of the brain need to be obtained, with the number of layers ranging from 100 to 200 and the layer thickness generally ranging from 1 to 2 mm;

[0044] 4. Repetition time (TR): TR is the time interval between two pulse sequences. The selection of its value will affect the contrast of the image. It is generally selected between 2-5 seconds;

[0045] 5. Echo time (TE): TE is the time from pulse emission to echo signal arrival, which affects the contrast of the image. The value of TE is usually between 20-80 milliseconds.

[0046] After determining the imaging parameters, the scanning range and imaging time are set. The scanning range should include the entire brain, from the top of the skull to the base of the skull, to ensure that complete brain structure information is obtained. The imaging time needs to consider the number of imaging layers. Generally, it takes 10-20 minutes to complete a full-brain structural MRI scan. Under the premise of ensuring image quality, the imaging time can be shortened by optimizing the scanning parameters.

[0047] Among them, clinical data include but are not limited to height, weight, age, gender, education level, previous surgical history, anesthesia method, etc.

[0048] S120 , preprocessing and feature extraction are performed on the brain image data and the clinical data respectively to obtain brain image features and clinical features.

[0049] In one embodiment, the preprocessing and feature extraction of the brain image data to obtain brain structure features includes:

[0050] Retaining data without artifacts and tumor lesions in the brain image data, and preprocessing the retained brain image data to obtain preprocessed brain image data;

[0051] Among them, the data without artifacts and tumor lesions in the brain imaging data are retained, and the retained brain imaging data are processed using a preprocessing method based on the design matrix. Specifically, in brain imaging data analysis, covariates usually represent non-interesting signal sources that need to be controlled or removed. The purpose of removing covariates in the original data set X is to reduce the impact of non-neurogenic signals such as head movement, white matter, and cerebrospinal fluid signals on the analysis results. The design matrix M is a matrix used to represent covariates in statistical analysis. In brain imaging data analysis, the design matrix contains various covariates related to time series data. These covariates may include head movement parameters, physiological signals (such as white matter, cerebrospinal fluid signals), etc. Each column of the design matrix represents a specific covariate, and each row represents a time point. The design matrix M can be expressed as:

[0052] [m 11 ,m 12 ,...,m 1k ]

[0053] [m 21 ,m 22 ,...,m 2k ]

[0054] [m n1 ,m n2 ,...,m nk ]

[0055] Based on the design matrix, find a coefficient vector β to minimize the residual sum of squares RSS. The residual sum of squares RSS is expressed by the following formula:

[0056] RSS=‖X-Mβ‖ 2

[0057] Where ‖.‖ represents the Frobenius norm, β represents a coefficient vector representing the influence coefficient of the covariate on the noise signal of the retained brain imaging data X, and M represents the design matrix.

[0058] The method to minimize the residual sum of squares (RSS) can be to differentiate β and set the derivative to zero to obtain the following formula:

[0059]

[0060] Then, the normal equation is obtained by differentiating β:

[0061] M T Mβ=M T X

[0062] Among them, M Trepresents the transpose of M, β represents a coefficient vector, representing the influence coefficient of the covariate on the noise signal of the data in the retained brain imaging data X, M represents the design matrix, and X represents the retained brain imaging data.

[0063] Finally, the noise signal influence coefficient β is obtained through the normal equation. The noise signal influence coefficient β is expressed by the following formula:

[0064] β=(M T M) -1 M T X

[0065] The influence of the covariate can be removed from the retained brain image data X by using the noise signal influence coefficient β to obtain the preprocessed brain image data, where the preprocessed brain image data is X residual The following formula is expressed:

[0066] X residual =X-Mβ

[0067] Among them, X residual represents the preprocessed brain image data, X represents the retained brain image data, β represents the noise signal influence coefficient, and M represents the design matrix.

[0068] The pre-processed brain image data are extracted using brain structure analysis software to obtain brain structure features.

[0069] Specifically, the data without artifacts and tumor lesions in the brain image data (i.e., MRI data) obtained through the above steps are retained, and preprocessed in the above manner to obtain preprocessed brain image data. Then, brain image analysis software is used to remove noise, correct head movement, align and align the preprocessed brain image data for feature extraction to obtain brain structure features. Among them, brain structure features include the volume of each part of the brain area, brain surface area, cerebral cortex thickness, brain morphological features, etc., among which brain morphological features include curvature, folding and other features.

[0070] It should be noted that different types of brain imaging analysis software can be selected according to actual needs. Among them, brain imaging analysis software includes FreeSurfer, Statistical Parametric Mapping (SPM), Analysis of Functional NeuroImages (AFNI), and FMRIB's Software Library (FSL). The advantages and disadvantages of each brain imaging analysis software are described in detail below:

[0071] FreeSurfer is a commonly used brain structure analysis software that can automatically segment brain tissue and calculate indicators such as volume and surface area. The advantages are high automation and reliable results; the disadvantage is that it requires high computing resources.

[0072] SPM is a widely used brain imaging analysis software that can perform image preprocessing, feature extraction, and statistical analysis. Its advantages are comprehensive functions and strong community support; its disadvantage is that the learning curve is steep.

[0073] FSL is an open source brain imaging analysis software that includes a variety of image processing tools. Its advantages are open source, free, and scalable; its disadvantages are relatively less documentation and community support.

[0074] AFNI is another popular brain imaging analysis software, especially suitable for fMRI data analysis. Its advantages are fast speed and support for GPU acceleration; its disadvantage is that its interface is relatively complex.

[0075] In one embodiment, the preprocessing and feature extraction of the clinical data to obtain clinical features includes:

[0076] The clinical data are supplemented, and the supplemented clinical data are feature processed to obtain clinical features.

[0077] Specifically, the clinical data obtained may be incomplete, such as lack of gender or surgical history, etc. It is necessary to complete the clinical data obtained to obtain complete clinical data, and then extract features from the clinical data to obtain clinical features. Clinical feature extraction can be digital encoding of the obtained data, or conversion into a data type that can be recognized by the model. Exemplarily, the completion method can be the Multiple Imputation (MI) method: multiple imputation values ​​(usually m) are provided for each missing value, generating m complete data sets, which are analyzed separately and then the effects are merged. This method takes into account the uncertainty of missing values ​​and can provide unbiased and effective estimates.

[0078] S130. Using a first preset screening method to screen the brain structure features to obtain target screened brain structure features.

[0079] Among them, the first preset screening method includes at least one of the t-test method, the Mann-Whitney Utest statistical test method, the Pearson correlation coefficient method and the minimum absolute shrinkage and selection operator method, and the brain structure characteristics are screened using at least one of the above first preset screening methods to obtain the target screened brain structure characteristics.

[0080] In a specific embodiment, the first preset screening method is used to screen the brain structure feature to obtain the target screening brain structure feature, including:

[0081] Using t-test or Mann-Whitney Utest statistical test method to screen the brain structure characteristics to obtain the first screening brain structure characteristics;

[0082] The first screening feature is screened using the Pearson correlation coefficient method to obtain a second screening brain structure feature;

[0083] The two screening brain structure features are screened using the minimum absolute shrinkage and selection operator method to obtain the target screening brain structure features.

[0084] Exemplarily, according to data distribution and continuity, the t-test method or Mann-Whitney Utest statistical test method is used to screen out brain structural features with a probability of difference of less than 0.05 as the first screening brain structural features; specifically, the t-test method is used to screen out brain structural features with significant differences. The t-test is used to determine whether there is a significant difference in the means of two independent brain structural feature samples. Here, it is assumed that the comparison is between patient group X 1 and control group Y 2 structural characteristics of the brain.

[0085]

[0086] in Represent the mean values ​​of brain structure characteristic samples of the patient group and the control group, S p Indicates the combined standard deviation of two groups of samples, n 1 Represents sample X 1 The number of 2 Represents sample Y 2 The number of

[0087] Then, the Pearson correlation coefficient method is used to screen out the first screening brain structure features with a threshold greater than 9.0 as the second screening brain structure features; the Pearson correlation coefficient method can further screen out features with high correlation. Specifically, for two groups of consecutive first screening brain structure feature samples X 1 and X 2 , the linear correlation between two continuous variables was further determined by the Pearson correlation coefficient method.

[0088]

[0089] in, Respectively represent the mean of the first screening brain structure feature samples, X 1,i is a sample X 1 The data points in X 2,i is a sample X 2 The data points in rpearson Represents the Pearson correlation coefficient, which ranges from -1 to 1. The closer the value is to ±1, the stronger the linear relationship between the variables.

[0090] Finally, the minimum absolute shrinkage and selection operator method is used to screen the second screening brain structure features to obtain the target screening brain structure features.

[0091] S140, constructing an image feature model based on the patient label and the target screened brain structure features.

[0092] Among them, the patient label refers to whether the patient has nausea and vomiting after surgery. The patient label (whether there is nausea and vomiting after surgery) and the target screening brain structure features are input into the machine learning model, and the machine learning model is trained to obtain the image feature model. It can be understood that the input of the image feature model is the target screening brain structure features, and the output is the score of whether the patient has nausea and vomiting after surgery, that is, the image feature score. The machine learning models include support vector machine (SVM), random forest (Random Forest) and XGBoost classifier models.

[0093] Exemplarily, the image feature model of the present application can be a machine learning model constructed based on an adjustable hybrid kernel function, and the expression of the adjustable hybrid kernel function is:

[0094]

[0095] Among them, K(x i ,x j ) represents an adjustable hybrid kernel function, x i is the data point in the target screening brain structure feature, x j is the data point in the target screening brain structure characteristics, α is the kernel function factor, 0≤α≤1, which is used to control the influence of the kernel function on the model, σ is the width factor of the kernel function, and exp is the exponential function.

[0096] The expression of the image feature model is:

[0097]

[0098] Among them, f(x) represents the image feature model, represents the optimal parameter of the Lagrange multiplier, b * represents the optimal bias term of the image feature model, x i is the data point in the target screening brain structure feature, x j is the data point in the target screening brain structure feature, y i represents the true label of the i-th target brain structure feature sample, K(x i ,x j ) shows xi and x j An adjustable hybrid kernel function between , sign represents the sign function, When , the sign output is 1, When , the sign output is 0,

[0099] Among them, the expression of the loss function of the image feature model is:

[0100]

[0101] in, represents the loss function, represents the optimal parameter of the Lagrange multiplier, b * represents the optimal bias term of the image feature model, x i is the data point in the target screening brain structure feature, x j is the data point in the target screening brain structure feature, y i represents the true label of the i-th target brain structure feature sample, y j represents the true label of the j-th target screening brain structure feature sample, K(x i ,x j ) shows x i and x j is an adjustable hybrid kernel function between , and γ represents the penalty factor of the loss function.

[0102] The optimal parameters can be obtained by minimizing the loss function Through the optimal parameters and adjustable hybrid kernel function K(x i ,x j ) can get the optimal bias term b of the image feature model * ,

[0103]

[0104] Among them, b * represents the optimal bias term of the image feature model, represents the optimal parameters of the Lagrange multiplier, x i is the data point in the target screening brain structure feature, x j is the data point in the target screening brain structure feature, y i represents the true label of the i-th target brain structure feature sample, K(x i ,x j ) shows x i and x j An adjustable hybrid kernel function between .

[0105] Based on the loss function, the optimal parameters of the image feature model are determined. The optimal parameters are substituted into the image feature model to obtain a trained image feature model. When screening brain structure features for a new target, the trained image feature model can be used to output the result 0 or 1 to determine whether there is nausea and vomiting after surgery. 0 means no nausea and vomiting, and 1 means nausea and vomiting.

[0106] S150. Use the second preset screening method to screen the clinical characteristics to obtain target screening clinical characteristics.

[0107] Among them, the second preset screening method includes single factor analysis method, multi-factor analysis method, etc.;

[0108] In one embodiment, the second preset screening method is used to screen clinical characteristics to obtain target screening clinical characteristics, including:

[0109] The clinical characteristics were screened using univariate statistical analysis to obtain the target screening clinical characteristics.

[0110] Specifically, on the basis of the above-mentioned embodiment, the clinical characteristics were analyzed by univariate statistical analysis, and the clinical characteristics that were significantly correlated with the occurrence of postoperative nausea and vomiting (P<0.05) were screened out as the target screening clinical characteristics.

[0111] S160, constructing a clinical feature model based on the patient label and the target screening clinical features;

[0112] Among them, the patient label refers to whether the patient has nausea and vomiting after surgery. The patient label and the target screening clinical features are input into the machine learning model, and the machine learning model is trained to obtain the clinical feature model. It can be understood that the input of the clinical feature model is the target screening clinical features, and the output is the score of whether the patient has nausea and vomiting after surgery, that is, the clinical feature score. The machine learning models include support vector machine (SVM), random forest (Random Forest) and XGBoost classifier models.

[0113] It should be noted that the construction method of the clinical feature model is consistent with the construction method of the imaging feature model, and will not be repeated here.

[0114] S170. Determine a logistic regression model based on the image feature model and the clinical feature model.

[0115] In one embodiment, determining a logistic regression model based on the image feature model and the clinical feature model includes:

[0116] Obtaining an image feature score output by the image feature model;

[0117] Obtaining a clinical feature score output by the clinical feature model;

[0118] A logistic regression model is constructed based on the image feature scores and the clinical feature scores.

[0119] Specifically, the image feature model can use the XGBoost machine learning model to analyze and model the target screening brain structure features to generate an image feature score (Rad_Sig). This score represents the contribution of the image data to the predicted target. The clinical feature model can use the XGBoost machine learning model to analyze and model the target screening clinical features to generate a clinical feature score (Clinic_Sig). This score represents the contribution of the clinical data to the predicted target. The target screening brain structure features and target screening clinical features are standardized using Z-score regularization, and the standardized target screening brain structure features and target screening clinical features are input into the constructed and trained image feature model and clinical feature model to obtain the image feature score and clinical feature score. In order to facilitate data processing, the image feature score and clinical feature score can be normalized, and the normalized range is 0-1.

[0120] Then, a logistic regression model was constructed using the imaging feature scores corresponding to the target screening brain structure features and the clinical feature scores corresponding to the target screening clinical features. The constructed and trained logistic regression model was then used to predict and analyze patient nausea and vomiting to obtain the probability of the predicted risk of nausea and vomiting in patients.

[0121] S180, determining an imaging-clinical nomogram based on the logistic regression model;

[0122] Among them, the logistic regression model can be a mathematical model, and the image-clinical nomogram is essentially a way to visualize the logistic regression model, which can be specifically expressed by the following formula:

[0123] Risk=sigmoid(β0+β1×Clinic_Sig+β2×Rad_Sig)

[0124] Among them, Risk represents the predicted risk probability, β0 represents the intercept of the logistic regression model, β1 represents the coefficient of the clinical characteristic score, β2 represents the coefficient of the imaging characteristic score, Clnic_Sig represents the clinical characteristic score, and Rad_Sig represents the imaging characteristic score.

[0125] Exemplary, reference Figure 2As shown, Point represents the score, Clinic_Sig represents the clinical characteristic score, Rad_Sig represents the image characteristic score, Risk represents the predicted risk probability, and Total Points represents the total score, wherein the total score represents the sum of the score Point corresponding to the image characteristic score Clinic_Sig and the score Point corresponding to the clinical characteristic score Rad_Sig.

[0126] S190. Obtain prediction results of nausea and vomiting based on the imaging-clinical nomogram.

[0127] Among them, the nausea and vomiting prediction result can be the nausea and vomiting risk probability value output by the imaging-clinical nomogram. Based on the above embodiment, continue to refer to Figure 2 As shown, the score Point corresponding to the patient's image feature score Clinic_Sig and the score Point corresponding to the clinical feature score Rad_Sig are determined, the score Point corresponding to the image feature score Clinic_Sig and the score Point corresponding to the clinical feature score Rad_Sig are summarized, and the summarized total score is converted into a predicted risk probability Risk, that is, the nausea and vomiting prediction result.

[0128] Specifically, refer to Figure 2 , assuming that the Clinic_Sig value is 0.7, the corresponding Point is 42, the Rad_Sig value is 0.7, the corresponding Point is 60, the total score Total Points is 102, and the preset threshold is 0.5. At this time, the total score TotalPoints is 102, corresponding to a predicted risk probability of 0.7, indicating that the patient will have a 70% probability of postoperative nausea and vomiting.

[0129] The present embodiment provides a method for predicting postoperative nausea and vomiting of patients, which obtains brain imaging data and clinical data of the patients; preprocesses and extracts features of the brain imaging data and clinical data respectively to obtain brain structure features and clinical features; then uses a first preset screening method to screen the brain structure features to obtain target screening brain structure features; constructs an imaging feature model based on patient labels and target screening brain structure features; uses a second preset screening method to screen clinical features to obtain target screening clinical features; constructs a clinical feature model based on patient labels and target screening clinical features; determines a logistic regression model based on the imaging feature model and the clinical feature model; determines an imaging-clinical nomogram based on the logistic regression model; obtains nausea and vomiting prediction results based on the imaging-clinical nomogram, improves data accuracy through preprocessing, and then uses a screening algorithm to screen important features of vomiting, thereby improving the accuracy of predicting postoperative nausea and vomiting of patients, and uses a machine learning model for automated prediction to reduce manual judgment and improve efficiency.

[0130] Figure 3 FIG. 1 is a schematic diagram of a device for predicting postoperative nausea and vomiting in an embodiment of the present disclosure. Figure 3 As shown: the device includes: a first acquisition module 310, a second acquisition module 320, a third acquisition module 330, a first construction module 340, a fourth acquisition module 350, a second construction module 360, a first determination module 370, a second determination module 380 and a fifth acquisition module 390.

[0131] A first acquisition module is used to acquire brain imaging data and clinical data of the patient;

[0132] A second acquisition module is used to preprocess and extract features from the brain image data and the clinical data to obtain brain structure features and clinical features;

[0133] A third acquisition module is used to screen the brain structure feature using a first preset screening method to obtain a target screening brain structure feature;

[0134] A first building module is used to build an image feature model based on the patient label and the target screened brain structure features;

[0135] A fourth acquisition module is used to screen the clinical characteristics using the second preset screening method to obtain target screening clinical characteristics;

[0136] A second building module is used to build a clinical feature model based on the patient label and the target screening clinical features;

[0137] A first determination module is used to determine a logistic regression model based on the image feature model and the clinical feature model;

[0138] A second determination module is used to determine an imaging-clinical nomogram based on the logistic regression model;

[0139] The fifth acquisition module is used to obtain nausea and vomiting prediction results based on the imaging-clinical nomogram.

[0140] Among them, the expression of the image feature model is:

[0141]

[0142] Among them, f(x) represents the image feature model, represents the optimal parameter of the Lagrange multiplier, b * represents the optimal bias term of the image feature model, x i is the data point in the target screening brain structure feature, x j is the data point in the target screening brain structure feature, y i represents the true label of the i-th target brain structure feature sample, K(xi ,x j ) represents x i and x j An adjustable hybrid kernel function between , sign represents the sign function.

[0143] Wherein, the expression of the adjustable hybrid kernel function is:

[0144]

[0145] Among them, K(x i ,x j ) represents an adjustable hybrid kernel function, x i is the data point in the target screening brain structure feature, x j is the data point in the target screening brain structure characteristics, α is the kernel function factor, 0≤α≤1, which is used to control the influence of the kernel function on the model, σ is the width factor of the kernel function, and exp is the exponential function.

[0146] Optionally, the second acquisition module is further used to retain data without artifacts and tumor lesions in the brain image data, and preprocess the retained brain image data to obtain preprocessed brain image data;

[0147] The pre-processed brain image data are extracted using brain structure analysis software to obtain brain structure features.

[0148] Optionally, the third acquisition module is further used to screen the brain structure feature by using a t-test method or a Mann-Whitney Utest statistical test method to obtain a first screened brain structure feature;

[0149] The first screening feature is screened using the Pearson correlation coefficient method to obtain a second screening brain structure feature;

[0150] The two screening brain structure features are screened using the minimum absolute shrinkage and selection operator method to obtain the target screening brain structure features.

[0151] Optionally, the second acquisition module is further used to complete the clinical data, and perform feature processing on the completed clinical data to obtain clinical features.

[0152] Optionally, the fourth acquisition module is also used to screen the clinical characteristics using a univariate statistical analysis method to obtain target screening clinical characteristics.

[0153] Optionally, the first determining module is further used to obtain the image feature score output by the image feature model;

[0154] Obtaining a clinical feature score output by the clinical feature model;

[0155] A logistic regression model is constructed based on the image feature scores and the clinical feature scores.

[0156] Wherein, the expression of the logistic regression model is:

[0157] Risk=sigmoid(β0+β1×Clinic_Sig+β2×Rad_Sig)

[0158] Among them, Risk represents the predicted risk probability, β0 represents the intercept of the logistic regression model, β1 represents the coefficient of the clinical characteristic score, β2 represents the coefficient of the imaging characteristic score, Clinic_Sig represents the clinical characteristic score, and Rad_Sig represents the imaging characteristic score.

[0159] The device for predicting postoperative nausea and vomiting of patients provided in the embodiment of the present disclosure can execute the steps of the method for predicting postoperative nausea and vomiting of patients provided in the embodiment of the method of the present disclosure, and the execution steps and beneficial effects are no longer repeated here.

[0160] Figure 4 Schematic diagram of the structure of an electronic device in the embodiment of the present disclosure. Figure 4 , which shows a structural schematic diagram of an electronic device 500 suitable for implementing the embodiments of the present disclosure. Figure 4 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0161] like Figure 4 As shown, the electronic device 500 may include a processing device 501, a ROM 502, a RAM 503, a bus 504, an input / output (I / O) interface 505, an input device 506, an output device 507, a storage device 508, and a communication device 509. The processing device (e.g., a central processing unit, a graphics processor, etc.) 501 can perform various appropriate actions and processes to implement the method of the embodiment described in the present disclosure according to the program stored in the read-only memory (ROM) 502 or the program loaded from the storage device 508 to the random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the electronic device 500 are also stored. The processing device 501, the ROM 502, and the RAM 503 are connected to each other through the bus 504. The input / output (I / O) interface 505 is also connected to the bus 504.

[0162] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program contains a program code for executing the method shown in the flowchart, thereby implementing the patient postoperative nausea and vomiting prediction method as described above. In such an embodiment, the computer program can be downloaded and installed from the network through the communication device 509, or installed from the storage device 508, or installed from the ROM 502. When the computer program is executed by the processing device 501, the above-mentioned functions defined in the method of the embodiment of the present disclosure are executed.

[0163] It should be noted that the computer-readable medium disclosed above may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0164] The computer-readable medium may be included in the electronic device; or it may exist independently without being assembled into the electronic device. The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device: obtains the patient's brain imaging data and clinical data; preprocesses and extracts features of the brain imaging data and clinical data respectively to obtain brain structure features and clinical features; uses a first preset screening method to screen the brain structure features to obtain target screening brain structure features; constructs an imaging feature model based on the patient's label and the target screening brain structure features; uses a second preset screening method to screen the clinical features to obtain target screening clinical features; constructs a clinical feature model based on the patient's label and the target screening clinical features; determines a logistic regression model based on the imaging feature model and the clinical feature model; determines an imaging-clinical nomogram based on the logistic regression model; and obtains nausea and vomiting prediction results based on the imaging-clinical nomogram.

[0165] Optionally, when the above one or more programs are executed by the electronic device, the electronic device may also execute other steps described in the above embodiments.

[0166] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0167] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other to form a technical solution.

Claims

1. A method for predicting postoperative nausea and vomiting in patients, characterized in that: The method comprises: Obtain brain imaging data and clinical data of patients; Preprocessing and feature extraction are performed on the brain imaging data and the clinical data respectively to obtain brain structural features and clinical features; Using a first preset screening method to screen the brain structure feature to obtain a target screening brain structure feature; Based on the patient label and the target, screening brain structure features builds an imaging feature model; The clinical characteristics are screened using a second preset screening method to obtain target screening clinical characteristics; Constructing a clinical feature model based on the patient label and the target screening clinical features; Determining a logistic regression model based on the image feature model and the clinical feature model; Based on the logistic regression model, determining an imaging-clinical nomogram; Based on the imaging-clinical nomogram, the prediction results of nausea and vomiting were obtained. Wherein, the brain structure feature is screened by a first preset screening method to obtain a target screened brain structure feature, including: Using t-test or Mann-Whitney Utest statistical test method to screen the brain structure characteristics to obtain the first screening brain structure characteristics; The first screening brain structure feature is screened using the Pearson correlation coefficient method to obtain the second screening brain structure feature; The second screening brain structure feature is screened by using the minimum absolute shrinkage and selection operator method to obtain the target screening brain structure feature, Among them, the second preset screening method is used to screen the clinical characteristics to obtain the target screening clinical characteristics, including: The clinical characteristics were screened using univariate statistical analysis to obtain the target screening clinical characteristics. Among them, the expression of the image feature model is: ， in, represents the image feature model, represents the optimal parameters of the Lagrange multiplier, represents the optimal bias term of the image feature model, is the data point in the target screening brain structure feature, is the data point in the target screening brain structure feature, Indicates The goal is to filter the true labels of brain structure feature samples. express An adjustable hybrid kernel function between , sign represents the sign function, Wherein, the expression of the adjustable hybrid kernel function is: ， in, represents an adjustable hybrid kernel function, is the data point in the target screening brain structure feature, is the data point in the target screening brain structure feature, is the kernel function factor, , used to control the influence of the kernel function on the model, is the width factor of the kernel function, and exp is the exponential function.

2. The method according to claim 1, characterized in that Preprocessing and feature extraction are performed on the brain image data to obtain brain structure features, including: Retaining data without artifacts and tumor lesions in the brain image data, and preprocessing the retained brain image data to obtain preprocessed brain image data; The pre-processed brain image data are extracted using brain structure analysis software to obtain brain structure features.

3. The method according to claim 1, characterized in that The clinical data is preprocessed and feature extracted to obtain clinical features, including: The clinical data are supplemented, and the supplemented clinical data are feature processed to obtain clinical features.

4. The method according to claim 1, characterized in that: Determining a logistic regression model based on the image feature model and the clinical feature model includes: Obtaining an image feature score output by the image feature model; Obtaining a clinical feature score output by the clinical feature model; A logistic regression model is constructed based on the image feature scores and the clinical feature scores.

5. The method according to claim 1, characterized in that The expression of the logistic regression model is: ， in, represents the predicted risk probability, represents the intercept of the logistic regression model, The coefficient representing the clinical characteristic score, The coefficient representing the image feature score, represents the clinical characteristic score, Represents the image feature score.

6. A device for predicting postoperative nausea and vomiting of patients, using the method for predicting postoperative nausea and vomiting of patients according to any one of claims 1 to 5, characterized in that: include: A first acquisition module is used to acquire brain imaging data and clinical data of the patient; A second acquisition module is used to preprocess and extract features from the brain image data and the clinical data to obtain brain structure features and clinical features; A third acquisition module is used to screen the brain structure feature using a first preset screening method to obtain a target screening brain structure feature; A first building module is used to build an image feature model based on the patient label and the target screened brain structure features; A fourth acquisition module is used to screen the clinical characteristics using the second preset screening method to obtain target screening clinical characteristics; A second building module is used to build a clinical feature model based on the patient label and the target screening clinical features; A first determination module is used to determine a logistic regression model based on the image feature model and the clinical feature model; A second determination module is used to determine an imaging-clinical nomogram based on the logistic regression model; The fifth acquisition module is used to obtain nausea and vomiting prediction results based on the imaging-clinical nomogram.

Images