CN112669929B

CN112669929B - Crohn's disease infliximab drug effect prediction method and terminal equipment

Info

Publication number: CN112669929B
Application number: CN202011625901.1A
Authority: CN
Inventors: 黄炳升; 袁程朗; 李雪华; 钟英奎; 张乃文; 张洪源; 罗梓欣; 冯盛宇; 曾英候; 陈白莉; 冯仕庭; 陈旻湖; 李子平
Original assignee: Shenzhen University
Current assignee: Shenzhen University
Priority date: 2020-12-30
Filing date: 2020-12-30
Publication date: 2021-08-03
Anticipated expiration: 2040-12-30
Also published as: CN112669929A

Abstract

The invention relates to a method for predicting drug effect of Rituximab of Crohn's disease and terminal equipment, wherein the method comprises the following steps: acquiring an artery phase scanning image of an intestinal tract focus, and extracting an image omics feature from the artery phase scanning image; and inputting the characteristics of the image group into a trained prediction model to obtain a drug effect prediction result, wherein the prediction model is established by using a support vector machine classifier. The method comprises the steps of obtaining an artery phase scanning image of the intestinal tract focus, extracting the image omics characteristics from the artery phase scanning image, and inputting the image omics characteristics into a trained prediction model, so that the drug effect of the Riximab infliximab caused by the Crohn disease can be accurately obtained.

Description

Crohn's disease infliximab drug effect prediction method and terminal equipment

Technical Field

The invention relates to the technical field of drug effect monitoring, in particular to a drug effect prediction method of Crohn's disease infliximab, a computer-readable storage medium and terminal equipment.

Background

Crohn's Disease (CD) is a chronic Inflammatory Bowel Disease (IBD) whose etiology and pathogenesis are not completely clear, and the clinical characteristics of IBD are chronic course, easy recurrence, many complications and high disability rate, which seriously affect the quality of life of a sample.

At present, no method for completely curing CD exists clinically. Moderate or severe active CD is usually treated with glucocorticoid, but some samples are not effective in hormonal treatment and some samples are hormone-dependent. None of the existing consensus suggests the use of hormones as long-term maintenance therapy for CD, and these parts of the sample that are ineffective or dependent on hormonal therapy have traditionally been treated with immunosuppressive agents. Thiopurine drugs (azathioprine or mercaptopurine) are the main drugs for maintenance therapy at present, and can effectively maintain long-term release of evacuation hormones. However, thiopurine drugs have dose-effect relationship and great individual difference, low dose can affect curative effect, large dose can increase the incidence rate of serious adverse reactions, and adverse reactions need to be strictly monitored in the treatment process.

The targeting of biological agents to specific inflammatory pathways provides a new alternative to the selection of traditional treatment regimens. Evidence-based medical evidence suggests that biologies are suitable for CD specimens where hormonal or immunosuppressive therapy is ineffective, hormone dependent, or intolerant; on the other hand, early biologics-enhanced treatment in moderate to severe CD has been shown to achieve mucosal healing. The monoclonal antibody of anti-Tumor necrosis factor alpha (TNF-alpha) can effectively promote the healing of the intestinal mucosa of a CD sample and reduce the incidence rate of disabling complications, and the Infliximab monoclonal antibody (IFX) is the most widely used monoclonal antibody of anti-Tumor necrosis factor alpha at present. However, 13% -40% of samples treated with IFX do not respond to the drug early in the treatment, and IFX is expensive, has unpredictable side effects, including increased risk of infection and lymphoproliferative disease, nervous system damage, and the like.

Therefore, how to predict the drug effect of the Rib fever infliximab in Crohn's disease in advance is a problem which needs to be solved urgently.

Disclosure of Invention

In view of the above-mentioned shortcomings of the prior art, the present invention aims to provide a method for predicting the efficacy of infliximab caused by crohn's disease, a computer-readable storage medium, and a terminal device, which are intended to predict and screen out samples that are not susceptible to IFX in advance.

The embodiment of the invention provides a method for predicting drug effect of England lizumab for Crohn's disease in the first aspect, which comprises the following steps:

acquiring an artery phase scanning image of an intestinal tract focus, and extracting an image omics feature from the artery phase scanning image;

and inputting the characteristics of the image group into a trained prediction model to obtain a drug effect prediction result, wherein the prediction model is established by using a support vector machine classifier.

In a second aspect of the embodiments of the present invention, a computer-readable storage medium is provided, where a drug effect prediction program of infliximab due to crohn's disease is stored, and when the drug effect prediction program is executed by a processor, the steps in the drug effect prediction method of infliximab due to crohn's disease are implemented.

A third aspect of the embodiments of the present invention provides a terminal device, where the terminal device includes a processor, a memory, and a drug effect prediction program of infliximab due to crohn's disease, the drug effect prediction program being stored in the memory and executable on the processor, and the processor implements the steps in the drug effect prediction method of infliximab due to crohn's disease as described above when executing the drug effect prediction program of infliximab due to crohn's disease.

Has the advantages that: the invention provides a method for predicting the drug effect of Riximab infliximab caused by Crohn's disease.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without any inventive work.

Fig. 1 is a flowchart of a method for predicting drug efficacy of infliximab caused by crohn's disease according to an embodiment of the present invention;

fig. 2 is a general flowchart of a drug efficacy prediction system for infliximab caused by crohn's disease according to an embodiment of the present invention;

FIG. 3 is a diagram of ROI and VOI of a CET image provided by an embodiment of the present invention;

FIG. 4 is a statistical chart showing the difference between the laboratory index CRP in the responder group and the non-responder group according to the embodiment of the present invention;

FIG. 5 is a statistical plot of the difference between the laboratory index ALB in the responder group and the non-responder group, provided by an embodiment of the present invention;

fig. 6 shows that features are screened according to feature skewness, and cross validation is performed by using a leave-one-out method to establish an image omics prediction model according to the feature skewness;

figure 7 clinical features + imagomics Nomogram provided by an embodiment of the present invention;

FIG. 8a is a schematic representation of the strength characteristics of an embodiment of the present invention; 8b is a schematic diagram of shape characteristics; 8c is a schematic diagram of texture features; 8d is a wavelet characteristic diagram;

fig. 9 is a schematic diagram of a terminal device structure according to an embodiment of the present invention.

Detailed Description

To facilitate an understanding of the invention, the invention will now be described more fully with reference to the accompanying drawings. Preferred embodiments of the present invention are shown in the drawings. This invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

Previous studies show that some indicators such as clinical data, serum, stool markers, etc. can be used to predict the efficacy of IFX, but the conclusion is not consistent.

Based on this, the present invention provides a solution to the above technical problem, and the details thereof will be explained in the following embodiments.

Referring to fig. 1, fig. 1 is a flowchart of a method for predicting drug efficacy of infliximab due to crohn's disease according to an embodiment of the present invention. The method comprises the following steps:

s10, obtaining an artery phase scanning image of the intestinal tract focus, and extracting the iconomics characteristics from the artery phase scanning image.

Specifically, in conjunction with FIG. 3, an arterial phase scan image of the intestinal tract lesion may be obtained by semi-automatically segmenting the intestinal tract lesion as a Region of Interest (ROI) of the study using ITK-SNAP software (open source software; www.itk-SNAP. org) on the CTE enhanced scan arterial phase image, and then generating the VOI based on a Region growing algorithm.

In this embodiment, the acquisition of the artery phase scan image of the intestinal tract lesion may also be obtained by identification of a trained image acquisition model.

In this embodiment, a cinematology feature is extracted from the arterial phase scan image, wherein the cinematology feature comprises: the intensity, shape, texture and wavelet features are quantized as shown in figures 8a to 8 d. It is readily understood that by means of the extracted influenceric features, the characteristics of the lesion read therefrom can be read.

And S20, inputting the characteristics of the image group into a trained prediction model to obtain a drug effect prediction result, wherein the prediction model is built by using a support vector machine classifier.

Specifically, the acquired influencer characteristics are input into a trained prediction model, and a pharmacodynamic prediction result is obtained through recognition of the prediction model. It is easy to understand that the prediction model is obtained after the training of the influencer characteristics, so that the iconomics characteristics to be predicted can be input into the prediction model, and then the pharmacodynamic prediction result is obtained.

In this embodiment, the prediction model is built for a Support Vector Machines (SVM) classifier. The SVM is a generalized linear classifier for binary classification of sample data according to a supervised learning mode, and can better solve the problems of small samples, nonlinearity, high-dimensional pattern recognition and the like. Wherein the modeling process can be implemented in Python3.5, and each classifier is provided by a sklern library (https:// scimit-lern. org/stable). No a priori knowledge is available for reference due to the choice of the prediction model parameters. Therefore, the parameter g and the penalty coefficient C of the gaussian kernel Function (RBF) of the SVM classifier are optimized by a grid optimization method. The mathematical form of the gaussian kernel function is shown as follows:

wherein x and y represent two samples, | | x-y | | Y circuitry²σ is the width parameter of the function as the Euclidean distance of samples x and y.

In this embodiment, the trained prediction model is used to predict the artery scan image of the intestinal tract lesion of the sample, so as to screen out the sample that is not susceptible to IFX. Therefore, the CD sample can accurately predict the IFX early curative effect.

In an implementation manner of this embodiment, the training process of the prediction model includes the following steps:

s1, obtaining a plurality of artery phase scanning images of the intestinal tract focus and a plurality of evaluation results of the curative effect of the inflixb monoclonal antibody.

Specifically, the images of arterial phase scans of multiple intestinal lesions and clinical data such as IFX injection per rule and review of laboratory indices, collecting clinical information such as sample gender, age, height, weight, disease course, montreal's typing, surgical history, smoking history, medication history, and laboratory indices such as hematocrit (Hct), C-reactive protein (CRP), Albumin (ALB), White Blood Cells (WBC), Red Blood Cells (RBC) can be collected.

Based on endoscopic results and clinical symptoms, CD samples were evaluated for IFX response half a year after initial IFX treatment (week 26) and classified as "response" and "no response" as gold standards for detection. If Simple Endoscopic CD Score (SES-CD) is reduced more than half way from baseline level and the sample achieves clinical remission, i.e. CD Activity Index (CDAI) <150, then response to IFX is defined; failure to meet the above criteria is considered as no response (primary no response).

And S2, based on the artery phase scanning image, generating a corresponding VOI image by using a region growing algorithm, extracting VOI image omics characteristics from the VOI image, taking the VOI image omics characteristics and the evaluation result as training samples, and training the prediction model by adopting a leave-one-out cross-validation method to obtain the corresponding prediction probability.

Specifically, the acquired arterial phase scanning image of the lesion is used for generating a corresponding VOI image by using a region growing algorithm, and a visual group image is extracted from the VOI image. For example, based on the CET artery scan image of each lesion and the corresponding VOI thereof, 1130 cine omics features are extracted in total, wherein the cine omics features include 14 shape features, 18 first-order features, 24 gray-level walk matrices (GLCM), 16 gray-level run matrices (GLRLM), 16 gray-level region matrices (GLSZM), 5 field gray-level matrices (NGTDM), 14 gray-level correlation matrices (GLDM), 744 wavelet features (wavelet), and 279 laplacian gaussian filter features (log).

With reference to fig. 6, in the present embodiment, feature screening is performed according to the skewness of each dimension feature, wherein the skewness (class imbalance) is a main determining factor of classification performance. Each dimension feature refers to each category feature, such as a shape feature, a gray area matrix, and the like.

In this embodiment, a Z-score normalization method is used to speed up gradient descent and optimal solution. The training sample feature values for each dimension were normalized by setting the mean to 0 and the standard deviation to 1. And finishing the training and testing of the prediction model by adopting a leave-one-out cross-validation strategy. And (3) a leave-one-out cross-validation strategy is that one sample in the input samples is selected in sequence for testing, the other samples are used as training samples to train the classifier, and the process is circulated until each sample in the input samples is used as a testing sample to be tested once, and corresponding prediction probability is obtained. Among them, the Z-score normalization method is also called standard deviation normalization, and is a method for normalizing data by giving mean and standard deviation (standard deviation) of raw data.

S3, setting the maximum iteration times of the prediction model, and stopping training the prediction model when the maximum iteration times is reached to obtain the trained prediction model.

Specifically, by training a prediction model, the classification probability output after training is recorded as a shadowmics score (Rad-score), and the significant difference P of the Rad-score of a response sample and a non-response sample is less than 0.001.

In one embodiment of this embodiment, the training sample further includes a laboratory index; the laboratory index is obtained by performing clinical characteristic analysis on clinical data of the plurality of samples.

For example, baseline data for 144 samples were collected, with 118 samples as the modeling set and 26 samples as the validation set. 88 of 118 samples in the modeling set responded to IFX treatment, 30 did not respond, and the response rate was 74.6%; while 17 out of 26 samples in the validation set responded to IFX treatment, 9 others did not respond, with a response rate of 65.4%. Of all CD samples included in the study, 73.6% were male, 81.2% were lesions located in the ileum, 47.2% were non-stenotic and non-penetrating, 43.8% were treated with immunosuppressants simultaneously, and 5.6% were operated. The baseline data of the modeling set and the validation set were subjected to one-way analysis (as shown in table 1), and the results showed that there was no statistical difference between the modeling set and the validation set in clinical baseline data including age, gender, CRP, montreal typing, drug therapy, etc. (P ═ 0.168-1.000), except for the statistical difference between the albumin level (P ═ 0.006) and the prior immunosuppressant therapy (P ═ 0.028). The baseline data for the response and non-response groups in the modeling and validation sets, respectively, were subjected to one-way analysis (see table 2). After 26 weeks of IFX treatment, single factor analysis results of baseline data of the modeling set show that only two laboratory indexes of CRP and ALB are related to the early treatment effect of IFX, and P values are 0.036 and 0.002 (as shown in figures 4 to 5).

TABLE 1 Baseline clinical and laboratory index characterization of modeling and validation set CD samples

a non-normally distributed continuous variables, expressed as median (quartile), row Mann-Whitney U test.

b, the classification variable is expressed by frequency (proportion) and the chi-square test is carried out.

c normally distributing continuous variables, expressing by mean plus or minus standard deviation, and performing independent sample t test.

Abbreviations: BMI body index, Hct hematocrit, CRP C-reactive protein, ESR erythrocyte sedimentation rate, RBC erythrocytes, WBC leukocytes, PLT platelets, ALT glutamic-pyruvic transaminase, AST glutamic-oxaloacetic transaminase, ALB albumin, SES-CD simplified crohn's disease endoscopic score, CDAI crohn's disease activity index.

TABLE 2 Single-factor analysis of clinical data between IFX early response combination non-responder groups in modeling and validation sets

The b classification variables are expressed by frequency (proportion), line chi-square test or Fisher test.

In this embodiment, only two test indexes of CRP and ALB have correlation with the IFX early stage efficacy, and multi-factor logistic regression analysis is used to screen effective prediction factors in the clinical indexes of CRP, ALB and imaging omics index Rad-score to construct a clinical feature + imaging omics Nomogram. The results show that ALB (P <0.05) and Rad-score (P <0.001) can be used for constructing a Nomogram model (as shown in FIG. 7), each prediction factor corresponds to a score of different proportions, and the probability corresponding to the total score is the probability of the sample appearing early non-response. The model has AUC up to 0.900, sensitivity up to 0.967 and specificity up to 0.727 in the modeling set. Among other things, the nogogrm is an easy-to-use clinical decision tool.

Based on the method for predicting the drug effect of the Riximab infliximab against the Crohn's disease, the invention also provides a computer-readable storage medium, wherein a drug effect prediction program of the Riximab against the Crohn's disease is stored in the computer-readable storage medium, and the steps in the method for predicting the drug effect of the Riximab against the Crohn's disease against the Riximab against the Crohn's disease are realized when the drug effect prediction program of the Riximab against the Crohn's disease against the Riximab is executed by a processor.

Based on the aforementioned method for predicting drug efficacy of the crohn disease infliximab, the present invention further provides a terminal device, as shown in fig. 9, which includes at least one processor (processor)30 and a memory (memory)31, and may further include a communication interface (communications interface)32 and a bus 33. The processor 30, the memory 31 and the communication interface 32 may communicate with each other via a bus 33. Communication interface 32 may communicate information. The processor 30 may call logic instructions in the memory 31 to perform the method in the above embodiments. In addition, the logic instructions in the memory 31 may be implemented in the form of software functional units and stored in a readable storage medium when the logic instructions are sold or used as independent products. The memory 31 is a readable storage medium and may be configured to store a software program, such as program instructions or modules corresponding to the methods in the embodiments of the present invention. The processor 30 executes the functional application and data processing by executing the software program, instructions or modules stored in the memory 31, that is, implements the method in the above-described embodiment. The memory 31 may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created according to the use of the terminal, and the like. Further, the memory 31 may include a high-speed random access memory, and may also include a nonvolatile memory. For example, a variety of media that can store program codes, such as a usb disk, a removable hard disk, a Read-only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk, may also be transient storage media. In addition, the specific processes loaded and executed by the instruction processors in the storage medium and the terminal are described in detail in the method, and are not stated herein.

In summary, the present invention provides a method for predicting drug efficacy of infliximab due to crohn's disease, a computer-readable storage medium, and a terminal device. A prediction model is established by using a support vector machine classifier, a corresponding VOI image is generated by using an arterial phase scanning image of a sample and a region growing algorithm, VOI image omics characteristics are extracted from the VOI image, the VOI image omics characteristics, an IFX treatment evaluation result of the sample and an experimental index of the sample are used as training samples, a clinical characteristic combination influence omics Nonogram prediction model is obtained by training, and the IFX curative effect of the sample is predicted by using the prediction model to obtain a prediction result. Therefore, samples which are not easy to IFX are screened out, and accurate prediction of early IFX curative effect by the CD samples is realized.

It is to be understood that the invention is not limited to the examples described above, but that modifications and variations may be effected thereto by those of ordinary skill in the art in light of the foregoing description, and that all such modifications and variations are intended to be within the scope of the invention as defined by the appended claims.

Claims

1. A method for predicting drug efficacy of England Rib-ximab for Crohn's disease is characterized by comprising the following steps:

acquiring an artery phase scanning image of an intestinal tract lesion to be predicted, and extracting an image omics feature from the artery phase scanning image;

inputting the image omics characteristics into a pre-trained prediction model for calculation to obtain a drug effect prediction result, wherein the prediction model is obtained by training image omics characteristics corresponding to arterial phase scanning images of a plurality of intestinal lesions and an evaluation result of the therapeutic effect of the inflixb monoclonal antibody;

generating a corresponding VOI image by using a region growing algorithm based on an artery phase scanning image of the intestinal tract focus;

extracting VOI (VoI cinematology) features from the VOI images; the VOI imagery omics characteristics comprise quantization intensity, shape, texture and wavelet characteristics;

the training method of the prediction model comprises the following steps:

obtaining a plurality of artery phase scanning images of the intestinal tract focus and an evaluation result of the curative effect of the inflixb monoclonal antibody;

based on the artery phase scanning image, generating a corresponding VOI image by using a region growing algorithm, extracting VOI image omics characteristics from the VOI image, taking the VOI image omics characteristics and the evaluation result as training samples, and training the prediction model by adopting a leave-one-out cross-validation method to obtain corresponding prediction probability;

setting the maximum iteration times of a prediction model, and stopping training the prediction model when the maximum iteration times is reached to obtain the trained prediction model;

the training sample further comprises laboratory metrics; the laboratory index is obtained by performing clinical characteristic analysis on the clinical data of the plurality of samples; the laboratory metrics include: c-reactive protein index and albumin index.

2. The method for predicting the pharmacodynamic effect of infliximab due to crohn's disease according to claim 1, wherein the training of the prediction model by a leave-one-out cross-validation method using the VOI imaging omic features and the evaluation results as training samples comprises:

the VOI omics features and the assessment results were normalized using a Z-score normalization method.

3. The method of claim 1, wherein the prediction model is built using a support vector machine classifier, and the gaussian kernel function g and the penalty coefficient C of the support vector machine classifier are optimized by a grid optimization method; the mathematical form of the gaussian kernel function is shown as follows:

4. A computer-readable storage medium storing a drug efficacy prediction program of infliximab crohn disease, which when executed by a processor, performs the steps of the drug efficacy prediction method of infliximab crohn disease according to any one of claims 1 to 3.

5. A terminal device, comprising a processor, a memory, and a drug efficacy prediction program of infliximab crohn disease stored in the memory and executable on the processor, wherein the processor implements the steps of the drug efficacy prediction method of infliximab crohn disease as set forth in any one of claims 1 to 3 when executing the drug efficacy prediction program.