CN111951962A - Bevacizumab curative effect prediction model for treating radioactive brain injury and construction method thereof - Google Patents
Bevacizumab curative effect prediction model for treating radioactive brain injury and construction method thereof Download PDFInfo
- Publication number
- CN111951962A CN111951962A CN202010734699.XA CN202010734699A CN111951962A CN 111951962 A CN111951962 A CN 111951962A CN 202010734699 A CN202010734699 A CN 202010734699A CN 111951962 A CN111951962 A CN 111951962A
- Authority
- CN
- China
- Prior art keywords
- wavelet
- treatment
- brain injury
- bevacizumab
- glszm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10088—Magnetic resonance imaging [MRI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30016—Brain
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30096—Tumor; Lesion
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Data Mining & Analysis (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Epidemiology (AREA)
- Pathology (AREA)
- Databases & Information Systems (AREA)
- Primary Health Care (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention discloses a bevacizumab therapeutic radiation brain injury curative effect prediction model and a construction method thereof, and provides an MRI-based imaging group classifier and clinical characteristics, and a method for constructing the bevacizumab therapeutic radiation brain injury curative effect prediction model. The invention also provides a prediction model constructed according to the method, the model has good efficacy, can effectively predict the effect of bevacizumab on treating radioactive brain injury of different patients, assists doctors to provide personalized accurate treatment, adjusts a treatment scheme, avoids under-treatment or over-treatment, improves the prognosis of patients and improves the quality of life, and has high clinical value and significance as an effective and easy-to-use guiding tool for non-invasive clinical decision.
Description
Technical Field
The invention belongs to the field of biological medicine, and relates to a prediction model of curative effect of bevacizumab on radioactive brain injury and a construction method thereof.
Background
Nasopharyngeal carcinoma is one of the common head and neck tumors, has obvious regional aggregative property, and is called Guangdong tumor in China south China, especially Guangdong province, with high incidence. Worldwide, about 8.7 ten thousand new cases occur each year, about 5.1 ten thousand death cases occur each year, and the health of human beings is seriously threatened. A large body of high-level evidence-based medical evidence suggests that radiotherapy or chemotherapy is the primary treatment for early or locally advanced nasopharyngeal carcinoma. Although various radiotherapy methods are aimed at reducing the radiation damage to the normal brain tissue adjacent to the irradiation target, the radiation damage is still inevitable to cause different degrees of brain parenchyma damage, and the mild patients have no clinical symptoms, and the severe patients can have radioactive brain damage. Radioactive brain damage occurs 3 months to 10 years after radiation therapy in about 3-24% of nasopharyngeal carcinoma patients, most of which occur within 2 years after radiation therapy.
Once a radioactive brain injury occurs, the brain injury will be progressively aggravated if no effective treatment is taken. The clinical manifestations of radioactive brain injury can be divided into three types according to the focus, namely hemispheric brain, brainstem and cranial nerve type, and mixed type. Cerebral hemispheres are common clinically, and patients mainly show that cognitive function is reduced, functions of language, memory, learning, execution and the like are damaged, intracranial hypertension symptoms such as headache, nausea and the like are caused, or brain injury is gradually aggravated to cerebral hernia, coma and even death. Patients with dry brain and cranial nerve mainly have posterior cranial nerve injury, and dysphagia and choking cough due to drinking water are mainly clinically manifested. Dysphagia caused by radiotherapy refers to dysphagia caused by chronic reactions after radiotherapy, including posterior cranial neuropathy, structural rigidity of swallowing organs, loss of function, tissue fibrosis and the like. Dysphagia can ultimately be life threatening if aspiration pneumonia, sepsis or severe malnutrition occurs. Approximately one third of dysphagia patients develop aspiration pneumonia with a mortality rate of 20% -65%.
The current treatment for radioactive brain injury is mainly based on glucocorticoid. The application of glucocorticoid can inhibit the generation and development of inflammatory reaction in brain of patients, reduce the degree of neurological impairment, reduce the focus volume, and is helpful for improving the long-term prognosis of patients. Glucocorticoids commonly used in the treatment of radioactive brain injury include: dexamethasone, methylprednisolone, prednisone, etc., of which methylprednisolone is most widely used. However, hormone therapy is accompanied by side effects such as cushing-like syndrome, peptic ulcer, osteoporosis, immunosuppression, and mental disorder, and only 35% of patients are sensitive to hormone therapy. In addition, some auxiliary treatments, such as antiplatelet, anticoagulation, hyperbaric oxygen, circulation improvement and the like, have been researched and shown to have an improvement effect on brain injury. Anticoagulant therapy with heparin and warfarin can improve the clinical symptoms of brain injury, possibly by reversing capillary endothelial injury. Hyperbaric oxygen therapy aims at increasing the oxygen partial pressure of tissue cells to excite a vascular repair mechanism, reducing vascular permeability and relieving normal tissue necrosis around tumors caused by radiotherapy, and researches show that hyperbaric oxygen can relieve cerebral edema. Craniotomy surgery and excision of radiation-induced lesions are currently only aimed at refractory brain injuries and those with conservative treatment failures.
The radioactive brain injury is closely related to blood brain barrier damage, vascular permeability increase, inflammatory reaction, VEGF increase and the like, and the increased VEGF plays an important role in the development of cerebral edema and necrosis. Bevacizumab, a humanized anti-VEGF monoclonal antibody, has been proved to be effective in improving radiation brain injury in many studies, reducing edema focus volume by 52-64%, and the effective rate is about 66%. Moreover, a randomized controlled trial of bevacizumab-treated radiation brain injury previously developed demonstrated that bevacizumab was more effective in reducing edema lesion volume and improving clinical symptoms relative to hormones. Therefore, bevacizumab is increasingly applied to the treatment of radioactive brain injury, and has wide potential to become a primary treatment mode. However, some patients are not sensitive to bevacizumab treatment, so that treatment is ineffective or even the disease condition is aggravated, some patients have serious side effects, and bevacizumab is expensive and only injection but no oral preparation is available, so that the popularization of bevacizumab in clinical treatment and application is limited. Therefore, it is particularly necessary to predict and distinguish bevacizumab treatment-effective and non-effective patients before treatment, provide personalized and accurate treatment, and avoid unnecessary side effect risks of treatment-ineffective patients.
Disclosure of Invention
In view of the above problems, the present invention aims to provide a prediction model for predicting the efficacy of bevacizumab in treating radioactive brain injury.
In order to achieve the purpose, the invention adopts the technical scheme that: a method for constructing a prediction model of curative effect of bevacizumab on radioactive brain injury comprises the following steps:
(1) collecting clinical data and MRI image data of a patient with radioactive brain injury treated by bevacizumab;
(2) analyzing the MRI image data collected in the step (1), segmenting the focus, extracting image features, screening the image features related to treatment response through LASSO logical analysis, and obtaining a corresponding image omics score calculation formula;
(3) screening the imaging omics scores and clinical characteristics in the clinical data collected in the step (1) through multi-factor logistic regression analysis and maximum likelihood ratio test by adopting a regression method to obtain characteristics related to treatment response;
(4) and (4) constructing and obtaining a corresponding composite prediction model based on the image omics score and the clinical characteristics based on the characteristics obtained in the step (3).
Imaging omics is a popular methodology in recent years, and research contents include acquisition of image images, segmentation and reconstruction of images, high-throughput extraction of a large number of image features from medical radiographic images, and determination of the association between image features and potential pathophysiological features. The imaging omics can be combined with clinical data, laboratory examination indexes, proteomics, genomics and the like, is applied to a clinical decision system to assist in diagnosing diseases, evaluating prognosis, predicting treatment response and the like, and is receiving more and more attention in the field of tumor research, wherein a plurality of researches show that an MRI image before treatment can be used for predicting the treatment response of certain diseases such as colon cancer, cervical cancer, esophageal cancer, glioma and the like, so that the method is favorable for clinically predicting the treatment effect, adjusting the treatment scheme, avoiding insufficient treatment or excessive treatment and achieving the aim of accurate treatment. The invention is used for constructing a model for evaluating the curative effect of bevacizumab on the radioactive brain injury patient by combining the imaging omics and clinical characteristics, can provide more effective evaluation opinions for whether the radioactive brain injury patient selects bevacizumab, and reduces or even avoids the risk of unnecessary side effects on the patient who is not treated effectively.
As a preferred embodiment of the present invention, the method further comprises the step (5): and (4) evaluating the composite prediction model obtained in the step (4).
More preferably, the assessment method comprises the steps of assessing the calibration degree of the imaging group-clinical compound prediction model by drawing a calibration curve, assessing the goodness of fit of the prediction model by using a Hosmer-Lemeshow test, quantitatively assessing the discrimination degree of the prediction model by using AUC (AUC), and finally assessing the clinical applicability of the prediction model by using Decision Curve Analysis (DCA).
The prediction efficiency of the constructed model can be known through evaluation, and the actual application of the model is further guided more effectively according to the evaluation result, so that doctors and patients are better helped to make clinical decisions.
As a preferred embodiment of the present invention, the inclusion criteria for the patient with radioactive brain injury after radiotherapy of nasopharyngeal carcinoma treated with bevacizumab in step (1) include: because pathological diagnosis nasopharyngeal carcinoma receives radiotherapy, imaging evidence supports radioactive brain injury diagnosis without tumor recurrence or metastasis, bevacizumab is received for 4 courses of treatment, and MRI evaluates brain injury focuses before and after treatment.
The bevacizumab treatment is: the patient received bevacizumab for 4 courses of treatment, 5mg/kg i.v. instillation, every two weeks.
Further, it is desirable to eliminate patients with artifacts on MRI images and incomplete clinical data to reduce unnecessary errors.
As a preferred embodiment of the present invention, the clinical data of step (1) comprises the following clinical characteristics: age, gender, height and weight, presence or absence of cranial nerve damage, hypersensitive C-reactive protein levels, time interval for radiation therapy and diagnosis of radioactive brain damage, time interval for diagnosis of radioactive brain damage and bevacizumab treatment, maximum dose of irradiation to temporal lobe, total dose of irradiation to neck, method of irradiation and LENT/SOMA scale scores.
In a preferred embodiment of the present invention, the image data in step (1) includes: the patient had two sequence findings, T2-Weighted FLAIR and T1 post-gadolinium, from a cranial MRI examination conducted 3 days prior to treatment with bevacizumab at treatment 1 and 2 weeks after the end of treatment 4.
In a preferred embodiment of the present invention, the MRI image feature extraction method in step (2) includes: dividing two-dimensional interested areas, and stacking continuous two-dimensional interested areas to form a three-dimensional interested volume, namely a brain injury focus; and extracting image characteristics of the brain injury focus.
On the FLAIR image, 1301 image features can be extracted from each lesion. The extraction of image features is carried out through a 3D Slicer software Pyradiomics platform.
More preferably, the process of extracting the two-dimensional region-of-interest image is to automatically segment the approximate outline of the two-dimensional region-of-interest by 3D Slicer software, and then fine-tune the edge of the region-of-interest by an imaging physician.
The adjustment process of the imaging physician can be performed by a plurality of qualified imaging physicians, and the repeatability of the extracted result is evaluated by randomly extracting patient data, wherein the higher the repeatability is, the better the repeatability is.
In a preferred embodiment of the present invention, when the image feature or clinical feature screening related to the therapeutic response is performed in steps (2) and (3), the effective therapeutic response is defined as: (1) the neurological symptoms or signs were stable, the FLAIR sequence showed a reduction in brain injury volume of at least 25%; or (2) a reduction in the level of improvement in neurological symptoms or signs by at least a level 1 as an improvement, as assessed by the LENT/SOMA rating scale.
The invention also claims a composite predictive model constructed according to the method.
As a preferred embodiment of the present invention, the composite prediction model includes a bevacizumab treatment effectiveness calculation formula, wherein the formula is: treatment response score 3.436 x imaging score-0.024 x time interval between radiotherapy and radioactive brain injury diagnosis (month) -0.043 x time interval between radioactive brain injury diagnosis and bevacizumab treatment (month) + 0.451; wherein:
the imagination score is 0.483308872
+0.996788068×LoG(σ＝1)_Firstorder_Maximum
-0.225130674×LoG(σ＝1)_GLSZM_LAHGLE
-0.227908898×LoG(σ＝2)_GLSZM_SALGLE
-1.174540209×LoG(σ＝3)_GLSZM_LAHGLE
+1.207131780×LoG(σ＝4)_GLSZM_SZNN
+2.523507898×Wavelet(HHH)_Firstorder_Skewness
+1.144733635×Wavelet(HHH)_Firstorder_Median
-0.401726675×Wavelet(HHH)_Firstorder_Kurtosis
-0.049392860×Wavelet(HHH)_GLRLM_LRE
+0.757594985×Wavelet(HHH)_GLSZM_SZNN
+0.416040381×Wavelet(HLL)_Firstorder_Median
+0.019379737×Wavelet(HLL)_Firstorder_Mean
-0.840176612×Wavelet(LHH)_GLDM_LDLGLE
-0.168430560×Wavelet(LHH)_GLSZM_SALGLE
-1.134782147×Wavelet(LLH)_GLSZM_LAHGLE
-0.627111233×Wavelet(HHL)_GLCM_IDMN
-0.573236112×Wavelet(HLH)_GLCM_Correlation
-0.946488391×GLCM_InverseVariance。
The composite prediction model provided by the invention is obtained by the model construction method based on the data of 118 patients (152 focuses in total) with nasopharyngeal carcinoma radiotherapy-diagnosed radioactive brain injuries which accord with the standards, the image characteristics are extracted by 2 depth image doctors, the correlation coefficient between operators is 0.912 (+/-SD (+/-0.142), and the composite prediction model has good stability; obtaining 18 image characteristics related to treatment reaction in the image characteristic screening, wherein the image characteristics are first order _ Maximum, GLSZM _ LAHGLE, GLSZM _ SALGLE, GLSZM _ LAHGLE, GLSZM _ SZNN, Wavelet (HHH) _ first order _ Skewness, Wavelet (HHH) _ first order _ Median, Wavelet (HHH) _ first order _ Kurtosis, Wavelet (HHH) _ GLRLM _ LRE, Wavelet (HHH) _ GLSZM _ SZNN, Wavelet (HLL) _ first order _ Mean, Wavelet (HLL) _ first order _ LGan, Wavelet (LHH) _ GLDM _ LDLE, Wavelet (LHH) _ LHS _ LHL), GLS _ SALG, GLS _ LMM _ GLS _ LML _ LMH _ GLS _ LMH _ GLS _ L _ GLS _ L _ GLS _; the analysis yielded 2 clinical features associated with treatment response, namely the interval between radiotherapy and diagnosis of radioactive brain Injury (IRB) and the interval between diagnosis of radioactive brain Injury and Bevacizumab (IBT).
Through evaluation, in a training set, a calibration curve shows that the effective possibility of the model is consistent with the actual possibility well, the AUC is 0.916 (95% CI, 0.857-0.976), the prediction model has good discrimination, and the result of the Hosmer-Lemeshow test shows that the prediction model has no significant difference (P is 0.484), and the prediction model has no deviation from perfect matching. Similarly, the prediction model in the validation group also showed good calibration and also had good discrimination (AUC 0.912, 95% CI 0.808-1.000). The result of the Hosmer-Lemeshow test is also no significant difference (P is 0.184). DCA shows that the composite prediction model has good clinical applicability, in a training group and a verification group, the threshold probability range of treatment effectiveness is 0% to 100% and 0% to 89.4%, and the model is applied to guide treatment and benefit more than that all patients are treated or all patients are not treated. The model construction of the invention is only brought into patients with nasopharyngeal carcinoma, and can eliminate the tumor recurrence and metastasis and other mixed factors except brain injury, so that the prediction model is more accurate and has the potential advantages of being applied to other tumors.
The invention also claims the application of image characteristics and clinical characteristics in the combined construction of a bevacizumab therapeutic radiation brain injury curative effect prediction model, wherein the image characteristics comprise first support _ Maximum, GLSZM _ LAHGLE, GLSZM _ salgl, GLSZM _ LAHGLE, GLSZM _ SZNN, Wavelet (HHH) _ first support _ skwness, Wavelet (HHH) _ first support _ media, Wavelet (HHH) _ GLRLM _ LRE, Wavelet (HHH) _ szm _ szrn, Wavelet (HLL) _ med support _ first, lgvel (HLL) _ first support _ Mean, Wavelet (HHH) _ lhlhlhlhlhlhs _ szm, gldlz _ slzn, Wavelet (HLL) _ medsfield _ first support _ gll, glldm _ glld _ g _ horizontal, glld _ horizontal _ glld _ g _ s _ bright; the clinical features include the time interval between diagnosis of radioactive brain injury and bevacizumab treatment.
The invention provides a method for constructing a prediction model of curative effect of bevacizumab on radioactive brain injury based on an MRI (magnetic resonance imaging) imaging omics classifier and clinical characteristics, and provides an imaging omics-clinical composite prediction model constructed for predicting the curative effect of radioactive brain injury after bevacizumab on nasopharyngeal carcinoma radiotherapy. Overall, the invention has several advantages as follows: first, the use of high-throughput imaging features provides more detailed information about the radioactive brain injury foci, thereby enabling more sensitive and comprehensive prediction of treatment response. Second, segmenting a three-dimensional volume of interest from the image plane more accurately demonstrates the heterogeneity of the entire lesion than does a two-dimensional region of interest. Thirdly, the research object is the focus of brain injury and is not the individual patient, so the model has higher prediction accuracy, and can particularly provide more accurate treatment direction for patients with bilateral brain injury lesion. The model provided by the invention has good efficacy, can effectively predict the effect of bevacizumab on treating radioactive brain injury of different patients, assists doctors in providing personalized accurate treatment, adjusts the treatment scheme, avoids under-treatment or over-treatment, improves the prognosis of patients and improves the quality of life, and has high clinical value and significance as an effective and easy-to-use guiding tool for non-invasive clinical decision.
Drawings
Fig. 1 is a result of evaluating the prediction performance of the combined imaging omics-clinical prediction model of bevacizumab efficacy; a is the composite prediction model calibration degree evaluated in the training group and the verification group respectively through calibration curves, and B is the composite prediction model discrimination degree evaluated in the training group and the verification group respectively through ROC analysis.
FIG. 2 is a graph of the results of an evaluation of the clinical utility of the imaging omics-clinical composite predictive model; a is the evaluation result in the training set, and B is the evaluation result in the validation set.
Detailed Description
To better illustrate the objects, aspects and advantages of the present invention, the present invention will be further described with reference to the accompanying drawings and specific embodiments.
All statistical analyses of the invention were performed in software R (version 3.4.2). Using χ2The categorical variables are compared by testing or Fisher's exact probability method. For normally distributed continuous variables, a comparison between groups was performed using the t-test; for non-normally distributed continuous variables, comparisons between groups were performed using the Mann-Whitney U test. Logistic regression models are used to perform single and multi-factor analyses. The "glmnet" package is used to perform LASSO logistic regression analysis and multifactorial logistic regression analysis. The "pROC" package was used to plot ROC curves. The "rms" packet was used to plot the calibration curve, and the "vcdExtra" packet was used to perform the Hosmer-Lemeshow test. The "DCA" packet is used to plot a decision analysis curve. Two sides P<0.05 was considered to be significantly different.
1. Patient screening and data collection
A total of 118 patients diagnosed with post-radiotherapy radioactive brain damage to nasopharyngeal carcinoma (152 lesions in total) were enrolled during the period from 7 months 2012 to 3 months 2019 and received bevacizumab (Avastin, Genentech, South San Francisco, CA, USA). The inclusion criteria were: (1) radiation therapy is received for nasopharyngeal carcinoma due to pathological diagnosis; (2) imaging evidence supports diagnosis of radioactive brain injury without tumor recurrence or metastasis; (3) receiving bevacizumab for 4 courses of treatment; (4) the lesion of brain injury is evaluated by MRI before and after treatment. Exclusion criteria were: (1) MRI image examination before and after treatment is not carried out in the hospital; (2) artifacts are present on MRI images; (3) clinical data are not fully available.
All patients received bevacizumab treatment every two weeks, 5mg/kg i.v. drip, for a total of 4 courses. The brain injury lesions of all patients are divided into two groups according to grouping time, wherein the training group is from 7 months to 12 months in 2012 to 2017, the verification group is from 1 month to 3 months in 2018, and 101 lesions and 51 lesions are respectively arranged in the final training group and the verification group.
Prior to bevacizumab treatment, the following clinical data were collected by outpatient and hospitalization medical records: patient age, sex, height and weight (used to calculate body mass index, BMI), whether there is cranial nerve damage, high sensitivity C-reactive protein (hs-CRP) level, time interval between radiation therapy and diagnosis of radioactive brain damage (interval between radiation therapy and diagnosis of brain necrosis, IRB), time interval between diagnosis of radioactive brain damage and bevacizumab therapy (interval between radiation diagnosis and diagnosis of brain necrosis and treatment with therapy, IBT), maximum dose to temporal lobe, total dose to neck, and method of irradiation (conventional or conformal-modulated radiation therapy, IMRT) and LENT/SOMA scale.
Cranial MRI examinations were performed 3 days before bevacizumab treatment on day 1 and 2 weeks after the end of treatment on day 4, including two sequence examinations T2-Weighted FLAIR and T1 post-gadolinium.
2. Obtaining MRI image, segmenting focus, extracting image features
All patients use an 8-channel phased array head coil and a 1.5T whole-body nuclear magnetic resonance imaging system to carry out MR imaging before and after bevacizumab treatment, and the machine model is Gyroscan Intera; philips, Aachen, Germany. The parameters for the skull coronal FLAIR sequence acquisition are as follows: layer thickness, 5 mm; the difference, 1.5 mm; echo time, 180 ms; repetition time, 11000 ms; inversion time, 2800 ms; echo train length, 48. All patients entering the group are obtained through a hospital Imaging system according to DICOM (digital Imaging and Communications in medicine) format skull MRI images for Imaging and feature extraction.
Segmentation of brain lesion lesions was done by semi-automated segmentation using a GrowCut plug-in of open source 3D Slicer software (version 4.9.0). The approximate outline of a two-dimensional region of interest (ROIs) is automatically segmented by software by using a threshold segmentation method, and then the edge of the region of interest is finely adjusted by an imaging physician. Successive ROIs are stacked to form a volumetric three-dimensional volume of interest (VOIs), i.e., a brain lesion. All lesion segmentation was done by a imaging physician a with 10 years of clinical experience and reviewed and modified by a senior imaging physician B with 22 years of clinical experience. Then, image features of the segmented brain lesion are extracted. On the FLAIR image, 1301 image features can be extracted from each lesion.
In addition, image segmentation and feature extraction were performed by physician a and physician B, respectively, by randomly selecting 20 patients (25 lesions in total), and inter-operator correlation coefficients were calculated for reproducibility evaluation. Through tests, the correlation coefficient between operators of the image features is 0.912 (+ -SD, + -0.142), and the image segmentation and feature extraction have good stability.
3. Construction of image omics classifier and discrimination evaluation
The LASSO (least squares similarity and selection operator) algorithm is an algorithm suitable for processing high-throughput data based on cross validation and penalty idea. In the training group, an LASSO logistic regression algorithm is adopted to screen image features related to treatment response from the extracted image features, and an image omics score is calculated and used for reflecting the effective treatment response probability of each brain injury focus.
The effective therapeutic response is defined as: (1) the neurological symptoms or signs were stable, the FLAIR sequence showed a reduction in brain injury volume of at least 25%; or (2) improvement in neurological symptoms or signs (improvement is at least a 1-degree reduction in the LENT/SOMA assessment scale).
In the training set, the discrimination of the image omics classifier was evaluated by the area under the curve (AUC) of the receiver operator characteristic curve (ROC) and verified in the verification set. In the training set, the AUC was 0.873 (95% CI, 0.804-0.943), indicating good discrimination between the predictive models. Good discrimination of the model was confirmed in the validation group to have an AUC of 0.869 (95% CI 0.755-0.982).
Within the training set, penalty parameter values in the LASSO logistic regression analysis were determined by ten-fold cross-validation. Through multi-factor logistic regression analysis and maximum likelihood ratio test of a regression method, 18 image features relevant to treatment response are screened out from 1301 image features, wherein the image features comprise: first datum _ Maximum, GLSZM _ LAHGLE, GLSZM _ SALGLE, GLSZM _ LAHGLE, GLSZM _ SZNN, Wavelet (HHH) _ first datum _ Skewness, Wavelet (HHH) _ first datum _ media, Wavelet (HHH) _ first datum _ Kurtosis, Wavelet (HHH) _ GLRLM _ LRE, Wavelet (HHH) _ GLSZM _ SN, Wavelet (HLL) _ first datum _ media, Wavelet (HLL) _ first datum _ Mean, Wavelet (LHH) _ GLDM _ LGLE, Wavelet (LHH) _ GLSZM _ SALE, Wavelet (WallH) _ GLSZM _ SALL, Wavelet (WallL) _ GLS _ LAHG, Wavelet _ GLCM _ HH _ GLS _ TM _ GLS _ S _;
remarking: each feature is named according to "filter _ feature class _ feature name". For filter "LoG", the values in parentheses represent the filter width for the gaussian kernel; for the filter "Wavelet", the values in parentheses indicate the type of filtering in the x, y and z directions (H: high pass; L: low pass), respectively.
Abbreviation LoG Laplacian of Gaussian; GLSZM is Gray Level Size Zone Matrix; GLRLM Gray Level Run Length Matrix; GLDM is Gray Level dependency Matrix; GLCM Gray Level Cooccurrence Matrix; LAHGLE, Large Area High Gray Level Emphasis; SALGLE, Small Area Low Gray Level Emphasis; SZNN, Size-Zone Non-unity Normalized; LRE, Long Run Emphasis; LDLGLE, Large Dependence Low Gray Level Emphasis; IDMN is Inverse Difference norm.
Further, a method for weighting and then linearly adding the screened imaging characteristics through coefficients obtained in the LASSO logical regression is adopted to obtain an imaging group score calculation formula, wherein the calculation method comprises the following steps:
the imagination score is 0.483308872
+0.996788068×LoG(σ＝1)_Firstorder_Maximum
-0.225130674×LoG(σ＝1)_GLSZM_LAHGLE
-0.227908898×LoG(σ＝2)_GLSZM_SALGLE
-1.174540209×LoG(σ＝3)_GLSZM_LAHGLE
+1.207131780×LoG(σ＝4)_GLSZM_SZNN
+2.523507898×Wavelet(HHH)_Firstorder_Skewness
+1.144733635×Wavelet(HHH)_Firstorder_Median
-0.401726675×Wavelet(HHH)_Firstorder_Kurtosis
-0.049392860×Wavelet(HHH)_GLRLM_LRE
+0.757594985×Wavelet(HHH)_GLSZM_SZNN
+0.416040381×Wavelet(HLL)_Firstorder_Median
+0.019379737×Wavelet(HLL)_Firstorder_Mean
-0.840176612×Wavelet(LHH)_GLDM_LDLGLE
-0.168430560×Wavelet(LHH)_GLSZM_SALGLE
-1.134782147×Wavelet(LLH)_GLSZM_LAHGLE
-0.627111233×Wavelet(HHL)_GLCM_IDMN
-0.573236112×Wavelet(HLH)_GLCM_Correlation
-0.946488391×GLCM_InverseVariance。
4. Construction of imaging omics-clinical composite prediction model
In the training group, the imaging omic score, the time interval between radiotherapy and diagnosis of radioactive brain Injury (IRB), the time interval between diagnosis of radioactive brain injury and bevacizumab (interval between diagnosis of radioactive brain injury and treatment of bevacizumab, IBT) were selected from the imaging omic score and a plurality of clinical characteristic variables as independent predictors of treatment response by multi-factor logistic regression analysis and maximum likelihood ratio test using the regression method.
The effective therapeutic response is defined as: (1) the neurological symptoms or signs were stable, the FLAIR sequence showed a reduction in brain injury volume of at least 25%; or (2) improvement in neurological symptoms or signs (improvement is at least a 1-degree reduction in the LENT/SOMA assessment scale).
And evaluating the relationship between the imagemics score and the candidate clinical variables and the treatment response based on multi-factor logistic regression analysis, and establishing a composite prediction model comprising the imagemics score and the clinical variables. The model comprises the following calculation formula: therapeutic response score x ═ 3.436 × imaging score-0.024 × IRB (month) -0.043 × IBT (month) + 0.451.
Further, the probability that the corresponding bevacizumab treatment is effective y ═ 1/(1+ e) can be obtained from the treatment response fraction x-x)。
5. Prediction performance evaluation of imaging omics-clinical composite prediction model
In the training group, the calibration degree of the imaging group-clinical composite prediction model is evaluated by drawing a calibration curve, the goodness of fit of the prediction model is evaluated by using a Hosmer-Lemeshow test, and the discrimination degree of the prediction model is quantitatively evaluated by using AUC. And (3) verifying the prediction efficiency of the imaging omics-clinical composite prediction model in a verification group, evaluating the calibration degree by drawing a calibration curve, carrying out Hosmer-Lemeshow test, and calculating the AUC evaluation discrimination. The results are shown in fig. 1, where a is the evaluation of the degree of calibration of the composite predictive model in the training and validation sets by the evaluation of a calibration curve reflecting the degree of calibration of the model by comparing the agreement between the actual predicted performance and the ideal predicted performance. The 45-degree dotted line represents the perfect prediction efficiency of an ideal model, the solid line represents the prediction efficiency of the model constructed in the research in the training set and the verification set respectively, and the solid line closer to 45-degree represents higher prediction accuracy. And B, evaluating the discrimination of the composite prediction model in a training group and a verification group respectively through ROC analysis.
And finally, evaluating the clinical applicability of the prediction model by adopting Decision Curve Analysis (DCA) in a training group and a verification group respectively. The results are shown in fig. 2, and the clinical utility of the imaging omics-clinical composite prediction model was evaluated by applying DCA in the training group (a) and the validation group (B). The Y-axis represents net gain. "all" represents the net benefit of adopting the corresponding treatment strategy assuming that all patients are therapeutically effective. "none" means the net benefit of adopting the corresponding treatment strategy assuming all patients are ineffectively treated. The "model" represents the net benefit under the guidance of the composite predictive model. The X-axis represents the threshold probability. The threshold probability is the probability that the treatment is effective when the expected benefit of the treatment is equal to the expected benefit of the untreated. For example, if the likelihood that a patient is therapeutically effective exceeds a threshold probability, a corresponding treatment strategy should be employed that is expected to be therapeutically effective.
The construction of the imaging omics-clinical composite prediction model and the prediction efficiency evaluation result show that: in the training set, the calibration curve shows that the prediction model has good consistency of the predicted treatment effective possibility and the actual possibility, and the AUC is 0.916 (95% CI, 0.857-0.976), which shows that the prediction model has good discrimination. The result of the Hosmer-Lemeshow test is no significant difference (P is 0.484), which indicates that the prediction model has no deviation from perfect matching. Similarly, the prediction model in the validation group also showed good calibration and also had good discrimination (AUC 0.912, 95% CI 0.808-1.000). The result of the Hosmer-Lemeshow test is also no significant difference (P is 0.184).
The results of the clinical applicability evaluation of the imaging omics-clinical composite prediction model show that: DCA shows that the composite prediction model has good clinical applicability, and when the threshold probability ranges of effective treatment in a training group and a verification group are 0% to 100% and 0% to 89.4%, the patient can benefit by applying the composite prediction model to assist clinical decision compared with the method that all patients are regarded as good in treatment effect or all patients are regarded as poor in treatment effect and relevant treatment measures are taken.
Under the guidance of a composite prediction model, if the patient is expected to have good treatment response to bevacizumab, bevacizumab treatment is recommended, especially for the patient with bilateral brain injury focus and the expected treatment effect is good. Conversely, glucocorticoids may be an advantageous alternative if patients with poor treatment of bilateral or unilateral lesions are expected. For patients with inconsistent treatment results of bilateral lesions, the clinical manifestations, lesion location and volume of the patients should be considered comprehensively to balance the risk and disadvantage of each regimen, and finally to determine the application of bevacizumab or other treatment methods.
Finally, it should be noted that the above embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the protection scope of the present invention, although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions can be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.
Claims (10)
1. A method for constructing a prediction model of curative effect of bevacizumab on radioactive brain injury is characterized by comprising the following steps:
(1) collecting clinical data and MRI image data of a patient with radioactive brain injury treated by bevacizumab;
(2) analyzing the MRI image data collected in the step (1), segmenting the focus, extracting image features, screening the image features related to treatment response through LASSO logical analysis, and obtaining a corresponding image omics score calculation formula;
(3) screening the imaging omics scores and clinical characteristics in the clinical data collected in the step (1) through multi-factor logistic regression analysis and maximum likelihood ratio test by adopting a regression method to obtain characteristics related to treatment response;
(4) and (4) constructing and obtaining a corresponding composite prediction model based on the image omics score and the clinical characteristics based on the characteristics obtained in the step (3).
2. The method of claim 1, further comprising step (5): and (4) evaluating the composite prediction model obtained in the step (4).
3. The method of claim 1, wherein the inclusion criteria for the patient with radioactive brain injury after radiation therapy for nasopharyngeal carcinoma treated with bevacizumab in step (1) comprises: because pathological diagnosis nasopharyngeal carcinoma receives radiotherapy, imaging evidence supports radioactive brain injury diagnosis without tumor recurrence or metastasis, bevacizumab is received for 4 courses of treatment, and MRI evaluates brain injury focuses before and after treatment.
4. The method of claim 1, wherein the clinical data of step (1) comprises the following clinical characteristics: age, gender, height and weight, presence or absence of cranial nerve damage, hypersensitive C-reactive protein levels, time interval for radiation therapy and diagnosis of radioactive brain damage, time interval for diagnosis of radioactive brain damage and bevacizumab treatment, maximum dose of irradiation to temporal lobe, total dose of irradiation to neck, method of irradiation and LENT/SOMA scale scores.
5. The method of claim 1, wherein the image data of step (1) comprises: the patient had two sequence findings, T2-Weighted FLAIR and T1 post-gadolinium, from a cranial MRI examination conducted 3 days prior to treatment with bevacizumab at treatment 1 and 2 weeks after the end of treatment 4.
6. The method according to claim 1, wherein the MRI image feature extraction method in step (2) is: dividing two-dimensional interested areas, and stacking continuous two-dimensional interested areas to form a three-dimensional interested volume, namely a brain injury focus; and extracting image characteristics of the brain injury focus.
7. The method of claim 1, wherein the screening for image characteristics or clinical characteristics associated with a therapeutic response in steps (2) and (3) is performed such that an effective therapeutic response is defined as: (1) the neurological symptoms or signs were stable, the FLAIR sequence showed a reduction in brain injury volume of at least 25%; or (2) a reduction in the level of improvement in neurological symptoms or signs by at least a level 1 as an improvement, as assessed by the LENT/SOMA rating scale; .
8. A composite predictive model constructed according to the method of any one of claims 1 to 7.
9. The composite predictive model of claim 8, including the formula for calculating the effectiveness of bevacizumab treatment, said formula being: treatment response score 3.436 x imaging score-0.024 x time interval between radiotherapy and radioactive brain injury diagnosis (month) -0.043 x time interval between radioactive brain injury diagnosis and bevacizumab treatment (month) + 0.451; wherein:
the imagination score is 0.483308872
+0.996788068×LoG(σ＝1)_Firstorder_Maximum
-0.225130674×LoG(σ＝1)_GLSZM_LAHGLE
-0.227908898×LoG(σ＝2)_GLSZM_SALGLE
-1.174540209×LoG(σ＝3)_GLSZM_LAHGLE
+1.207131780×LoG(σ＝4)_GLSZM_SZNN
+2.523507898×Wavelet(HHH)_Firstorder_Skewness
+1.144733635×Wavelet(HHH)_Firstorder_Median
-0.401726675×Wavelet(HHH)_Firstorder_Kurtosis
-0.049392860×Wavelet(HHH)_GLRLM_LRE
+0.757594985×Wavelet(HHH)_GLSZM_SZNN
+0.416040381×Wavelet(HLL)_Firstorder_Median
+0.019379737×Wavelet(HLL)_Firstorder_Mean
-0.840176612×Wavelet(LHH)_GLDM_LDLGLE
-0.168430560×Wavelet(LHH)_GLSZM_SALGLE
-1.134782147×Wavelet(LLH)_GLSZM_LAHGLE
-0.627111233×Wavelet(HHL)_GLCM_IDMN
-0.573236112×Wavelet(HLH)_GLCM_Correlation
-0.946488391×GLCM_InverseVariance。
10. The application of image characteristics and clinical characteristics in the combined construction of a bevacizumab therapeutic radiation brain injury curative effect prediction model is characterized in that the image characteristics comprise first order _ Maximum, GLSZM _ LAHGLE, GLSZM _ SALGLE, GLSZM _ LAHGLE, GLSZM _ SZNN, Wavelet (HHH) _ first order _ Skewness, Wavelet (HHH) _ first order _ Median, Wavelet (HHH) _ GLRLM _ LRE, Wavelet (HHH) _ SZM _ SZNN, Wavelet (HLL) _ first order _ Kurtosis, Wavelet (HLL) _ first order _ Merton, Wavelet (HLH) _ Walsh _ LH, GLSZM _ SALGLE, GLVLLE _ GLVLL _ GLISE, Wavelet _ GLISE, and GLISHMM _ GLISE; the clinical features include the time interval between diagnosis of radioactive brain injury and bevacizumab treatment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010734699.XA CN111951962B (en) | 2020-07-27 | 2020-07-27 | Bevacizumab curative effect prediction model for treating radioactive brain injury and construction method thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010734699.XA CN111951962B (en) | 2020-07-27 | 2020-07-27 | Bevacizumab curative effect prediction model for treating radioactive brain injury and construction method thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111951962A true CN111951962A (en) | 2020-11-17 |
| CN111951962B CN111951962B (en) | 2022-03-11 |
Family
ID=73339662
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010734699.XA Active CN111951962B (en) | 2020-07-27 | 2020-07-27 | Bevacizumab curative effect prediction model for treating radioactive brain injury and construction method thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111951962B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113160965A (en) * | 2021-01-08 | 2021-07-23 | 福建医科大学附属第一医院 | Construction method and prediction method of small liver cancer early recurrence probability prediction model |
| CN113793324A (en) * | 2021-09-17 | 2021-12-14 | 山东师范大学 | Nasopharyngeal carcinoma induced chemotherapy curative effect prediction method and system |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120184840A1 (en) * | 2009-04-07 | 2012-07-19 | Kayvan Najarian | Automated Measurement of Brain Injury Indices Using Brain CT Images, Injury Data, and Machine Learning |
| US20170202519A1 (en) * | 2005-05-31 | 2017-07-20 | Algeomics, Llc | Method for integrating large scale biological data with imaging |
| CN109583447A (en) * | 2018-12-11 | 2019-04-05 | 山东省肿瘤防治研究院（山东省肿瘤医院） | A kind of image group credit analysis model building method and analysis method |
| CN110197236A (en) * | 2018-10-17 | 2019-09-03 | 中山大学附属第三医院 | A kind of prediction and verification method based on image group to glucocorticoid curative effect |
| CN110265095A (en) * | 2019-05-22 | 2019-09-20 | 首都医科大学附属北京佑安医院 | For HCC recurrence and construction method and the application of the prediction model and nomogram of RFS |
| CN110808093A (en) * | 2019-09-16 | 2020-02-18 | 中山大学孙逸仙纪念医院 | Prediction model for clinical prognosis of radioactive encephalopathy and construction method thereof |
| CN111128328A (en) * | 2019-10-25 | 2020-05-08 | 中山大学肿瘤防治中心（中山大学附属肿瘤医院、中山大学肿瘤研究所） | Nasopharyngeal carcinoma structured image report and data processing system and method |
| CN111311558A (en) * | 2020-02-09 | 2020-06-19 | 华中科技大学同济医学院附属协和医院 | Construction method of imaging omics model for pancreatic cancer prediction |
| CN111657945A (en) * | 2020-06-16 | 2020-09-15 | 中南大学湘雅医院 | Nasopharyngeal carcinoma prognosis auxiliary evaluation method based on enhanced MRI (magnetic resonance imaging) imaging omics |
-
2020
- 2020-07-27 CN CN202010734699.XA patent/CN111951962B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170202519A1 (en) * | 2005-05-31 | 2017-07-20 | Algeomics, Llc | Method for integrating large scale biological data with imaging |
| US20120184840A1 (en) * | 2009-04-07 | 2012-07-19 | Kayvan Najarian | Automated Measurement of Brain Injury Indices Using Brain CT Images, Injury Data, and Machine Learning |
| CN110197236A (en) * | 2018-10-17 | 2019-09-03 | 中山大学附属第三医院 | A kind of prediction and verification method based on image group to glucocorticoid curative effect |
| CN109583447A (en) * | 2018-12-11 | 2019-04-05 | 山东省肿瘤防治研究院（山东省肿瘤医院） | A kind of image group credit analysis model building method and analysis method |
| CN110265095A (en) * | 2019-05-22 | 2019-09-20 | 首都医科大学附属北京佑安医院 | For HCC recurrence and construction method and the application of the prediction model and nomogram of RFS |
| CN110808093A (en) * | 2019-09-16 | 2020-02-18 | 中山大学孙逸仙纪念医院 | Prediction model for clinical prognosis of radioactive encephalopathy and construction method thereof |
| CN111128328A (en) * | 2019-10-25 | 2020-05-08 | 中山大学肿瘤防治中心（中山大学附属肿瘤医院、中山大学肿瘤研究所） | Nasopharyngeal carcinoma structured image report and data processing system and method |
| CN111311558A (en) * | 2020-02-09 | 2020-06-19 | 华中科技大学同济医学院附属协和医院 | Construction method of imaging omics model for pancreatic cancer prediction |
| CN111657945A (en) * | 2020-06-16 | 2020-09-15 | 中南大学湘雅医院 | Nasopharyngeal carcinoma prognosis auxiliary evaluation method based on enhanced MRI (magnetic resonance imaging) imaging omics |
Non-Patent Citations (1)
| Title |
|---|
| 周晶 等: "MRI 影像组学在脑胶质瘤分级中的价值研究", 《中国微侵袭神经外科杂志》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113160965A (en) * | 2021-01-08 | 2021-07-23 | 福建医科大学附属第一医院 | Construction method and prediction method of small liver cancer early recurrence probability prediction model |
| CN113160965B (en) * | 2021-01-08 | 2023-11-17 | 福建医科大学附属第一医院 | Construction method and prediction method of small liver cancer early recurrence probability prediction model |
| CN113793324A (en) * | 2021-09-17 | 2021-12-14 | 山东师范大学 | Nasopharyngeal carcinoma induced chemotherapy curative effect prediction method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111951962B (en) | 2022-03-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mu et al. | Radiomics of 18 F-FDG PET/CT images predicts clinical benefit of advanced NSCLC patients to checkpoint blockade immunotherapy | |
| Charron et al. | Automatic detection and segmentation of brain metastases on multimodal MR images with a deep convolutional neural network | |
| Win et al. | Pulmonary 18 F-FDG uptake helps refine current risk stratification in idiopathic pulmonary fibrosis (IPF) | |
| CN111951962B (en) | Bevacizumab curative effect prediction model for treating radioactive brain injury and construction method thereof | |
| Castillo et al. | Pre-radiotherapy FDG PET predicts radiation pneumonitis in lung cancer | |
| Lee et al. | Prediction of IDH1 mutation status in glioblastoma using machine learning technique based on quantitative radiomic data | |
| US20190247000A1 (en) | Prediction Model For Grouping Hepatocellular Carcinoma, Prediction System Thereof, And Method For Determining Hepatocellular Carcinoma Group | |
| Luo et al. | A nomogram based on pretreatment CT radiomics features for predicting complete response to chemoradiotherapy in patients with esophageal squamous cell cancer | |
| US9585627B2 (en) | Histological differentiation grade prediction of hepatocellular carcinoma in computed tomography images | |
| Yang et al. | Treatment response prediction using ultrasound-based pre-, post-early, and delta radiomics in neoadjuvant chemotherapy in breast cancer | |
| Wen et al. | Normal tissue complication probability (NTCP) models for predicting temporal lobe injury after intensity-modulated radiotherapy in nasopharyngeal carcinoma: A large registry-based retrospective study from China | |
| Tran et al. | Predictive quantitative ultrasound radiomic markers associated with treatment response in head and neck cancer | |
| Harat et al. | Combining amino acid PET and MRI imaging increases accuracy to define malignant areas in adult glioma | |
| Ye et al. | Classification of gliomas and germinomas of the basal ganglia by transfer learning | |
| Zhang et al. | A deep learning model for the differential diagnosis of benign and malignant salivary gland tumors based on ultrasound imaging and clinical data | |
| Wang et al. | Automatic segmentation of vestibular schwannomas from T1-weighted MRI with a deep neural network | |
| Jinghua et al. | Clinical prospective study of Gallium 68 (68Ga)–labeled fibroblast-activation protein inhibitor PET/CT in the diagnosis of biliary tract carcinoma | |
| Cao et al. | Application value of multiparametric MRI for evaluating iron deposition in the substantia nigra in Parkinson's disease | |
| CN115797308A (en) | DCE-MRI-based breast tumor segmentation method | |
| Prats-Climent et al. | Artificial intelligence on FDG PET images identifies mild cognitive impairment patients with neurodegenerative disease | |
| Cui et al. | MRI-based texture analysis for differentiate between pediatric posterior fossa ependymoma type A and B | |
| Chiang et al. | Treatment and survival of patients with small cell lung cancer and brain metastasis | |
| Wang et al. | Combination of hematoma volume and perihematoma radiomics analysis on baseline CT scan predicts the growth of perihematomal edema | |
| Wang et al. | Deep transfer learning radiomics based on two-dimensional ultrasound for predicting the efficacy of neoadjuvant chemotherapy in breast cancer | |
| Wang et al. | Deep Learning-Enabled Clinically Applicable CT Planbox for Stroke With High Accuracy and Repeatability |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |