ABSTRACT
@#Objective To construct a radiomics model for identifying clinical high-risk carotid plaques. Methods A retrospective analysis was conducted on patients with carotid artery stenosis in China-Japan Friendship Hospital from December 2016 to June 2022. The patients were classified as a clinical high-risk carotid plaque group and a clinical low-risk carotid plaque group according to the occurrence of stroke, transient ischemic attack and other cerebrovascular clinical symptoms within six months. Six machine learning models including eXtreme Gradient Boosting, support vector machine, Gaussian Naive Bayesian, logical regression, K-nearest neighbors and artificial neural network were established. We also constructed a joint predictive model combined with logistic regression analysis of clinical risk factors. Results Finally 652 patients were collected, including 427 males and 225 females, with an average age of 68.2 years. The results showed that the prediction ability of eXtreme Gradient Boosting was the best among the six machine learning models, and the area under the curve (AUC) in validation dataset was 0.751. At the same time, the AUC of eXtreme Gradient Boosting joint prediction model established by clinical data and carotid artery imaging data validation dataset was 0.823. Conclusion Radiomics features combined with clinical feature model can effectively identify clinical high-risk carotid plaques.
ABSTRACT
@#Objective To investigate the radiomics features to distinguish invasive lung adenocarcinoma with micropapillary or solid structure. Methods A retrospective analysis was conducted on patients who received surgeries and pathologically confirmed invasive lung adenocarcinoma in our hospital from April 2016 to August 2019. The dataset was randomly divided into a training set [including a micropapillary/solid structure positive group (positive group) and a micropapillary/solid structure negative group (negative group)] and a testing set (including a positive group and a negative group) with a ratio of 7∶3. Two radiologists drew regions of interest on preoperative high-resolution CT images to extract radiomics features. Before analysis, the intraclass correlation coefficient was used to determine the stable features, and the training set data were balanced using synthetic minority oversampling technique. After mean normalization processing, further radiomics features selection was conducted using the least absolute shrinkage and selection operator algorithm, and a 5-fold cross validation was performed. Receiver operating characteristic (ROC) curves were depicted on the training and testing sets to evaluate the diagnostic performance of the radiomics model. Results A total of 340 patients were enrolled, including 178 males and 162 females with an average age of 60.31±6.69 years. There were 238 patients in the training set, including 120 patients in the positive group and 118 patients in the negative group. There were 102 patients in the testing set, including 52 patients in the positive group and 50 patients in the negative group. The radiomics model contained 107 features, with the final 2 features selected for the radiomics model, that is, Original_ glszm_ SizeZoneNonUniformityNormalized and Original_ shape_ SurfaceVolumeRatio. The areas under the ROC curve of the training and the testing sets of the radiomics model were 0.863 (95%CI 0.815-0.912) and 0.857 (95%CI 0.783-0.932), respectively. The sensitivity was 91.7% and 73.7%, the specificity was 78.8% and 84.0%, and the accuracy was 85.3% and 78.4%, respectively. Conclusion There are differences in radiomics features between invasive pulmonary adenocarcinoma with or without micropapillary and solid structures, and the radiomics model is demonstrated to be with good diagnostic value.
ABSTRACT
Objective To utilize sophisticated CT-driven radiomics to prognosticate the mutation situation of KRAS in patients with colorectal cancer(CRC).Methods A total of 393 patients who underwent KRAS mutation testing and preoperative triphasic enhanced CT were analyzed retrospectively.All patients were divided into training group(n=276)and validation group(n=117)with a ratio of 7∶3.The characteristics tightly associated with KRAS mutation were extracted and screened to construct three models,include clinical,radiomics,and clinical-radiomics fusion models for prediction of KRAS mutation.The performance and clinical utility of these three models were assessed via receiver operating characteristic(ROC)curve and decision curve analysis(DCA).Results The study identified significant correlations between KRAS mutation and CEA,CA199,and a set of 13 radiomics features,respective-ly.Based on these clinical indicators and radiomics features,clinical,radiomics,and clinical-radiomics fusion models were constructed to prognosticate KRAS mutation.The radiomics model construc-ted in this study had good performance for the prediction of KRAS mutation status in CRC patients.Most notably,a clinical-radiomics nomogram that amalgamated both clinical risk factors and radiomics parameters emerged as the most effective predictor of KRAS mutation,with an area under the curve(AUC)of 0.782 and 0.744 in the training group and validation group,respectively.Conclusion The refined CT radiomics model serves as a robust,non-invasive,quantitative tool for the assessment of KRAS mutation status in CRC patients.
ABSTRACT
Objective To investigate the diagnostic value of diffusion weighted imaging(DWI)-based radiomics model to identify small hepatocellular carcinoma(<2 cm)(SHCC)and dysplastic nodule(DN)in the background of hepatitis cirrhosis.Methods A total of 93 cases SHCC and 25 cases with DN with complete enhanced MRI images and surgically pathologically confirmed were collected retrospectively.Chi-square test was performed to analyze the signal characteristics of DWI and enhanced triphasic MRI images between the two groups.ITK-SNAP was used to draw the region of interest(ROI)on DWI,and FAE software was applied for extraction,selec-tion,and construction of support vector machine(SVM)models(dividing into training set and test set according to the ratio of 7∶3).The diagnostic performance of model was evaluated by receiver operating characteristic(ROC)curve.Results There were statisti-cally significant differences in enhanced triphasic MRI and DWI between SHCC and DN(P<0.05).The area under the curve(AUC)of the DWI-SVM model training set was 0.936,and sensitivity,specificity and accuracy was 95.4%,88.2%and 93.9%,respec-tively,and the AUC of the test set was 0.911,and sensitivity,specificity and accuracy was 85.7%,87.5%and 86.1%,respectively,which were all significantly better than the diagnostic efficacy of DWI(AUC=0.720).Conclusion DWI-SVM model with signifi-cantly higher AUC and specificity can effectively differentiate SHCC from DN in the background of hepatitis cirrhosis.
ABSTRACT
Objective To construct a radiomics nomogram combining clinical and a radiomics signature for distinguishing type Ⅱpapillary renal cell carcinoma(pRCC)from atypical clear cell renal cell carcinoma(ccRCC).Methods Clinical and CT data of patients with pathologically confirmed type Ⅱ pRCC(62 cases)and atypical ccRCC(56 cases)were analyzed.A random sample was divided into a training set(82 cases)and a test set(36 cases)in a ratio of 7∶3.Clinical factors were screened to construct clinical factor models.A total of 1 595 radiomics features of tumors were extracted from the corticomedullary phase CT images and based on the most effective features to construct a radiomics signature and calculate the radiomics score(Rad-score).A radiomics nomogram was constructed by combining the Rad-score and independent clinical factors.Receiver operating characteristic(ROC)curve was used to assess the clini-cal usefulness of the models.Decision curve analysis(DCA)was used to assess the difference between the models.Results The radiomics signature showed good discrimination in training set area under the curve(AUC)0.894[95%confidence interval(CI)0.834-0.947]and test set AUC 0.879(95%CI 0.774-0.963).The AUC of the clinical factors model in training set and test set were 0.725(95%CI 0.646-0.804)and 0.698(95%CI 0.567-0.819).The AUC of the radiomics nomogram in training set and test set were 0.901(95%CI 0.840-0.953)and 0.901(95%CI 0.809-0.975).DCA demonstrated the radiomics nomogram outmatched the clinical factors model and radiomics signature in the aspects of clinical usefulness.Conclusion Radiomics nomogram based on enhanced CT can provide good prediction of type Ⅱ pRCC and atypical ccRCC preoperatively,improve the diagnostic accuracy and provide guidance for future clinical treatment.
ABSTRACT
Objective To investigate the significance of intratumoral and peritumoral radiomics models in predicting occult lymph node metastasis in stage T1 non-small cell lung cancer(NSCLC)and to compare the predictive accuracy in different peritumoral radiomics models.Methods The CT images and clinical data of 211 patients without lymph node metastasis on preoperative CT examination and pathologically confirmed NSCLC after surgery were collected.The radiomics features were derived from the three-dimensional volume of interest(VOI)of the intratumoral and peritumoral at 3-,5-,and 10-mm following lesion segmentation on CT images of each patient.The feature data of all nidus were radomly divide into training set and validation set with a ratio of 7︰3.The Pearson or Spearman correlation test was performed to remove redundancy.Dimensionality was reduced by the least absolute shrinkage and selection operator(LASSO)regression analysis.The linear combination of selected features and corresponding coefficients were used to construct the Radiomics score(Radscore).The clinical model and comprehensive model were constructed by logistic regression analysis.The conprehensive model was visualized with the nomogram,and its performance was evaluated.Results Among the peritumoral radiomics models,the peritumoral 5-mm model showed the best predictive efficacy[validation set,area under the curve(AUC)0.771].The comprehensive model containing Radscore,CT image features and CEA exhibited the best performance(validation set,AUC 0.850).Conclusion Intratumoral and peritumoral radiomics models perform efficiently in predicting occult lymph node metastasis in stage T1 NSCLC,and nomogram can effectively and noninvasively predict occult lymph node metastasis in NSCLC.
ABSTRACT
Objective To investigate the value of multimodal MRI radiomics in predicting muscle-invasive bladder cancer.Methods A total of 178 patients with pathology diagnosis of bladder cancer were retrospectively collected,including 31 cases of muscle invasive bladder cancer(MIBC)and 147 cases of non-muscle invasive bladder cancer(NMIBC).Patients were randomly divided into training group and testing group at a ratio of 7︰3.The range of bladder tumors in T2WI,diffusion weighted imaging(DWI)and apparent diffusion coefficient(ADC)images were segmented as volume of interest(VOI)by using ITK-SNAP software.Radiomics features were extracted through A.K software.The optimal radiomics features were obtained through radiomics algorithm and least absolute shrinkage and selection operator(LASSO)method.Finally,the logistic regression analysis method and random forest model method were used to construct prediction models.The performance of prediction models was evaluated by the receiver operating characteristic(ROC)curve.Results This study constructed four groups of models containing T2WI prediction model,DWI prediction model,ADC prediction model,and T2WI+DWI+ADC prediction model.The area under the curve(AUC)of T2WI,DWI,and ADC prediction models for identifying MIBC and NMIBC were separately 0.920,0.914,and 0.954 in the training group while those were respectively 0.881,0.773,and 0.871 in the testing group.There was no statistical significance between T2WI,DWI,and ADC prediction models.In training and testing groups,the AUC of T2WI+DWI+ADC prediction model were respectively 0.959 and 0.909,which were higher than the single sequence prediction model.The sensitivity and specificity of the training group were 0.905 and 0.853 and the sensitivity and specificity of the testing group were 0.778 and 0.795.Conclusion MRI radiomics prediction model can effectively differentiate MIBC and NMIBC.The T2WI+DWI+ADC prediction model shows better prediction efficiency.
ABSTRACT
Objective To explore the value of dual-phase enhanced CT radiomics in predicting post-acute pancreatitis diabetes mellitus(PPDM-A).Methods A total of 145 patients with acute pancreatitis(AP)were retrospectively collected,including 62 patients in PPDM-A group and 83 patients in non-PPDM-A group.The patients were randomly divided into training set and test set at a ratio of 7︰3,the pancreatic parenchyma in arterial phase and venous phase was delineated and the radiomics features were extracted.Vari-ance threshold method,univariate selection method and least absolute shrinkage and selection operator(LASSO)were used to screen radiomics features.The prediction performance of the model was evaluated by the area under the curve(AUC).The DeLong test was used to compare the prediction efficiency between the models,and the calibration curve and decision curve were used to evaluate the prediction efficiency of the model.Results The AUC of arterial phase model,venous phase model,combined arterial venous phase model,clinical model and radiomics combined clinical model in the training set were 0.845,0.792,0.829,0.656 and 0.862,respec-tively.The DeLong test results showed that only the difference between the radiomics combined clinical model and the clinical model in the training set and the test set was statistically significant(P<0.05).The decision curve showed that the radiomics combined clinical model had high clinical practicability in a certain range,and the calibration curve showed that the radiomics combined clinical model had the best fitting degree with the actual observation value.Conclusion Based on the dual-phase enhanced CT radiomics combined clinical model,PPDM-A can be predicted more accurately,and it can provide a certain reference value for the clinical development of per-sonalized treatment programs.
ABSTRACT
Objective To investigate the correlation between intra-and peri-tumoral radiomics features and the response to con-current chemoradiotherapy(CCRT)in cervical squamous cell carcinoma,and to explore the difference of predictive performance between 2D and 3D radiomics models.Methods The imaging data of 132 patients were analyzed retrospectively and randomly divided into training set(n=92)and validation set(n=40).Radiomics features were extracted based on the dynamic contrast-enhanced magnetic resonance imaging(DCE-MRI),the correlation analysis and least absolute shrinkage and selection operator(LASSO)algorithm were used for dimensionality reduction and screening,then the radiomics score was calculated and the logistic model was constructed.The receiver operating characteristic(ROC)curve,internal validation of Bootstrap and Brier score were used to evaluate the discrimina-tion and calibration of the model,and the improvement in predictive performance of 3D model compared with 2D model was evaluated by the integrated discrimination improvement(IDI).Results In the training set,the ROC curve showed that the area under the curve(AUC)of the models(2D-intratumoral,3D-intratumoral,3D-peritumoral,3D-combined)ranged from 0.774 to 0.893.The internal validation of Bootstrap showed the AUC were 0.772,0.860,0.847 and 0.888,respectively,while in the validation set,the AUC were 0.757,0.849,0.824 and 0.887,respectively.The Brier scores indicated that the models were well calibrated.In the training set and the validation set,the IDI values were 0.155 and 0.179,respectively,and the differences were statistically significant(P<0.05).Conclusion The radiomics analysis based on the tumor volume can fully explore the tumor heterogeneity.The intra-and peri-tumoral radiomics combined model shows the best predictive performance,which is important to assist clinicians in developing individualized therapies.
ABSTRACT
Objective To investigate the application value of quantitative computed tomography(QCT)-based imaging histology in the diagnosis of clinical osteoporosis.Methods A total of 182 patients who underwent QCT scans of the chest or abdomen were ana-lyzed retrospectively,and the patients were divided into osteoporosis group(56 cases)and non-osteoporosis group(126 cases)accord-ing to the bone mineral density(BMD)values measured by QCT.The cases were randomly divided into training set(110 cases)and validation set(72 cases)at a ratio of 6︰4.The L1-L2 vertebrae were outlined with the region of interest(ROI)and the image features were extracted using ITK-SNAP 3.6.0 and A.K.software.(1)Radiomics score(Rad-score)model was established using maximum relevance minimum redundancy(mRMR)and least absolute shrinkage and selection operator(LASSO)algorithm.(2)The L1 and L2 vertebrae BMD value and T-value from the dual energy X-ray absorptiometry(DXA)examination of the patient were obtained to build the model separately for analysis and for comparison with the Rad-score model.(3)The discriminative ability,clinical applica-tion performance and calibration ability of Rad-score in the training and validation sets were evaluated using receiver operating charac-teristic(ROC)curves,decision curve analysis(DCA)and calibration curves.(4)Data on sensitivity,specificity,accuracy and area under the curve(AUC)were used to compare the predictive ability of DXA and Rad-score.DeLong test was used for comparison of differences between Rad-score and DXA models.Results Four optimal ima-ging histology features were finally selected to create Rad-score model,which confirmed the significant correlation between Rad-score and QCT for the diagnosis of osteoporosis.The calibration curve showed that Rad-score model had a good fit in both the training set and the validation set.The results of the DeLong test showed that the AUC of Rad-score model were greater than those of DXA model.Conclusion The QCT-based imaging histology model has high sen-sitivity,specificity and accuracy,with outstanding advantages and good performance for osteoporosis diagnosis,and is superior to the DXA model.
ABSTRACT
Objective To explore the value of radiomics in differential diagnosis of small cell lung cancer(SCLC)and non-small cell lung cancer(NSCLC).Methods Literature on the differential diagnosis of SCLC and NSCLC using radiomics was searched in Chinese and English databases.After literature screening and data extraction,Meta-DiSc1.4 and State16.0 SE software were used for analysis.Results A total of 910 patients were included in 8 studies.Meta-analysis results showed that the radiomics differential diag-nosis of SCLC and NSCLC had high co-sensitivity(Sen)and specificity(Spe),0.87[95%confidence interval(CI)0.83-0.91]and 0.88(95%CI 0.85-0.90),respectively.Meta-regression analysis showed that heterogeneity was not caused by feature extraction software type,joint machine learning,image pattern,brain metastasis,and sample size.Publication bias results didn't show any sig-nificant publication bias.Conclusion The radiomics method can differentiate and diagnose SCLC from NSCLC more accurately.When Matlab software is used to extract MRI image features combined with machine learning,and the sample size is large enough,the radiomics can differentiate and diagnose SCLC from NSCLC more accurately.
ABSTRACT
BACKGROUND:Previous studies on cervical instability failed to explain the dynamic and static interaction relationship and pathological characteristics changes in the development of cervical lesions under the traditional imaging examination.In recent years,the emerging nuclear magnetic resonance imaging(MRI)radiomics can provide a new way for in-depth research on cervical instability. OBJECTIVE:To investigate the application value of MRI radiomics in the study of cervical instability. METHODS:Through recruitment advertisements and the Second Department of Spine of Wangjing Hospital,China Academy of Chinese Medical Sciences,young cervical vertebra unstable subjects and non-unstable subjects aged 18-45 years were included in the cervical vertebra nuclear magnetic image collection.Five specific regions of interest,including the intervertebral disc region,the facet region,the prevertebral muscle region,the deep region of the posterior cervical muscle group,and the superficial region of the posterior cervical muscle group,were manually segmented to extract and screen the image features.Finally,the cervical instability diagnosis classification model was constructed,and the effectiveness of the model was evaluated using the area under the curve. RESULTS AND CONCLUSION:(1)A total of 56 subjects with cervical instability and 55 subjects with non-instability were included,and 1 688 imaging features were extracted for each region of interest.After screening,300 sets of specific image feature combinations were obtained,with 60 sets of regions of interest for each group.(2)Five regions of interest diagnostic classification models for cervical instability were initially established.Among them,the support vector machine model for the articular process region and the support vector machine model for the deep cervical muscle group had certain accuracy for the classification of instability and non-instability,and the average area under the curve of ten-fold cross-validation was 0.719 7 and 0.703 3,respectively.(3)The Logistic model in the intervertebral disc region,the LightGBM model in the prevertebral muscle region,and the Logistic model in the superficial posterior cervical muscle region were generally accurate in the classification of instability and non-instability,and the average area under the curve of ten-fold cross-validation was 0.650 4,0.620 7,and 0.644 2,respectively.(4)This study proved the feasibility of MRI radiomics in the study of cervical instability,further deepened the understanding of the pathogenesis of cervical instability,and also provided an objective basis for the accurate diagnosis of cervical instability.
ABSTRACT
Ductal carcinoma in situ(DCIS),a pathological type of breast cancer that is limited to the terminal ducts of the breast without breaking through the basement membrane,is considered as the precursor of invasive ductal carcinoma(IDC).When DCIS breaks through the basement membrane and invades surrounding tissues,it can form infiltrating lesions.If the maximum diameter of a single infiltrating lesion is less than 1mm or the maximum diameter of multiple infiltrating lesions is less than 1mm,it is defined as ductal carcinoma in situ with microinvasion(DCIS-Mi).About 12%-40%of untreated and intervened DCIS will progress to IDC,and DCIS and IDC can also coexist.However,there is a considerable portion of DCIS that never progresses with good prognosis.Recently,overdiagnosis and overtreatment of DCIS have become the research hotspots.The histological grade of DCIS is mainly based on the morphology of the nucleus,which is divided into three nuclear levels:low,medium,and high.There are also significant differences in receptor expression and molecular type distribution between DCIS,DCIS-Mi,and IDC.For DCIS with or without microinvasion as well as different histological grades,there are many controversies about the treatment regimen,clinical prognosis and risk.The development of modern imaging technology has achieved preliminary evaluation of histological grading,infiltration status,and prognosis prediction of DCIS.The most commonly used breast imaging techniques in clinical practice currently include mammography(MG),ultrasound(US),and magnetic resonance imaging(MRI).The imaging principles of these three techniques are different,and each has its own advantages and disadvantages in breast disease imaging diagnosis.However,they can complement each other and play an important role in disease diagnosis,treatment,and prognosis evaluation.Mammography has the advantages of safety,reliability and good repeatability.It is the preferred screening method for breast cancer recommended by international guidelines.The main manifestations of DCIS on MG can be divided into non calcified lesions and calcified lesions.On US,the main manifestations are lesions and non-lesion type,which can be further divided into hypoechoic changes,calcification,ductal changes,and structural disorders and distortions.MRI has higher sensitivity in detecting DCIS without calcification and multifocal DCIS compared with MG,and has higher accuracy in evaluating the lesion range.However,there are also shortcomings such as low diagnostic specificity and insensitivity to microcalcification display.In addition,radiomics has great potential in the histopathological evaluation,prediction,and guidance of individualized precision treatment of DCIS.In the current era of precision medicine,image features,histopathology,molecular genes,etc.are increasingly significant in predicting the prognosis of breast cancer.The early accurate diagnosis and molecular type of DCIS are also extremely important in clinical work.It has become a consensus in clinical treatment to predict the potential benefits of different treatments through molecular typing,histological grade,and imaging findings,in order to develop the most suitable personalized treatment plan.This article reviewed the correlation between imaging features and the molecular subtype,histopathology and prognosis of DCIS.
ABSTRACT
In recent years,radiomics has been developed rapidly in the field of clinical medicine,and the artificial intelligence(AI)technology has been utilized to assist diagnosis.This paper introduced the background of radiomics,analyzed the basic research process of radiomics,and looked forward to its application in the identification of bone and joint injuries in the field of forensic medicine.Reviewing the three aspects is expected to provide a theoretical foundation of radiomics,which will be helpful to develop its application in forensic medicine.
ABSTRACT
Objective To observe the value of radiomics models based on gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid(Gd-EOB-DTPA)enhanced hepatobiliary phase(HBP)MRI for assessing clinical pathological stage of hepatic fibrosis(HF).Methods Data of 240 patients with pathologically/clinically diagnosed and clinical pathological staged HF who underwent Gd-EOB-DTPA enhanced MR examination were retrospectively analyzed.The liver-to-muscle signal intensity ratio(SIR1)and liver-to-spleen signal intensity ratio(SIR2)were measured based on HBP images.Radiomics features of HBP images were extracted and screened to construct radiomics models.The signal intensity ratio(SIR)-radiomics combined models were constructed based on SIR and radiomics signatures.Receiver operating characteristic(ROC)curves were drawn to evaluate the efficacy of each model for assessing clinical pathological stage of HF.Results The area under the curve(AUC)of SIR1 and SIR2 models for assessing clinical pathological stage of HF were 0.63-0.70 and 0.65-0.71,respectively.The most effective radiomics model for assessing HF,significant HF,advanced HF and early cirrhosis was support vector machine(SVM),SVM,light gradient boosting machine and K-nearest neighbor model,respectively,with the AUC in validation set of 0.87,0.82,0.81 and 0.80,respectively,while the AUC of SIR-radiomics combined models in validation set of 0.88,0.82,0.82 and 0.81,respectively.Conclusion The radiomics models based on Gd-EOB-DTPA enhanced HBP MRI were helpful for assessing clinical pathological stage of HF.Combining with HBP SIR could improve their efficacy.
ABSTRACT
Objective To explore the value of CT radiomics combined with clinical data and CT features for predicting TNM stage of thymic epithelial tumor(TET).Methods Data of 216 single TET patients confirmed by surgical pathology were retrospectively analyzed.Totally 151 cases with TNM stage Ⅰ TET were divided into early group,while 27 with TNM stage Ⅲ and 38 with TNM stage Ⅳ TET were divided into late group(n=65).Univariate analysis was used to analyze clinical data and chest CT manifestations.Based on non-contrast-enhanced CT(NECT)and contrast-enhanced CT(CECT),the best radiomics features were extracted and screened to establish radiomics models(RMNECT,RMCECT)for predicting TNM stage of TET.RMNECT-clinic,RMCECT-clinic,RMNECT-clinic-CT and RMCECT-clinic-CT were constructed based on combination of clinical and CT features being significantly different between groups,respectively.The patients were divided into training set(n=151)and validation set(n=65)at the ratio of 7∶3.The above models were trained in the training set using repeated 5-fold cross validation method,and their efficacy were verified in the validation set.Results Significant differences of clinical symptoms and CT manifestations including fat infiltration around the lesion,mediastinal lymph node enlargement and pleural effusion were found between groups(all P<0.05).Based on NECT and CECT,2 and 9 best radiomics features were selected to construct the corresponding models.In validation set,the area under the curve(AUC)of RMNECT-clinic-CT for predicting TNM stage of TET(0.864)was higher than that of RMNECT and RMNECT-clinic(AUC=0.634,0.721,Z=3.081,2.937,P=0.002,0.003),while AUC of RMCECT-clinic-CT(0.920)was also higher than that of RMCECT and RMCECT-clinic(AUC=0.689,0.751,Z=2.698,2.390,P=0.007,0.017).Conclusion CT radiomics combined with clinical data and CT features could effectively predict TNM stage of TET.
ABSTRACT
Objective To observe the value of radiomics combined with CT features for distinguishing mycoplasma pneumonia(MP)and non-MP in children.Methods Data of 153 children with pneumonia were retrospectively analyzed.The children were divided into MP group(n=101)and non-MP group(n=52)according to mycoplasma RNA detection,and also were divided into training set(n=107,including 71 MP and 36 non-MP)and validation set(n=46,including 30 MP and 16 non-MP)at the ratio of 7∶3.CT findings were compared between groups.Six best CT features were selected in training set using F test algorithm,and a CT model was constructed using logistic regression(LR)method.The best radiomics features were extracted and screened in training set,and machine learning(ML)models were constructed using LR,support vector machine(SVM),random forest(RF),linear discriminant analysis(LDA)and stochastic gradient descent(SGD)classifiers,respectively.Based on the best CT features and radiomics features,CT-ML models were constructed using the above classifiers.Receiver operating characteristic curves were drawn,and the areas under the curve(AUC)were calculated,the efficacy of each model for distinguishing MP and non-MP was evaluated.Results Lesions involved the upper,middle and lower lobe of right lung,thickened bronchial wall,tree bud sign and edge retract sign were the best CT features.AUC of CTLR was 0.710,of MLLR,MLSVM,MLRF,MLLDA and MLSGD in validation set was 0.715,0.663,0.623,0.706 and 0.494,respectively,and MLLR was the optimal radiomics model.AUC of CT-MLLR,CT-MLSVM,CT-MLRF,CT-MLLDA and CT-MLSGD in validation set was 0.813,0.823,0.649,0.796 and 0.665,respectively,and CT-MLSVM was the optimal CT-ML model.In training set,AUC of CT-MLSVM(0.840)was higher than that of CTLR and MLLR model(AUC=0.713,0.740,both P<0.05).In validation set,no significant difference of AUC was found among CTLR,MLLR and CT-MLSVM(AUC=0.710,0.715 and 0.823,all P>0.05).Conclusion Radiomics combined with CT features could effectively distinguish MP and non-MP in children.
ABSTRACT
Objective To observe the value of intratumoral and peritumoral radiomics based on diffusion weighted imaging(DWI)for predicting histological grade of breast cancer.Methods Preoperative DWI data of 700 patients with single breast cancer diagnosed by pathology were retrospectively analyzed.The patients were divided into training set(n= 560,including 381 of grade Ⅰ+Ⅱ and 179 of grade Ⅲ)and test set(n=140,including 95 of grade Ⅰ+Ⅱ and 45 of grade Ⅲ)at the ratio of 8∶2.Intratumoral ROI(ROIintra)was manually delineated on DWI,which was automatically expanded by 3 mm and 5 mm to decline peritumoral ROI(ROIperi,including ROI3 mm and ROI5 mm),then intratumoral-peritumoral ROI(ROIintra+3 mm,ROIintra+5 mm)were obtained.The optimal radiomics features were extracted and screened,and the radiomics model(RM)for predicting the histological grade of breast cancer were constructed.Receiver operating characteristic curves were drawn,and the areas under the curve(AUC)were calculated to evaluate the predictive efficacy of each model.Calibration curve method was used to evaluate the calibration degree,while decision curve analysis(DCA)was performed to explore the clinical practicability of each model.Results AUC of RMintra,RM+3 mm,RM+5mm,RMintra+3 mm and RMintra+5 mm was 0.750,0.724,0.749,0.833 and 0.807 in training set,while was 0.723,0.718,0.736,0.759 and 0.782 in test set,respectively.In training set,significant differences of AUC was found(all P<0.01),while in test set,no significant difference of AUC was found among models(all P>0.05).The calibrations of models were all high.DCA showed that taken 0.02-0.88 as the threshold,the clinical net benefit of RMintra+per were greater in training set,while taken 0.40-0.72 as the threshold,the clinical net benefit of RMintra+per was greater in test set.Conclusion Both DWI intratumoral and peritumoral radiomics could effectively predict histological grade of breast cancer.Combination of intratumoral and peritumoral radiomics was more effective.
ABSTRACT
Radiomics is a new diagnostic and treatment technology,which can be employed to extract high-throughput radiomics features from CT,MR,and PET images and screen features closely related to diagnostic and treatment purposes,so as to accurately predict tumor or disease classification,prognosis,or genomic changes.Ki-67 is a type of nuclear protein,which is present only in nuclei of proliferative and dividing cells but not in those of quiescent phase cells;hence,it can be used as a predictor of cell proliferation and has been proven to be closely related to prognosis of lung cancer.This article reviews the mechanism and progress in radiomics research related to Ki-67 in lung cancer.
ABSTRACT
Objective To explore the predictive value of CT radiomics and morphological features for the prognosis and survival in non-small cell lung cancer(NSCLC)patients.Methods The clinic data of 300 NSCLC patients(300 lesions)were downloaded from the Cancer Imaging Archive,with 210 randomly selected as the training set and 90 as the test set.According to the prognosis and survival,the patients were divided into two groups with survival period≤3 and>3 years.3D Slicer software was used to delineate the regions of interest layer by layer in CT images,and the radiomics features were extracted from each region of interest.Both t-test and least absolute shrinkage and selection operator were utilized for radiomics feature screening.Three types of prediction models,namely radiomics model,morphological model and combined model,were constructed with Logistic regression,whose performances were evaluated using the receiver operating characteristic(ROC)curve.Results The differences in radiomics labels and mediastinal lymph node metastasis between the training set and the test set were statistically significant.For radiomics model,morphological model and combined model,the area under the ROC curve was 0.784(95%CI:0.722-0.847),0.734(95%CI:0.664-0.804)and 0.748(95%CI:0.680-0.815)in the training set,and 0.737(95%CI:0.630-0.844),0.665(95%CI:0.554-0.777)and 0.687(95%CI:0.578-0.797)in the test set,which demonstrated that radiomics model had the best diagnostic performance.Conclusion The CT radiomics model can effectively predict the prognosis and survival in NSCLC patients.