CT radiomics-based machine learning classification of atypical cartilaginous tumours and appendicular chondrosarcomas.
Ontology highlight
ABSTRACT: Background
Clinical management ranges from surveillance or curettage to wide resection for atypical to higher-grade cartilaginous tumours, respectively. Our aim was to investigate the performance of computed tomography (CT) radiomics-based machine learning for classification of atypical cartilaginous tumours and higher-grade chondrosarcomas of long bones.Methods
One-hundred-twenty patients with histology-proven lesions were retrospectively included. The training cohort consisted of 84 CT scans from centre 1 (n=55 G1 or atypical cartilaginous tumours; n=29 G2-G4 chondrosarcomas). The external test cohort consisted of the CT component of 36 positron emission tomography-CT scans from centre 2 (n=16 G1 or atypical cartilaginous tumours; n=20 G2-G4 chondrosarcomas). Bidimensional segmentation was performed on preoperative CT. Radiomic features were extracted. After dimensionality reduction and class balancing in centre 1, the performance of a machine-learning classifier (LogitBoost) was assessed on the training cohort using 10-fold cross-validation and on the external test cohort. In centre 2, its performance was compared with preoperative biopsy and an experienced radiologist using McNemar's test.Findings
The classifier had 81% (AUC=0.89) and 75% (AUC=0.78) accuracy in identifying the lesions in the training and external test cohorts, respectively. Specifically, its accuracy in classifying atypical cartilaginous tumours and higher-grade chondrosarcomas was 84% and 78% in the training cohort, and 81% and 70% in the external test cohort, respectively. Preoperative biopsy had 64% (AUC=0.66) accuracy (p=0.29). The radiologist had 81% accuracy (p=0.75).Interpretation
Machine learning showed good accuracy in classifying atypical and higher-grade cartilaginous tumours of long bones based on preoperative CT radiomic features.Funding
ESSR Young Researchers Grant.
SUBMITTER: Gitto S
PROVIDER: S-EPMC8170113 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:BackgroundAtypical cartilaginous tumour (ACT) and grade II chondrosarcoma (CS2) of long bones are respectively managed with watchful waiting or curettage and wide resection. Preoperatively, imaging diagnosis can be challenging due to interobserver variability and biopsy suffers from sample errors. The aim of this study is to determine diagnostic performance of MRI radiomics-based machine learning in differentiating ACT from CS2 of long bones.MethodsOne-hundred-fifty-eight patients with surgically treated and histology-proven cartilaginous bone tumours were retrospectively included at two tertiary bone tumour centres. The training cohort consisted of 93 MRI scans from centre 1 (n=74 ACT; n=19 CS2). The external test cohort consisted of 65 MRI scans from centre 2 (n=45 ACT; n=20 CS2). Bidimensional segmentation was performed on T1-weighted MRI. Radiomic features were extracted. After dimensionality reduction and class balancing in centre 1, a machine-learning classifier (Extra Trees Classifier) was tuned on the training cohort using 10-fold cross-validation and tested on the external test cohort. In centre 2, its performance was compared with an experienced musculoskeletal oncology radiologist using McNemar's test.FindingsAfter tuning on the training cohort (AUC=0.88), the machine-learning classifier had 92% accuracy (60/65, AUC=0.94) in identifying the lesions in the external test cohort. Its accuracies in correctly classifying ACT and CS2 were 98% (44/45) and 80% (16/20), respectively. The radiologist had 98% accuracy (64/65) with no difference compared to the classifier (p=0.134).InterpretationMachine learning showed high accuracy in classifying ACT and CS2 of long bones based on MRI radiomic features.FundingESSR Young Researchers Grant.
| S-EPMC8688587 | biostudies-literature
Project description:BackgroundAtypical cartilaginous tumour (ACT) and high-grade chondrosarcoma (CS) of long bones are respectively managed with active surveillance or curettage and wide resection. Our aim was to determine diagnostic performance of X-rays radiomics-based machine learning for classification of ACT and high-grade CS of long bones.MethodsThis retrospective, IRB-approved study included 150 patients with surgically treated and histology-proven lesions at two tertiary bone sarcoma centres. At centre 1, the dataset was split into training (n = 71 ACT, n = 24 high-grade CS) and internal test (n = 19 ACT, n = 6 high-grade CS) cohorts, respectively, based on the date of surgery. At centre 2, the dataset constituted the external test cohort (n = 12 ACT, n = 18 high-grade CS). Manual segmentation was performed on frontal view X-rays, using MRI or CT for preliminary identification of lesion margins. After image pre-processing, radiomic features were extracted. Dimensionality reduction included stability, coefficient of variation, and mutual information analyses. In the training cohort, after class balancing, a machine learning classifier (Support Vector Machine) was automatically tuned using nested 10-fold cross-validation. Then, it was tested on both the test cohorts and compared to two musculoskeletal radiologists' performance using McNemar's test.FindingsFive radiomic features (3 morphology, 2 texture) passed dimensionality reduction. After tuning on the training cohort (AUC = 0.75), the classifier had 80%, 83%, 79% and 80%, 89%, 67% accuracy, sensitivity, and specificity in the internal (temporally independent) and external (geographically independent) test cohorts, respectively, with no difference compared to the radiologists (p ≥ 0.617).InterpretationX-rays radiomics-based machine learning accurately differentiates between ACT and high-grade CS of long bones.FundingAIRC Investigator Grant.
| S-EPMC10884340 | biostudies-literature
Project description:To determine if apparent diffusion coefficients (ADC) can discriminate between posterior fossa brain tumours on a multicentre basis. A total of 124 paediatric patients with posterior fossa tumours (including 55 Medulloblastomas, 36 Pilocytic Astrocytomas and 26 Ependymomas) were scanned using diffusion weighted imaging across 12 different hospitals using a total of 18 different scanners. Apparent diffusion coefficient maps were produced and histogram data was extracted from tumour regions of interest. Total histograms and histogram metrics (mean, variance, skew, kurtosis and 10th, 20th and 50th quantiles) were used as data input for classifiers with accuracy determined by tenfold cross validation. Mean ADC values from the tumour regions of interest differed between tumour types, (ANOVA P < 0.001). A cut off value for mean ADC between Ependymomas and Medulloblastomas was found to be of 0.984 × 10-3 mm2 s-1 with sensitivity 80.8% and specificity 80.0%. Overall classification for the ADC histogram metrics were 85% using Naïve Bayes and 84% for Random Forest classifiers. The most commonly occurring posterior fossa paediatric brain tumours can be classified using Apparent Diffusion Coefficient histogram values to a high accuracy on a multicentre basis.
| S-EPMC7862387 | biostudies-literature
Project description:We explored whether radiomic features from T1 maps by cardiac magnetic resonance (CMR) could enhance the diagnostic value of T1 mapping in distinguishing health from disease and classifying cardiac disease phenotypes. A total of 149 patients (n = 30 with no heart disease, n = 30 with LVH, n = 61 with hypertrophic cardiomyopathy (HCM) and n = 28 with cardiac amyloidosis) undergoing a CMR scan were included in this study. We extracted a total of 850 radiomic features and explored their value in disease classification. We applied principal component analysis and unsupervised clustering in exploratory analysis, and then machine learning for feature selection of the best radiomic features that maximized the diagnostic value for cardiac disease classification. The first three principal components of the T1 radiomics were distinctively correlated with cardiac disease type. Unsupervised hierarchical clustering of the population by myocardial T1 radiomics was significantly associated with myocardial disease type (chi2 = 55.98, p < 0.0001). After feature selection, internal validation and external testing, a model of T1 radiomics had good diagnostic performance (AUC 0.753) for multinomial classification of disease phenotype (normal vs. LVH vs. HCM vs. cardiac amyloid). A subset of six radiomic features outperformed mean native T1 values for classification between myocardial health vs. disease and HCM phenocopies (AUC of T1 vs. radiomics model, for normal: 0.549 vs. 0.888; for LVH: 0.645 vs. 0.790; for HCM 0.541 vs. 0.638; and for cardiac amyloid 0.769 vs. 0.840). We show that myocardial texture assessed by native T1 maps is linked to features of cardiac disease. Myocardial radiomic phenotyping could enhance the diagnostic yield of T1 mapping for myocardial disease detection and classification.
| S-EPMC8654857 | biostudies-literature
Project description:BackgroundTo develop and validate an 18F-fluorodeoxyglucose (FDG) positron emission tomography/computed tomography (PET/CT) and clinico-biological features-based nomogram for distinguishing solid benign pulmonary nodules (BPNs) from malignant pulmonary nodules (MPNs).MethodsA total of 280 patients with BPN (n=128) or MPN (n=152) were collected retrospectively and randomized into the training set (n=196) and validation set (n=84). Pretherapeutic clinicobiological markers, PET/CT metabolic features and radiomic features were analyzed and selected to develop prediction models by the machine-learning method [Least Absolute Shrinkage and Selection Operator (LASSO) regression]. These prediction models were validated using the area under the curve (AUC) of the receiver-operator characteristic (ROC) analysis and decision curve analysis (DCA). Then, the factors of the model with the optimal predictive efficiency were used to constructed a nomogram to provide a visually quantitative tool for distinguishing BPN from MPN patients.ResultsWe developed 3 independent models (Clinical Model, Radiomics Model and Combined Model) to distinguish patients with BPN from those with MPN in the training set. The Combined Model was validated to hold the optimal efficiency and clinical utility with the lowest false positive rate (FPR) in classifying the solid pulmonary nodules in two sets (AUCs of 0.91 and 0.94, FPRs of 18.68% and 5.41%, respectively; P<0.05). Thus, the quantitative nomogram was developed based on the Combined Model, and a good consistency between the predictions and the actual observations was validated by the calibration curves.ConclusionsThis study presents a machine-learning nomogram integrated clinico-biologico-radiological features that can improve the efficiency and reduce the FPR in the noninvasive differentiation of BPN from MPN.
| S-EPMC9816842 | biostudies-literature
Project description:BackgroundTo develop a machine learning-based CT radiomics model is critical for the accurate diagnosis of the rapid spreading coronavirus disease 2019 (COVID-19).MethodsIn this retrospective study, a total of 326 chest CT exams from 134 patients (63 confirmed COVID-19 patients and 71 non-COVID-19 patients) were collected from January 20 to February 8, 2020. A semi-automatic segmentation procedure was used to delineate the volume of interest (VOI), and radiomic features were extracted. The Support Vector Machine (SVM) model was built on the combination of 4 groups of features, including radiomic features, traditional radiological features, quantifying features, and clinical features. By repeating cross-validation procedure, the performance on the time-independent testing cohort was evaluated by the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity.ResultsFor the SVM model built on the combination of 4 groups of features (integrated model), the per-exam AUC was 0.925 (95% CI 0.856 to 0.994) for differentiating COVID-19 on the testing cohort, and the sensitivity and specificity were 0.816 (95% CI 0.651 to 0.917) and 0.923 (95% CI 0.621 to 0.996), respectively. As for the SVM models built on radiomic features, radiological features, quantifying features, and clinical features, individually, the AUC on the testing cohort reached 0.765, 0.818, 0.607, and 0.739, respectively, significantly lower than the integrated model, except for the radiomic model.ConclusionThe machine learning-based CT radiomics models may accurately classify COVID-19, helping clinicians and radiologists to identify COVID-19 positive cases.
| S-EPMC8424152 | biostudies-literature
Project description:Object To evaluate the difference between multiple primary lung adenocarcinoma (MPLA) and solitary primary lung adenocarcinoma (SPLA) by delta-radiomics based machine learning algorithms in CT images. Methods A total of 1094 patients containing 268 MPLAs and 826 SPLAs were recruited for this retrospective study between 2014 to 2020. After the segmentation of volume of interest, the radiomic features were automatically calculated. The patients were categorized into the training set and testing set by a random proportion of 7:3. After feature selection, the relevant classifiers were constructed by the machine learning algorithms of Bayes, forest, k-nearest neighbor, logistic regression, support vector machine, and decision tree. The relative standard deviation (RSD) was calculated and the classification model with minimal RSD was chosen for delta-radiomics analysis to explore the variation of tumor during follow-up surveillance in the cohort of 225 MPLAs and 320 SPLAs. According to the different follow-up duration, it was divided into group A (3–12 months), group B (13–24 months), and group C (25–48 months). Then the corresponding delta-radiomics classifiers were developed to predict MPLAs. The area under the receiver operator characteristic curve (AUC) with 95% confidence interval (CI) was quantified to evaluate the efficiency of the model. Results To radiomics analysis, the forest classifier (FC-radio) with the minimal RSD showed the better stability with AUCs of 0.840 (95%CI, 0.810–0.867) and 0.670 (95%CI, 0.611–0.724) in the training and testing set. The AUCs of the forest classifier based on delta-radiomics (FC-delta) were higher than those of FC-radio. In addition, with the extension of follow-up duration, the performance of FC-delta in Group C were the best with AUCs of 0.998 (95%CI, 0.993–1.000) in the training set and 0.853 (95%CI, 0.752–0.940) in the testing set. Conclusions The machine-learning approach based on radiomics and delta-radiomics helped to differentiate SPLAs from MPLAs. The FC-delta with a longer follow-up duration could better distinguish between SPLAs and MPLAs. Supplementary Information The online version contains supplementary material available at 10.1186/s12885-022-10036-1.
| S-EPMC9440600 | biostudies-literature
Project description:Purpose To identify precise three-dimensional radiomics features in CT images that enable computation of stable and biologically meaningful habitats with machine learning for cancer heterogeneity assessment. Materials and Methods This retrospective study included 2436 liver or lung lesions from 605 CT scans (November 2010-December 2021) in 331 patients with cancer (mean age, 64.5 years ± 10.1 [SD]; 185 male patients). Three-dimensional radiomics were computed from original and perturbed (simulated retest) images with different combinations of feature computation kernel radius and bin size. The lower 95% confidence limit (LCL) of the intraclass correlation coefficient (ICC) was used to measure repeatability and reproducibility. Precise features were identified by combining repeatability and reproducibility results (LCL of ICC ≥ 0.50). Habitats were obtained with Gaussian mixture models in original and perturbed data using precise radiomics features and compared with habitats obtained using all features. The Dice similarity coefficient (DSC) was used to assess habitat stability. Biologic correlates of CT habitats were explored in a case study, with a cohort of 13 patients with CT, multiparametric MRI, and tumor biopsies. Results Three-dimensional radiomics showed poor repeatability (LCL of ICC: median [IQR], 0.442 [0.312-0.516]) and poor reproducibility against kernel radius (LCL of ICC: median [IQR], 0.440 [0.33-0.526]) but excellent reproducibility against bin size (LCL of ICC: median [IQR], 0.929 [0.853-0.988]). Twenty-six radiomics features were precise, differing in lung and liver lesions. Habitats obtained with precise features (DSC: median [IQR], 0.601 [0.494-0.712] and 0.651 [0.52-0.784] for lung and liver lesions, respectively) were more stable than those obtained with all features (DSC: median [IQR], 0.532 [0.424-0.637] and 0.587 [0.465-0.703] for lung and liver lesions, respectively; P < .001). In the case study, CT habitats correlated quantitatively and qualitatively with heterogeneity observed in multiparametric MRI habitats and histology. Conclusion Precise three-dimensional radiomics features were identified on CT images that enabled tumor heterogeneity assessment through stable tumor habitat computation. Keywords: CT, Diffusion-weighted Imaging, Dynamic Contrast-enhanced MRI, MRI, Radiomics, Unsupervised Learning, Oncology, Liver, Lung Supplemental material is available for this article. © RSNA, 2024 See also the commentary by Sagreiya in this issue.
| S-EPMC10982821 | biostudies-literature
Project description:PurposeMYCN amplification plays a critical role in defining high-risk subgroup of patients with neuroblastoma. We aimed to develop and validate the CT-based machine learning models for predicting MYCN amplification in pediatric abdominal neuroblastoma.MethodsA total of 172 patients with MYCN amplified (n = 47) and non-amplified (n = 125) were enrolled. The cohort was randomly stratified sampling into training and testing groups. Clinicopathological parameters and radiographic features were selected to construct the clinical predictive model. The regions of interest (ROIs) were segmented on three-phrase CT images to extract first-, second- and higher-order radiomics features. The ICCs, mRMR and LASSO methods were used for dimensionality reduction. The selected features from the training group were used to establish radiomics models using Logistic regression, Support Vector Machine (SVM), Bayes and Random Forest methods. The performance of four different radiomics models was evaluated according to the area under the receiver operator characteristic (ROC) curve (AUC), and then compared by Delong test. The nomogram incorporated of clinicopathological parameters, radiographic features and radiomics signature was developed through multivariate logistic regression. Finally, the predictive performance of the clinical model, radiomics models, and nomogram was evaluated in both training and testing groups.ResultsIn total, 1,218 radiomics features were extracted from the ROIs on three-phrase CT images, and then 14 optimal features, including one original first-order feature and eight wavelet-transformed features and five LoG-transformed features, were identified and selected to construct the radiomics models. In the training group, the AUC of the Logistic, SVM, Bayes and Random Forest model was 0.940, 0.940, 0.780 and 0.927, respectively, and the corresponding AUC in the testing group was 0.909, 0.909, 0.729, 0.851, respectively. There was no significant difference among the Logistic, SVM and Random Forest model, but all better than the Bayes model (p <0.005). The predictive performance of the Logistic radiomics model based on three-phrase is similar to nomogram, but both better than the clinical model and radiomics model based on single venous phase.ConclusionThe CT-based radiomics signature is able to predict MYCN amplification of pediatric abdominal NB with high accuracy based on SVM, Logistic and Random Forest classifiers, while Bayes classifier yields lower predictive performance. When combined with clinical and radiographic qualitative features, the clinics-radiomics nomogram can improve the performance of predicting MYCN amplification.
| S-EPMC8181422 | biostudies-literature
Project description:ObstructiveTo develop and validate radiomics and machine learning models for identifying encrusted stents and compare their recognition performance with multiple metrics.MethodsA total of 354 patients with ureteral stent placement were enrolled from two medical institutions and divided into the training cohort (n = 189), internal validation cohort (n = 81) and external validation cohort (n = 84). Based on features selected by Wilcoxon test, Spearman Correlation Analysis and least absolute shrinkage and selection operator (LASSO) regression algorithm, six machine learning models based on radiomics features were established with six classifiers (LR, DT, SVM, RF, XGBoost, KNN). After comparison with those models, the most robust model was selected. Considering its feature importance as radscore, the combined model and a nomogram were constructed by incorporating indwelling time. Accuracy, sensitivity, specificity, area under the curve (AUC), decision curve analysis (DCA) and calibration curve were used to evaluate the recognition performance of models.Results1,409 radiomics features were extracted from 641 volumes of interest (VOIs) and 20 significant radiomics features were selected. Considering the superior performance (AUC 0.810, 95%CI, 0.722-0.888) in the external validation cohort, feature importance of XGBoost was used as a radscore, constructing a combined model and a nomogram with indwelling time. The accuracy, sensitivity, specificity and AUC of the combined model were 98, 100, 97.3% and 0.999 for the training cohort, 83.3, 80, 84.5% and 0.867 for the internal cohort and 78.2, 76.3, 78.8% and 0.820 for the external cohort, respectively. DCA indicates the favorable clinical utility of models.ConclusionMachine learning model based on radiomics features enables to identify ureteral stent encrustation with high accuracy.
| S-EPMC10433756 | biostudies-literature