The receptor activator of the nuclear factor kappa B ligand (RANKL) is the therapeutic target of denosumab. In this study, we evaluated whether radiomics signature and machine learning analysis can predict RANKL status in spinal giant cell tumors of bone (GCTB). This retrospective study consisted of 107 patients, including a training set (n = 82) and a validation set (n = 25). Kaplan-Meier survival analysis was used to validate the prognostic value of RANKL status. Radiomic feature extraction of three heterogeneous regions (VOIentire, VOIedge, and VOIcore) from pretreatment CT were performed. Followed by feature selection using Selected K Best and least absolute shrinkage and selection operator (LASSO) analysis, three classifiers (random forest (RF), support vector machine, and logistic regression) were used to build models. The area under the curve (AUC), accuracy, F1 score, recall, precision, sensitivity, and specificity were used to evaluate the models’ performance. Classification of 75 patients with eligible follow-up based on RANKL status resulted in a significant difference in progression-free survival (p = 0.035). VOIcore-based RF classifier performs best. Using this model, the AUCs for the training and validation cohorts were 0.880 and 0.766, respectively. In conclusion, a machine learning approach based on CT radiomic features could discriminate prognostically significant RANKL status in spinal GCTB, which may ultimately aid clinical decision-making.
BackgroundPatients undergoing surgery for spinal metastasis are predisposed to hidden blood loss (HBL), which is associated with poor surgical outcomes but unpredictable.PurposeTo evaluate the role of MRI‐based radiomics models for assess the risk of HBL in patients undergoing spinal metastasis surgery.Study TypeRetrospective.Subjects202 patients (42.6% female) operated on for spinal metastasis with a mean age of 58 ± 11 years were divided into a training (n = 162) and a validation cohort (n = 40).Field Strength/Sequence1.5T or 3.0T scanners. Sagittal T1‐weighted and fat‐suppressed T2‐weighted imaging sequences.AssessmentHBL was calculated using the Gross formula. Patients were classified as low and high HBL group, with 1000 mL as the threshold. Radiomics models were constructed with radiomics features. The radiomics score (Radscore) was obtained from the optimal radiomics model. Clinical variables were accessed using univariate and multivariate logistic regression analyses. Independent risk variables were used to build a clinical model. Clinical variables combined with Radscore were used to establish a combined model.Statistical TestsPredictive performance was evaluated using area under the curve (AUC), accuracy, sensitivity, specificity, and F1 score. Calibration curves and decision curves analyses were produced to evaluate the accuracy and clinical utility.ResultsAmong the radiomics models, the fusion (T1WI + FS‐T2WI) model demonstrated the highest predictive efficacy (AUC: 0.744, 95% confidence interval [CI]: 0.576–0.914). The Radscore model (AUC: 0.809, 95% CI: 0.664–0.954) performs slightly better than the clinical model (AUC: 0.721, 95% CI: 0.524–0.918; P = 0.418) and the combined model (AUC: 0.752, 95% CI: 0.593–0.911; P = 0.178).Data ConclusionA radiomics model may serve as a promising assessment tool for the risk of HBL in patients undergoing spinal metastasis surgery, and guide perioperative planning to improve surgical outcomes.Level of Evidence3Technical EfficacyStage 2
Background: The aim of this study was to compare the ability of a standard magnetic resonance imaging (MRI)-based radiomics model and a semantic features logistic regression model in differentiating between predominantly osteolytic and osteoblastic spinal metastases.
Methods:We retrospectively analyzed standard MRIs and computed tomography (CT) images of 78 lesions of spinal metastases, of which 52 and 26 were predominantly osteolytic and osteoblastic, respectively. CT images were used as references for determining the sensitivity and specificity of standard MRI. Five standard MRI semantic features of each lesion were evaluated and used for constructing a logistic regression model to differentiate between predominantly osteolytic and osteoblastic metastases. For each lesion, 107 radiomics features were extracted. Six features were selected using a support vector machine (SVM) and were used for constructing classification models. Model performance was measured by means of the area under the curve (AUC) approach and compared using receiver operating characteristics (ROC) curve analysis.
Results:The signal intensity on T1-weighted (T1W), T2-weighted (T2W), and fat-suppressed T2-weighted (FS-T2W) MRI sequences were significantly different between predominantly osteolytic and osteoblastic spinal metastases (P<0.001), as is the case with the existence of soft-tissue masses. The overall prediction accuracy of the models based on radiomics and semantic features was 78.2% and 75.6%, respectively, with corresponding AUCs of 0.82 and 0.79, respectively.
Conclusions:The standard MRI-based radiomics model outperformed the semantic features logistic regression model with regard to differentiating predominantly osteolytic and osteoblastic spinal metastases.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.