Objective: To investigate whether machine learning analysis of multiparametric MR radiomics can help classify immunohistochemical (IHC) subtypes of breast cancer. Study design: One hundred and thirty-four consecutive patients with pathologically-proven invasive ductal carcinoma were retrospectively analyzed. A total of 2,498 features were extracted from the DCE and DWI images, together with the new calculated images, including DCE images changing over six time points (DCEsequential) and DWI images changing over three b-values (DWIsequential). We proposed a novel two-stage feature selection method combining traditional statistics and machine learning-based methods. The accuracies of the 4-IHC classification and triple negative (TN) vs. non-TN cancers were assessed. Results: For the 4-IHC classification task, the best accuracy of 72.4% was achieved based on linear discriminant analysis (LDA) or subspace discrimination of assembled learning in conjunction with 20 selected features, and only small dependent emphasis of Kendall-tau-b for sequential features, based on the DWIsequential with the LDA model, yielding an accuracy of 53.7%. The linear support vector machine (SVM) and medium k-nearest neighbor using eight features yielded the highest accuracy of 91.0% for comparing TN to non-TN cancers, and the maximum variance for DWIsequential alone, together with a linear SVM model, achieved an accuracy of 83.6%. Conclusions: Whole-tumor radiomics on MR multiparametric images, DCE images changing over time points, and DWI images changing over different b-values provide a non-invasive analytical approach for breast cancer subtype classification and TN cancer identification.
Background: Brain metastasis (BM) is most common in non-small cell lung cancer (NSCLC) patients. This study aims to enhance BM risk prediction within three years for advanced NSCLC patients by using a deep learning-based segmentation and computed tomography (CT) radiomics-based ensemble learning model. Methods: This retrospective study included 602 stage IIIA-IVB NSCLC patients, 309 BM patients and 293 non-BM patients, from two centers. Patients were divided into a training cohort (N = 376), an internal validation cohort (N = 161) and an external validation cohort (N = 65). Lung tumors were first segmented by using a three-dimensional (3D) deep residual U-Net network. Then, a total of 1106 radiomics features were computed by using pretreatment lung CT images to decode the imaging phenotypes of primary lung cancer. To reduce the dimensionality of the radiomics features, recursive feature elimination configured with the least absolute shrinkage and selection operator (LASSO) regularization method was applied to select the optimal image features after removing the low-variance features. An ensemble learning algorithm of the extreme gradient boosting (XGBoost) classifier was used to train and build a prediction model by fusing radiomics features and clinical features. Finally, Kaplan‒Meier (KM) survival analysis was used to evaluate the prognostic value of the prediction score generated by the radiomics–clinical model. Results: The fused model achieved area under the receiver operating characteristic curve values of 0.91 ± 0.01, 0.89 ± 0.02 and 0.85 ± 0.05 on the training and two validation cohorts, respectively. Through KM survival analysis, the risk score generated by our model achieved a significant prognostic value for BM-free survival (BMFS) and overall survival (OS) in the two cohorts (P < 0.05). Conclusions: Our results demonstrated that (1) the fusion of radiomics and clinical features can improve the prediction performance in predicting BM risk, (2) the radiomics model generates higher performance than the clinical model, and (3) the radiomics-clinical fusion model has prognostic value in predicting the BMFS and OS of NSCLC patients.
Background: To develop and validate a peritumoral vascular and intratumoral radiomics model to improve pretreatment predictions for pathologic complete responses (pCRs) to neoadjuvant chemoradiotherapy (NAC) in patients with triple-negative breast cancer (TNBC). Methods: A total of 282 TNBC patients (93 in the primary cohort, 113 in the validation cohort, and 76 in The Cancer Imaging Archive [TCIA] cohort) were retrospectively included. The peritumoral vasculature on the maximum intensity projection (MIP) from pretreatment DCE-MRI was segmented by a Hessian matrix-based filter and then edited by a radiologist. Radiomics features were extracted from the tumor and peritumoral vasculature of the MIP images. The LASSO method was used for feature selection, and the k-nearest neighbor (k-NN) classifier was trained and validated to build a predictive model. The diagnostic performance was assessed using the ROC analysis. Results: One hundred of the 282 patient (35.5%) with TNBC achieved pCRs after NAC. In predicting pCRs, the combined peritumoral vascular and intratumoral model (fusion model) yields a maximum AUC of 0.82 (95% confidence interval [CI]: 0.75, 0.88) in the primary cohort, a maximum AUC of 0.67 (95% CI: 0.57, 0.76) in the internal validation cohort, and a maximum AUC of 0.65 (95% CI: 0.52, 0.78) in TCIA cohort. The fusion model showed improved performance over the intratumoral model and the peritumoral vascular model, but not significantly (p > 0.05). Conclusion: This study suggested that combined peritumoral vascular and intratumoral radiomics model could provide a non-invasive tool to enable prediction of pCR in TNBC patients treated with NAC.
29 July 2019 in Journal of Computer Assisted Tomography
Objective: To identify the computed tomographic features to differentiate gastric schwannoma from gastric gastrointestinal stromal tumor. Methods: Computed tomographic images of 103 pathologically confirmed patients with gastric schwannoma (n = 23) or gastric gastrointestinal stromal tumor (n = 80) were retrospectively studied. Sex, morphology, location, border, growth pattern, enhancement pattern, necrosis, calcification, ulceration, and perigastric lymph nodes were analyzed. Age, short diameter, long diameter, and the degree of enhancement were measured. Statistical analyses were performed, and receiver operating characteristic curve, sensitivity, and specificity values were analyzed. Results: Female, age younger than 57.5 years, round or ovoid morphology, extraluminal growth, homogeneous enhancement, lack of necrosis, presence of perigastric lymph nodes, and an enhancement degree of less than 15.4 Hounsfield units in the arterial phase were significant variables for differentiating gastric schwannoma from gastric gastrointestinal stromal tumor (P < 0.05). Conclusions: Dual-phase contrast-enhanced computed tomography can help to differentiate gastric schwannoma from gastric gastrointestinal stromal tumor.