Development and validation of a CT-based radiomics model for differentiating pneumonia-like primary pulmonary lymphoma from infectious pneumonia: A multicenter study / 中华医学杂志(英文版)

BACKGROUND@#Pneumonia-like primary pulmonary lymphoma (PPL) was commonly misdiagnosed as infectious pneumonia, leading to delayed treatment. The purpose of this study was to establish a computed tomography (CT)-based radiomics model to differentiate pneumonia-like PPL from infectious pneumonia.@*METHODS@#In this retrospective study, 79 patients with pneumonia-like PPL and 176 patients with infectious pneumonia from 12 medical centers were enrolled. Patients from center 1 to center 7 were assigned to the training or validation cohort, and the remaining patients from other centers were used as the external test cohort. Radiomics features were extracted from CT images. A three-step procedure was applied for radiomics feature selection and radiomics signature building, including the inter- and intra-class correlation coefficients (ICCs), a one-way analysis of variance (ANOVA), and least absolute shrinkage and selection operator (LASSO). Univariate and multivariate analyses were used to identify the significant clinicoradiological variables and construct a clinical factor model. Two radiologists reviewed the CT images for the external test set. Performance of the radiomics model, clinical factor model, and each radiologist were assessed by receiver operating characteristic, and area under the curve (AUC) was compared.@*RESULTS@#A total of 144 patients (44 with pneumonia-like PPL and 100 infectious pneumonia) were in the training cohort, 38 patients (12 with pneumonia-like PPL and 26 infectious pneumonia) were in the validation cohort, and 73 patients (23 with pneumonia-like PPL and 50 infectious pneumonia) were in the external test cohort. Twenty-three radiomics features were selected to build the radiomics model, which yielded AUCs of 0.95 (95% confidence interval [CI]: 0.94-0.99), 0.93 (95% CI: 0.85-0.98), and 0.94 (95% CI: 0.87-0.99) in the training, validation, and external test cohort, respectively. The AUCs for the two readers and clinical factor model were 0.74 (95% CI: 0.63-0.83), 0.72 (95% CI: 0.62-0.82), and 0.73 (95% CI: 0.62-0.84) in the external test cohort, respectively. The radiomics model outperformed both the readers' interpretation and clinical factor model ( P <0.05).@*CONCLUSIONS@#The CT-based radiomics model may provide an effective and non-invasive tool to differentiate pneumonia-like PPL from infectious pneumonia, which might provide assistance for clinicians in tailoring precise therapy.

Preoperative MRI-based deep learning radiomics machine learning model for prediction of the histopathological grade of soft tissue sarcomas / 中华放射学杂志

Objective:To investigate the value of a preoperatively MRI-based deep learning (DL) radiomics machine learning model to distinguish low-grade and high-grade soft tissue sarcomas (STS).Methods:From November 2007 to May 2019, 151 patients with STS confirmed by pathology in the Affiliated Hospital of Qingdao University were enrolled as training sets, and 131 patients in the Affiliated Hospital of Shandong First Medical University and the Third Hospital of Hebei Medical University were enrolled as external validation sets. According to the French Federation Nationale des Centres de Lutte Contre le Cancer classification (FNCLCC) system, 161 patients with FNCLCC grades Ⅰ and Ⅱ were defined as low-grade and 121 patients with grade Ⅲ were defined as high-grade. The hand-crafted radiomic (HCR) and DL radiomic features of the lesions were extracted respectively. Based on HCR features, DL features, and HCR-DL combined features, respectively, three machine-learning models were established by decision tree, logistic regression, and support vector machine (SVM) classifiers. The area under the receiver operating characteristic curve (AUC) was used to evaluate the performance of each machine learning model and choose the best one. The univariate and multivariate logistic regression were used to establish a clinical-imaging factors model based on demographics and MRI findings. The nomogram was established by combining the optimal radiomics model and the clinical-imaging model. The AUC was used to evaluate the performance of each model and the DeLong test was used for comparison of AUC between every two models. The Kaplan-Meier survival curve and log-rank test were used to evaluate the performance of the optimal machine learning model in the risk stratification of progression free survival (PFS) in STS patients.Results:The SVM radiomics model based on HCR-DL combined features had the optimal predicting power with AUC values of 0.931(95%CI 0.889-0.973) in the training set and 0.951 (95%CI 0.904-0.997) in the validation set. The AUC values of the clinical-imaging model were 0.795 (95%CI 0.724-0.867) and 0.615 (95%CI 0.510-0.720), and of the nomogram was 0.875 (95%CI 0.818-0.932) and 0.786 (95%CI 0.701-0.872) in the training and validation sets, respectively. In validation set, the performance of SVM radiomics model was better than those of the nomogram and clinical-imaging models ( Z=3.16, 6.07; P=0.002,<0.001). Using the optimal radiomics model, there was statistically significant in PFS between the high and low risk groups of STS patients (training sets: χ2=43.50, P<0.001; validation sets: χ2=70.50, P<0.001). Conclusion:Preoperative MRI-based DL radiomics machine learning model has accurate prediction performance in differentiating the histopathological grading of STS. The SVM radiomics model based on HCR-DL combined features has the optimal predicting power and was expected to undergo risk stratification of prognosis in STS patients.