Differentiation of pulmonary mucosa-associated lymphoid tissue lymphoma and pulmonary adenocarcinoma by radiomics / 中华放射学杂志

Objective To differentiate between pulmonary mucosa-associated lymphoid tissue lymphoma (MALT) and adenocarcinoma by radiomics, and then evaluate the diagnostic value of this novel approach. Methods We retrospectively analyzed CT images of pulmonary MALT lymphoma (n=16) and invasive pulmonary adenocarcinoma (n=41) and all these cases were confirmed by pathology in the Second Affiliated Hospital of Zhejiang University School of Medicine from June 2012 to June 2017. After we delineated the lesions as region of interest (ROI), sixty-one radiomics features were extracted from each individual's CT images by Radcloud 1.0. All cases in each group were randomly divided into training set (70％cases) and testing set(30％cases), with 7 features (Wilcoxon test) of which showed group differences and were used to train and validate a support vector machine (SVM) classifier. Results Seven of 61 radiomics features showed differences between the two groups, i.e. 10th percentile, mean, median, minimum, total energy, run length non uniformity, gray level non uniformity. Using these 7 features, the resulted SVM successfully differentiated two diseases. The SVM showed high performance with 90％precision, recall 0.89, F1-score 0.87, ROC 0.75. Conclusions Pulmonary MALT and adenocarcinoma differ in radiomics features and machine learning can utilize these features to differentiate between pulmonary MALT and adenocarcinoma. Combination of radiomics and machine learning is promising in the differential diagnosis of these two diseases.

Role of EGFR as a prognostic factor for survival in head and neck cancer: a meta-analysis.

The prognostic role of epidermal growth factor receptor (EGFR) in head and neck squamous cell carcinoma (HNSCC) remains controversial. The goal of this study was to summarize existing evidence regarding whether EGFR overexpression is a prognostic factor in HNSCC. Relevant studies were identified using Pubmed, Ovid, and Web of Science databases. A meta-analysis was conducted on the prognostic value of EGFR expression for overall survival (OS) and disease-free survival (DFS). Thirty-seven studies were included. Primary analysis indicated that EGFR overexpression was associated with reduced OS (hazard ratio [HR]: 1.694, 95% confidence interval [CI]: 1.432­2.004). DFS, on the other hand, was not associated with EGFR expression after adjusting for publication bias (HR: 1.084, 95% CI: 0.910­1.290). Subgroup analysis gave a statistically significant pooled HR for OS in laryngeal carcinoma (HR: 2.519, 95% CI: 1.615­3.928) and in oropharyngeal carcinoma (HR: 2.078, 95% CI: 1.605­2.690). The pooled HR was statistically significant for DFS with respect to oropharyngeal carcinoma (HR: 1.055, 95% CI: 1.020­1.092), but not laryngeal carcinoma (HR: 1.750, 95% CI: 0.911­3.360). When dividing studies based on the immunohistochemistry (IHC) scoring system, only the group that evaluated EGFR expression according to the intensity and extent of staining showed no between-study heterogeneity for both OS and DFS. Overall, EGFR overexpression was associated with shortened OS, but not DFS. Future studies are needed that stratify patients by specific tumor sites. Furthermore, when estimating protein level by the IHC method, it is advisable to consider both intensity and extent of staining.