Integrating the radiomics signature the clinical and
Integrating the radiomics signature, the clinical and morphological features, the integrated model was built consisting of maximum diameter, location, sex and radiomics signature. The integrated model showed good performance with higher AUC, better than the radiomics signature and the clinical models in both training and validation cohorts. This finding suggested that the integrated model based on radiomics is more effective to improve the pre-therapy personalized prediction of EGFR mutation. This finding was similar to the previous study by Liu et al. , which selected five radiomics features to construct a radiomics model for prediction of EGFR mutation in lung adenocarcinomas. And they found the predictive ability of clinical model increased when combining with radiomics features. Zhang et al.  also found the combined model was superior to the model generated with clinical features alone and the model with radiomics features alone. Although the inclusion criteria, sample size, feature selection approach or modeling methods of previous studies were different from ours, we have also established radiomics-based EGFR prediction models with good predictive performance. Therefore, it is indicated that EGFR mutation status may be indeed related with the imaging phenotypes of tumors, and it is feasible to predict EGFR mutation status by mining intratumoral heterogeneity from medical images. Most of prior studies didn’t evaluate morphological characteristics comprehensively. Rios et al.  only incorporated two size-based characteristics (maximum diameter and tumor volume) into their radiomics study about predicting EGFR mutation, and found radiomics signature was superior to size-based characteristics. In contrast, we assessed more comprehensive morphological features than those previous studies.
Conflict of interest
Introduction Primary pulmonary lymphoma (PPL) is extremely rare, accounting for only 0.4% of all lymphomas and less than 0.5% of all primary lung tumors [1,2]. PPL is defined as a lymphoma confined to the lung with or without hilar Relebactam node involvement at the time of diagnosis or up to 3 months thereafter, and predominantly affects adults with a median age of about 60 years [1,, , ]. Approximately 30% to 40% of patients are asymptomatic at initial presentation. Other patients may present non-specific clinical symptoms that are difficult to distinguish from other lung cancers, such as cough, dyspnea, hemoptysis and chest pain. Fever, diaphoresis, or weight loss were reported in 15% to 27% of patients with PPL . The radiological findings of PPL are heterogenous and present as solitary mass lesions, pulmonary infiltrates, and solitary/ multiple bilateral nodules [6,7]. The majority of PPL cases are of B-cell origin, among which mucosa-associated lymphoid tissue (MALT) lymphoma encompasses 70% to 80% . The survival of PPL patients is favorable, with 68% 5-year survival rate and 53% 10-year survival rate in MALT lymphoma . Secondary pulmonary lymphoma (SPL) is more frequent and defined as secondary involvement of pulmonary by a systemic lymphoma. The radiological features of SPL may present with solitary or multiple nodules, mass or mass-like consolidation, alveolar or interstitial infiltrates, pleural effusions and mediastinal lymph node enlargement. Diffuse large B-cell lymphoma (DLBCL) is the most common subtype of SPL, followed by mantle cell lymphoma (MCL), MALT, follicular lymphoma (FL), and Burkitt lymphoma (BL). The response and outcome of SPL are significantly worse compared to those without pulmonary involvement . National Comprehensive Cancer Network-International Prognostic Index (NCCN-IPI) considered lung as one of the major extranodal organs .
Patients and methods
Discussion Approximately 25% of non-Hodgkin’s lymphoma arises in extranodal sites, most often the gastrointestinal tract [12,13]. Primary pulmonary involvement is a very rare entity, accounting for only 3% of extranodal lymphoma . The majority of PPL is of B-cell origin, accounting for more than 90% of cases, with MALT lymphoma as the most common subtype. In our study, 67% of PPL were MALT, 27% were DLBCL, 5% were PTCL, and 1% was FL, confirming the epidemiologic data previously reported . Meanwhile, all histological subtypes of lymphoma may also secondarily involve the lung. Here, SPL patients were most commonly affected by DLBCL (48%), followed by PTCL (32%), MALT (6%), FL (5%), BL (3%), SLL (3%), and MCL (2%). As expected, we observed more frequent presence of B symptoms, advanced stage disease, intermediate-high or high risks of IPI and NCCN-IPI, elevated inflammatory parameters (AMC, CRP, β2-MG and ferritin) and cytokine levels (IL-2R, IL-6 and IL-10) in patients with SPL, compared to patients with PPL.