Characterization of Mediastinal Bulky Lymphomas with FDG-PET-Based Radiomics and Machine Learning Techniques

This manuscript aims to address the diagnostic challenges of mediastinal bulky lymphomas with the baseline value of 18F-FDG PET/CT metabolic, volumetric and texture parame- ters, also relying on machine learning techniques, in patients with grey zone lymphoma, primary diffuse large B-cell lymphoma of the mediastinum and classical Hodgkin lymphoma. Different types of histology demonstrated several baseline 18F-FDG PET/CT radiomics parameters that were significantly different from one another, suggesting the possibility of identifying potential histological heterogeneity and aggressive transformation. Moreover, using radiomics-based imaging biomarkers, machine learning techniques offer a solution for separating not completely disjoint histological types. To date, the gold standard for diagnosis is biopsy, but machine learning methods could be combined with radiomics to build a histological representation of mediastinal bulky masses that is able to suc- cessfully identify different types of lymphomas. Finally, this preliminary study supports the potential of metabolic texture analyses as future imaging biomarkers, with a growing role in clinical diagnosis.