Reply to Letter to the Editor: “Current status and quality of radiomics studies in lymphoma: a systematic review”
by Rong Tian (rongtiannuclear@126.com)
Current status and quality of radiomics studies in lymphoma: a systematic reviewDear Editor,
We thank Dr. Martina Sollini and colleagues for complementing our discussion on the methodologies of radiomic studies for lymphoma [1], and for highlighting the importance of lesion selection and segmentation given the great heterogeneity in the size, shape, and location of lymphoma lesions.
Dr. Martina Sollini and colleagues proved in their recent study that Hodgkin’s lymphoma (HL) lesions were not homogeneous within the patients in terms of radiomics signature [2]. Therefore, a random target lesion selection should not be adopted for radiomics applications and the use of only the largest lesion for analysis is not reliable since information coming from all exiting lesions contributes to patient outcome prediction. To set up a methodological framework for radiomics studies in lymphoma is an important and necessary next step to promote better research in this field.
As a core step in radiomic analysis for lymphoma, the optimal segmentation methodology has been explored by several studies. For instance, Hu H et al. proposed an entropy-based optimisation strategy to detect and segment lymphoma in PET images [3]. Hu X et al. proposed an automatic approach for ENKTL segmentation that was more stable than the traditional deep-learning segmentation [4]. Barrington SF et al. evaluated the best automated segmentation workflow in diffuse large B-cell lymphoma (DLBCL) and found segmentation using (a) the standardized uptake value (SUV), with a cut-off of 2.5, and (b) the majority vote, including voxels detected by ≥ 2 methods, are promising approaches [5]. Although studies with large samples are needed to evaluate the performance of these feature selection and segmentation strategies, we are excited to see the increasing interests on radiomics in lymphoma and the continuous progressions to better methodologies.
Thanks again for the comments.