Comparison Of The Prognostic Value Of The Traditional Parameters And Radiomic Features Of ¹⁸F-FDG PET For Immunochemotherapy-treated Diffuse Large B-cell Lymphoma

Objective:To explore the prognostic value of the traditional metabolic parameters and radiomic features of ¹⁸F-fluorodeoxyglucose（FDG）positron emission tomography（PET）in predicting the 2-year event-free survival（2y-EFS 0）of patients with immunochemotherapy-treated diffuse large B-cell lymphoma（DLBCL）.Methods:Patients with pathologically confirmed immunochemotherapy-treated DLBCL from June 2012 to August 2018 were retrospectively assessed.Total metabolic tumor volume（TMTV）,the maximum and mean standardized uptake value,and total lesion glycolysis（TLG）of each patient were recorded;a region of interest was delineated on the largest lymphoma lesion;and the radiomic features of ¹⁸F-FDG PET were extracted.Using stratified random sampling,the patients were divided into the training（n=108）and test（n=46）groups.2y-EFS 0 was considered the follow-up index,and the extracted radiomic features were screened and optimized using the least absolute shrinkage and selection operator regression model;next,the radiomic score（Rad-score）was established for predicting the 2y-EFS 0 of immunochemotherapy-treated DLBCL in the training group and its predictive efficacy was evaluated and verified in the test group.The receiver operating characteristic curve（ROC curve）was drown to compare the predictive performance of the traditional metabolic parameters and radiomic features of ¹⁸F-FDG PET.Results:In total,154 patients were included.The median follow-up duration was 25months,and the number of patients with 2y-EFS 0 was 82.Among the traditional metabolic parameters,TMTV has a fair value of predicting the 2y-EFS 0 of patients with immunochemotherapy-treated DLBCL（training group:AUC=0.71,test group:AUC=0.66）,followed by TLG（training group:AUC=0.68,test group:AUC=0.65）.The2y-EFS 0 of the patients with immunochemotherapy-treated DLBCL was predicted by the PET Rad-score based on two radiomic features（DISCRETIZED＿HISTO＿Skewness and NGLDM＿busyness）in the training group with the area under the curve（AUC）of0.87;the sensitivity was 75.0%,whereas the specificity was 90.4%.In the test group,the AUC was 0.84,the sensitivity was 72.0%,and the specificity was 85.7%.The predictive power of the radiomic features is better than that of the traditional metabolic parameters（P<0.05）.Conclusions:Compared with the traditional metabolic parameters（TMTV and TLG）,the radiomic features have a higher predictive value for 2y-EFS 0 in patients with immunochemotherapy-treated DLBCL.