Abstract

INTRODUCTION: 18F-FDG PET/CT whole-body tumor burden in lymphoma is not routinely performed due to the lack of fast quantification methods. Although the semi-automatic method is fast, it still lacks the necessary speed required to quantify tumor burden in daily clinical practice. PURPOSE: To evaluate the performance of the convolutional neural networks (CNN) software to localize neoplastic lesions in whole-body 18F-FDG PET/CT images of pediatric lymphoma patients. METHODS: This retrospective image data set, derived from the data pool under the IAEA (CRP# E12017), included 102 baseline staging 18F-FDG PET/CTs of pediatric lymphoma patients (mean age 11 yrs). Images were quantified to determine the whole-body (wb) tumor burden (wbMTV and wbTLG) using a semi-automatic (SEMI) software and an CNN-based software. Both were displayed as wbMTVSEMI & wbTLGSEMI and wbMTVCNN & TLGCNN. The intraclass correlation coefficient (ICC) was applied to evaluate concordance between the CNN-based software and the SEMI software. RESULTS: Twenty-six patients were excluded from the analyses because the software was unable to perform calculation. In the remaining 76 patients, wbMTVCNN and wbMTVSEMI whole-body tumor burden metrics were highly correlated (ICC=0.993; 95%CI: 0.989 -0.996; p-value<0.0001) as were wbTLGCNN and wbTLGSEMI (ICC=0.999; 95%CI: 0.998-0.999; p-value<0.0001). However, the time spent calculating these metrics was significantly (<0.0001) faster by CNN (mean = 19 seconds; 11 - 50 seconds) compared to the semi-automatic method (mean = 21.6 minutes; 3.2 - 62.1 minutes), especially in patients with advanced disease. CONCLUSION: Determining whole-body tumor burden in pediatric lymphoma patients using CNN is fast and feasible in clinical practice.

View details for DOI 10.2967/jnmt.121.262900

View details for PubMedID 35440476