Validation of Convolutional Neural Networks for Fast Determination of Whole-Body Metabolic Tumor Burden in Pediatric Lymphoma

Abstract

F-FDG PET/CT quantification of whole-body tumor burden in lymphoma is not routinely performed because of the lack of fast methods. Although the semiautomatic method is fast, it is not fast enough to quantify tumor burden in daily clinical practice. Our purpose was to evaluate the performance of convolutional neural network (CNN) software in localizing neoplastic lesions in whole-body 18 F-FDG PET/CT images of pediatric lymphoma patients. Methods: The retrospective image dataset, derived from the data pool of the International Atomic Energy Agency (coordinated research project E12017), included 102 baseline staging 18 F-FDG PET/CT studies of pediatric lymphoma patients (mean age, 11 y). The images were quantified to determine the whole-body tumor burden (whole-body metabolic tumor volume [wbMTV] and whole-body total lesion glycolysis [wbTLG]) using semiautomatic software and CNN-based software. Both were displayed as semiautomatic wbMTV and wbTLG and as CNN wbMTV and wbTLG. The intraclass correlation coefficient (ICC) was applied to evaluate concordance between the CNN-based software and the semiautomatic software. Results: Twenty-six patients were excluded from the analysis because the software was unable to perform calculations for them. In the remaining 76 patients, CNN and semiautomatic wbMTV tumor burden metrics correlated strongly (ICC, 0.993; 95% CI, 0.989 2 0.996; P , 0.0001), as did CNN and semiautomatic wbTLG (ICC, 0.999; 95% CI, 0.998-0.999; P , 0.0001). However, the time spent calculating these metrics was significantly (,0.0001) less by CNN (mean, 19 s; range, 11-50 s) than by the semiautomatic method (mean, 21.6 min; range, 3.2-62.1 min), 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.

The transformational potential of molecular radiomics

“…demonstrated application of DL multi‐focal segmentation in reliably determining MTV and TLG in lymphoma and found that it reduced time to 18% of manual approaches. In similar research, 18 F FDG PET MTV and TLG in paediatric lymphoma patients were highly correlated for computer‐assisted versus DL segmentation but showed faster processing time for DL with DL taking just 1.5% of the processing time of computer‐assisted approaches (mean of 19 sec versus 21.6 min) 30 . DL has also been reported to effectively localise lesions and then classify as suspicious or otherwise in lung cancer and lymphoma 31 .…”

“…In similar research, 18 F FDG PET MTV and TLG in paediatric lymphoma patients were highly correlated for computer-assisted versus DL segmentation but showed faster processing time for DL with DL taking just 1.5% of the processing time of computer-assisted approaches (mean of 19 sec versus 21.6 min). 30 DL has also been reported to effectively localise lesions and then classify as suspicious or otherwise in lung cancer and lymphoma. 31 DL has also been used for auto-segmentation of metastatic deposits on bone SPECT studies to improve staging and determination of tumour burden.…”

The Impact of PET/CT on Paediatric Oncology

“…There is growing evidence for the utility of using machine learning methods for semiquantification, such as convolutional neural networks (CNN). Recent work by Etchebehere et al demonstrated clearly that the determination of whole body tumour burden in patients with paediatric lymphoma using CNN is fast and feasible in implementation in clinical practice [19].…”