“…In our study, expert tumor delineation performed as the best (although not perfect) approach to truth. The evaluation of the voxel-wise similarity between the ground truth and the automatically segmented mask demonstrates that the state-of-the-art deep learning architecture nnU-Net can be used to detect and segment neuroblastic tumors on MR images, with a median DSC of 0.965 (±0.018 IQR), achieving a strong performance and surpassing the methods and results obtained in previous studies that approached the problem of neuroblastoma segmentation [ 9 , 10 , 11 ]. However, no previous literature has demonstrated the performance of a CNN-based solution in neuroblastic tumor.…”