A mirror-Unet architecture for PET/CT lesion segmentation
This work addresses lesion segmentation in medical imaging for oncology, presenting an incremental improvement by combining PET and CT information in a hybrid architecture.
The authors tackled the challenging task of automatic lesion segmentation from PET/CT scans by proposing a deep learning method based on two UNet-3D branches, achieving results validated on the AutoPET MICCAI 2023 Challenge dataset.
Automatic lesion detection and segmentation from [${}^{18}$F]FDG PET/CT scans is a challenging task, due to the diversity of shapes, sizes, FDG uptake and location they may present, besides the fact that physiological uptake is also present on healthy tissues. In this work, we propose a deep learning method aimed at the segmentation of oncologic lesions, based on a combination of two UNet-3D branches. First, one of the network's branches is trained to segment a group of tissues from CT images. The other branch is trained to segment the lesions from PET images, combining on the bottleneck the embedded information of CT branch, already trained. We trained and validated our networks on the AutoPET MICCAI 2023 Challenge dataset. Our code is available at: https://github.com/yrotstein/AutoPET2023_Mv1.