A mirror-Unet architecture for PET/CT lesion segmentation

TL;DR

This paper introduces a novel deep learning architecture called mirror-UNet, which combines two 3D UNet branches to improve automatic lesion segmentation in PET/CT scans, addressing challenges posed by lesion diversity and physiological uptake.

Contribution

The paper presents a new dual-branch UNet architecture that leverages CT and PET data for enhanced lesion segmentation in oncologic imaging.

Findings

Effective segmentation of lesions demonstrated on AutoPET dataset.

Improved accuracy over existing methods in PET/CT lesion segmentation.

Code implementation is publicly available for reproducibility.

Abstract

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.