Deep PET/CT Fusion with Dempster-Shafer Theory for Lymphoma Segmentation

Lymphoma detection and segmentation from whole-body Positron Emission Tomography/Computed Tomography (PET/CT) volumes are crucial for surgical indication radiotherapy. Designing automatic methods capable of effectively exploiting the information PET CT as well resolving their uncertainty remain a challenge. In this paper, we propose an lymphoma model using UNet with evidential PET/CT fusion layer. Single-modality trained separately to get initial maps layer is proposed fuse two pieces evidence Dempster-Shafer theory (DST). Moreover, multi-task loss function proposed: in addition use Dice segmentation, based on concordance between added constrain final segmentation. We evaluate our proposal database polycentric patients treated lymphoma, delineated by experts. Our method accurate results score 0.726, without any user interaction. Quantitative show that superior state-of-the-art methods.