Automated carotid artery atherosclerotic inflammation segmentation on PET-MRI: Mitigating partial volume effect

Motivation: Vulnerable atherosclerotic plaque in carotid artery is a significant contributor to cerebral mortality,with carotid wall inflammation being closely associated with plaque progression and rupture. Goal(s): This study aims to develop a deep learning-based approach to improve the segmentation of inflammation uptake in carotid PET images,utilizing MRI anatomy and a novel radiotracer to assess plaque inflammation. Approach: We employed a two-stage neural network with a multiscale Residual(MSR) backbone for segmentation of PET uptake.A simulation pipeline was created using PET-MR images to generate PET data for training. Results: Our approach demonstrated superior accuracy to identify PET uptake regions compared to three other deep learning models. Impact: By improving the assessment of carotid inflammation, this methodology has the potential to inform clinical decisions and interventions, ultimately reducing cardiovascular risk and mortality.