A Spline Model for Rv Registration from Cardiac Pet Images
Abstract
Background: The etiology of pulmonary hypertension (PAH) is poorly understood, and is associated with high morbidity and mortality (life expectancy <3 years). PAH leads to progressive enlargement of the right ventricle (RV) with a rapid decline in function. The utility of non-invasive molecular imaging to track PAH progression, evaluate disease etiology and monitor clinical therapy is currently limited by the lack of automated RV image analysis tools.
Objective: To develop a highly automated RV registration, sampling, and analysis tool for human and small animal positron emission tomography (PET) imaging.
Methods: We developed a spline-based model for registering the mid-RV myocardium from a PET uptake image. Model fitting was automated by optimizing a constrained cost function with optional operator intervention.
Inter- and intra-operator variability of FDG PET uptake activity measurements were evaluated using a dataset consisting of 7 PAH and 12 randomly selected non-PAH human subjects. The dataset was processed twice by each of two operators, a novice and an expert. The accuracy of RV cavity volumes and ejection fraction (EF) measurements from cardiac-gated PET images was evaluated by comparing with results from cardiac magnetic resonance imaging (CMR) in 5 PAH patients.
Results: 50% of cases assessed required operator intervention. In intra-operator variability analysis of relative uptake images, the reproducibility coefficient (RPC) for each operator was 5.6% and 6.4% for expert and novice respectively. Inter-operator uptake RPC was 8.2%. RV cavity volumes and EF agreed closely with CMR results (r2=0.954, n=10 and r2=0.965, n=5 respectively).
Conclusions: The RV can be automatically registered in uptake PET images and has performance characteristics that are comparable with established left ventricle analysis tools making it suitable for investigating RV molecular and cardiac functions. Additional work is required to improve automation and evaluate molecular function quantification with dynamic imaging.
