Bayesian Quantification of Contrast-Enhanced Ultrasound Images With Adaptive Inclusion of an Irreversible Component

摘要: Contrast Enhanced Ultrasound (CEUS) is a sensitive imaging technique to assess tissue vascularity and it can be particularly useful in early detection grading of arthritis. In recent study we have shown that Gamma-variate accurately quantify synovial perfusion flexible enough describe many heterogeneous patterns. However, some cases the heterogeneity kinetics such even Gamma model does not properly curve, with high number outliers. this work apply CEUS data single compartment recirculation (SCR) which takes explicitly into account trapping microbubbles contrast agent by adding its integral. The SCR model, originally proposed for dynamic-susceptibility magnetic resonance imaging, solved here at pixel level within Bayesian framework using Variational Bayes (VB). We also include automatic relevant determination (ARD) algorithm automatically infer complexity (SCR vs. model) from data. demonstrate inclusion best describes patterns 50% pixels, other fitted Gamma. Such results highlight necessity use ARD, exclude irreversible component where supported VB ARD returns precise estimates majority (88% total percentage pixels) limited computational time (on average, 3.6 min per subject). Moreover, impact additional has been evaluated differentiation rheumatoid non-rheumatoid patients, means support vector machine classifier backward feature selection. show parameter always present selected set, improves classification.