Motivation: We aim to conduct a comprehensive evaluation of a radial cardiac cine acquisition and deep learning reconstruction protocol. Goal(s): Our objective is to demonstrate the effectiveness and generalizability of the deep learning reconstruction for accelerated cine imaging via qualitative and quantitative assessment over a diverse cohort of volunteers and patients. Approach: We deploy a cardiac cine sequence and collect data from a large subject cohort. Collected data are processed with raw k-space preprocessing modules, followed by a deep learning reconstruction based on unrolled neural networks. The reconstruction quality is assessed via peak signal-to-noise ratio, structural similarity index and ejection fraction ratio. Impact: Free-breathing, radial cardiac cine acquisition and reconstruction approaches can mitigate motion artifacts and improve patient comfort and compliance. We perform a comprehensive evaluation of such a protocol to validate its effectiveness and validity on diverse populations including volunteers and patients.
Yurt et al. (Tue,) studied this question.