Self-Supervised Dynamic CT Perfusion Image Denoising With Deep Neural Networks

Dynamic computed tomography perfusion (CTP) imaging is a promising approach for acute ischemic stroke diagnosis and evaluation. Hemodynamic parametric maps of cerebral parenchyma are calculated from repeated CT scans the first pass iodinated contrast through brain. It necessary to reduce dose CTP routine applications due high radiation exposure scans, where image denoising achieve reliable diagnosis. In this article, we proposed self-supervised deep learning method denoising, which did not require any high-dose reference images training. The network was trained by mapping each frame an estimation its adjacent frames. Because noise in source target independent, could effectively remove noise. Being free training granted easier adaptation different scanning protocols. validated on both simulation public real dataset. achieved improved quality compared conventional methods. On data, also had spatial resolution contrast-to-noise ratio supervised data.