Optimizing Mobile Photon-Counting CT Image Quality via Deep Learning for Neuro Intensive Care Unit - Project Summary Mobile CT scanners are routinely used in the neuro intensive care unit (ICU) for critically ill patients to avoid morbidity and adverse events associated with patient transport. The image quality of mobile CT is inferior to a fixed MDCT in terms of image noise, spatial resolution, soft tissue contrast, and susceptibility to artifacts from beam hardening, motion, metallic implants, and truncation. Reduced image quality may compromise care and necessitate transport to a fixed scanner. For example, on a mobile CT scanner, it is difficult to diagnose small infarcts or hemorrhage, or to differentiate between intracranial hemorrhage and contrast extravasation after endovascular processes. In this project, we will leverage the benefits from an FDA-approved mobile photon counting CT (PCCT), as well as deep learning-based image reconstruction algorithms to improve the image quality of the mobile PCCT to or beyond that of a fixed scanner. The multi-spectral and high-resolution features of the mobile PCCT will be explored and combined with deep learning algorithms for noise and artifacts reduction. In this project, the following aims will be investigated to achieve our goal to match the image quality of a mobile PCCT to a fixed scanner: (1) We will develop low-dose, high-resolution deep learning-based reconstruction algorithms to reduce the noise and improve gray-white matter contrast in the mobile PCCT; (2) We will develop methods for material decomposition with reduced noise amplification and spectral optimization to overcome beam hardening artifacts and achieve discrimination between calcium/contrast/hemorrhage; (3) We will develop deep learning-based algorithms for image artifacts correction to tackle artifacts that are more frequent on a mobile CT, including motion, truncation, and metal artifacts; (4) To validate the methods, the optimized mobile PCCT images will be compared with fixed CT images by trained radiologists.
