Design of prehospital spectral CT with a rotation-free rectangular gantry and deep learning reconstruction

Abstract

Treatment of stroke depends on a narrow therapeutic time window and requires urgent intervention to be emergently pursued. Despite recent “fast” initiatives that have underscored “time is brain,” many patients still fail to present within the narrow time window to receive maximum treatment benefit from advanced stroke therapies. The convergence of emergency medical services, telemedicine, and mobile technology, including transportable computed tomography scanners, has presented a unique opportunity to advance patient stroke care in the prehospital field by shortening the time to hyperacute stroke treatment with a mobile stroke unit. However, in order to facilitate the integration of computed tomography (CT) scanners into small ambulances, their geometry should be streamlined and made compact. Additionally, it is imperative that radiation exposure from prehospital CT scans comply with established radiation protection guidelines to safeguard both patients and healthcare workers. In this study, we propose a newly designed CT system for a rotation-free rectangular gantry with a photon-counting detector (PCD) and a deep-learning reconstruction approach for undersampled projections. Undersampled projections obtained using the designed CT system were implemented using a deep learning method. The deep learning approach showed superior image quality without distortion compared to state-of-the-art reconstruction methods. Finally, we implemented imaging of brain microhemorrhages using low- and high-energy images obtained from the PCD. Our results indicate that the proposed scanner effectively shows potential for acute stroke detection in prehospital ambulances. Its effectiveness was validated by comparing its image performance with those of other methods such as FBP and compressed sensing.