Iterative reconstruction and deep learning algorithms for enabling low-dose computed tomography in midfacial trauma

ObjectivesThe objective of this study was to quantitatively assess the image quality Advanced Modeled Iterative Reconstruction (ADMIRE) and PixelShine (PS) deep learning algorithm for optimization low-dose computed tomography protocols in midfacial trauma.Study DesignSix fresh frozen human cadaver head specimens were scanned by using both standard scan protocols. Three iterative reconstruction strengths applied reconstruct bone soft tissue data sets these subsequently PS algorithm. Signal-to-noise ratios (SNRs) contrast-to-noise (CNRs) calculated each set noise measurements 10 consecutive slices from a standardized region interest template.ResultsThe protocol resulted 61.7% decrease radiation dose. Radiation dose reduction significantly reduced, improved, CNR sets. The algorithms improved after substantial reduction. greatest improvement SNRs CNRs found algorithm.ConclusionBoth ADMIRE trauma. Six template. Both