Image Quality of Non-Contrast Neurological Images on Energy-Integrating versus Silicon Photon-Counting Detector CT.
Authors
Affiliations (2)
Affiliations (2)
- From the Departments of Medical Physics and Radiology, University of Wisconsin, Madison; Department of Radiology, University of Wisconsin, Madison; Department of Radiology, University of Wisconsin, Madison; Department of Radiology, Harvard Medical School; Department of Radiology, Wake Forest University School of Medicine; GE HealthCare, Waukesha, WI; GE Health Care, Waukesha, WI; GE HealthCare, Waukesha, WI; Departments of Medical Physics & Radiology, University of Wisconsin, Madison, WI; and Departments of Biomedical Engineering, Medical Physics, and Radiology, University of Wisconsin, Madison.
- From the Departments of Medical Physics and Radiology, University of Wisconsin, Madison; Department of Radiology, University of Wisconsin, Madison; Department of Radiology, University of Wisconsin, Madison; Department of Radiology, Harvard Medical School; Department of Radiology, Wake Forest University School of Medicine; GE HealthCare, Waukesha, WI; GE Health Care, Waukesha, WI; GE HealthCare, Waukesha, WI; Departments of Medical Physics & Radiology, University of Wisconsin, Madison, WI; and Departments of Biomedical Engineering, Medical Physics, and Radiology, University of Wisconsin, Madison. [email protected].
Abstract
The purpose of this work was to evaluate image quality of non-contrast neurological CT scans on a commercially available energy-integrating detector (EID) CT compared to a prototype deep silicon (dSi) photon-counting detector CT (PCCT). A cohort of subjects scanned on the EID-CT (Apex CT, GE HealthCare, Waukesha, WI) was retrospectively recruited. A second cohort of prospectively recruited subjects was scanned on the prototype dSi-PCCT CT using deep learning (DL) recon. Image quality was assessed by four neuroradiologists on a 5-point Likert scale for sharpness, noise, and presence of artifacts, differentiation of gray-white matter, and structural confidence in the lentiform nucleus and insular ribbon. Reader agreement was assessed through a linearly weighted Cohen's kappa. Contrast-to-noise ratio (CNR) between gray and white matter was determined. Differences between the median Likert scores and CNRs of the two cohorts were assessed with the Mann-Whitney U test. Reader agreement was fair on average ([Formula: see text]). Median Likert scores of the dSi-PCCT with DL recon cohort were statistically superior compared to the median scores of the EID-CT scores for sharpness, noise, gray-white matter differentiation, and structural confidence of the lentiform nucleus and insular ribbon. No statistical difference was detected in the presence of artifacts between the cohorts. Average CNRs were 1.45±0.14 and 1.53±0.08 for the EID-CT and dSi-PCCT cohorts, respectively, but no significant difference was detected (p = 0.72). Increased sharpness, lower noise, and better ability to differentiate structures demonstrate improved image quality and the potential for utilizing PCCT in non-contrast neurological imaging tasks.