CT Resolution, Noise, and Dose Reference Levels Across a Multi-Center Patient Population
PURPOSE The diagnostic reference level (DRL) is useful as a first-order tool to compare radiation exposures of one's imaging clinic against those of one's peers in computed tomography. Ria et al. have advocated for the addition of a Noise Reference Level (NRL), Noise Reference Range (NRR), a Dose Reference Level (DoRL) and a Dose Reference Range (DoRR) as a means of extending the first steps of optimization of clinical operation to image quality in addition to radiation exposure (ICRP 135). In this work, we investigate and establish another reference level, for the case of resolution, called the Resolution Reference Level (RRL) and Resolution Reference Range (RRR). METHOD AND MATERIALS Over 10000 anonymized scans were sampled from 109 institutions in the United States which represented 13 large (estim. > 36,000 scans/yr), 32 medium (6000 - 36,000), 64 small (< 6000) imaging centers. CTDIvol, Noise and resolution (MTF) were measured for chest and abdominopelvic (AP) helical single-acquisitions of the trunk (extremity/head scans, multiple reconstructions excluded). Noise was taken as a modified version of the Global Noise Level measured in air. MTFs were measured from the air-skin interface. Patients were binned by their effective diameter. DoRL, NRL, and RRLs were measured for each size bin. Reference Levels for each were defined as the median values. Reference Ranges were defined as interquartile intervals. RESULTS Measurements of the f50 MTF in AP exams in all sizes had RRR lower bounds from 0.33 - 0.42 mm-1. The RRR upper bounds for these exams ranged from 0.38 - 0.46 mm-1. In chest exams these ranges were from 0.39 - 0.42 mm-1 and 0.45 - 0.46 mm-1 respectively. The lower bound of DoRR in the AP exams in all sizes ranged from 5.3 - 11.2 mGy. The upper bounds of DoRR ranged from 7.4 - 16.5 mGy. In chest exams these lower and upper bounds of the DoRR ranged from 5.4 - 11.5 mGy and 7.5 - 14.4 mGy. CONCLUSION An extension to the reference level methodology has been proposed. With this new RRL and RRR, and existing NRL, NRR, DoRL, and DoRR, a clinic can begin to make holistic decisions regarding population-based imaging-protocol adjustments with consideration to dose, noise, and resolution. CLINICAL RELEVANCE/APPLICATION This work extends the notion of NRL to the assessment of resolution. It represents a more generalizable sampling, with more vendors and kernels, and an order of magnitude increase in population.
