A Proposal for a Process from as Low as Reasonably Achievable to an Ultra-Low-Level Goal in Chest Computed Tomography.

To define and evaluate a radiation dose optimization process for chest computed tomography (CT) imaging. Data from unenhanced and enhanced chest CT acquisitions performed between June 2018 and January 2020 in adult patients were included in the study. Images were acquired on a Siemens SOMATOM Definition Edge CT. Dose values, including Dose.Length Product (DLP) and Volume CT Dose Index (CTDI), were collected. Low doses (LDs, 25th percentiles), achievable doses (ADs, 50th percentiles), and diagnostic reference levels (DRLs, 75th percentiles) were calculated before and after parameter modifications. A process was defined and applied to patient data. For unenhanced chest CT, data were differentiated according to three groups: high dose (HD), optimized dose (OD), and ultra-low dose (ULD). Dosimetric changes between protocols were expressed as mean CTDI % (CI95%). A Mann and Whitney statistical test was used. The diagnostic quality score (DQS) of a subset of 70 randomly selected CT examinations was evaluated by one radiologist. The DQS was scored according to a three-point Likert scale: (1) poor (definite diagnosis impossible), (2) fair (evaluation of major findings possible), and (3) excellent (exact diagnosis possible). : Data were collected from 1929 patients. For unenhanced chest CT protocols, only one process loop was run. A dose comparison between the chest CT protocol before the use of the process and the three groups showed a decrease of -38.3% (9.7%) and -93.4% (24.2%) for OD and ULD, respectively, and an increase of +29.4% (4.7%) for HD. For the enhanced chest CT protocol, two optimization loops were performed, and they resulted in a mean dose reduction of -50.0% (2.6%) compared to the pre-optimization protocol. For all protocols, the DQS was greater than or equal to 2. : We proposed a radiation dose optimization process for chest CT that could significantly reduce the dose without compromising diagnosis.