An Image Quality-informed Framework for CT Characterization

BACKGROUND: Lack of standardization in CT protocol choice contributes to radiation dose variation. PURPOSE: To create a framework to assess radiation doses within broad CT categories defined according to body region and clinical imaging indication and to cluster indications according to the dose required for sufficient image quality. MATERIALS AND METHODS: This was a retrospective study using Digital Imaging and Communications in Medicine metadata. CT examinations in adults from January 1, 2016 to December 31, 2019 from the University of California San Francisco International CT Dose Registry were grouped into 19 categories according to body region and required radiation dose levels. Five body regions had a single dose range (ie, extremities, neck, thoracolumbar spine, combined chest and abdomen, and combined thoracolumbar spine). Five additional regions were subdivided according to dose. Head, chest, cardiac, and abdomen each had low, routine, and high dose categories; combined head and neck had routine and high dose categories. For each category, the median and 75th percentile (ie, diagnostic reference level [DRL]) were determined for dose-length product, and the variation in dose within categories versus across categories was calculated and compared using an analysis of variance. Relative median and DRL (95% CI) doses comparing high dose versus low dose categories were calculated. RESULTS: Among 4.5 million examinations, the median and DRL doses varied approximately 10 times between categories compared with between indications within categories. For head, chest, abdomen, and cardiac (3 266 546 examinations [72%]), the relative median doses were higher in examinations assigned to the high dose categories than in examinations assigned to the low dose categories, suggesting the assignment of indications to the broad categories is valid (head, 3.4-fold higher [95% CI: 3.4, 3.5]; chest, 9.6 [95% CI: 9.3, 10.0]; abdomen, 2.4 [95% CI: 2.4, 2.5]; and cardiac, 18.1 [95% CI: 17.7, 18.6]). Results were similar for DRL doses (all P < .001). CONCLUSION: Broad categories based on image quality requirements are a suitable framework for simplifying radiation dose assessment, according to expected variation between and within categories. © RSNA, 2021 See also the editorial by Mahesh in this issue.