Imalogix, a radiology operations platform that handles roughly one in five CT studies performed in the United States, found in its 2026 data that protocols coming in under the achievable dose, the national median, were 35% more likely to fall short on image quality than is typical for their exam type. A protocol is considered optimized when it delivers a diagnostic quality image at an appropriate dose. Dose optimization requires both numbers, and yet CT has only ever had a national benchmark for dose.
Imalogix measured "typical" against noise reference levels, or NRLs, which the company describes as the image-quality analogue to the dose reference levels already in common use, and the first such levels established for CT. Set the way a DRL is, at the 75th percentile of the national distribution, an NRL gives image noise the one thing dose has had for years and quality has lacked: a concrete number to compare against.
ICRP Publication 135 warned that a low dose by itself says nothing about whether a protocol is optimized, that image quality is a separate question. What it never had was a way to answer that question at national scale, and thus dose remained the primary metric scored, from accreditation to Leapfrog's pediatric CT measure.
Imalogix, a radiology operations platform that handles roughly one in five CT studies performed in the United States, found in its 2026 data that protocols coming in under the achievable dose, the national median, were 35% more likely to fall short on image quality than is typical for their exam type. A protocol is considered optimized when it delivers a diagnostic quality image at an appropriate dose. Dose optimization requires both numbers, and yet CT has only ever had a national benchmark for dose.
Imalogix measured "typical" against noise reference levels, or NRLs, which the company describes as the image-quality analogue to the dose reference levels already in common use, and the first such levels established for CT. Set the way a DRL is, at the 75th percentile of the national distribution, an NRL gives image noise the one thing dose has had for years and quality has lacked: a concrete number to compare against.
ICRP Publication 135 warned that a low dose by itself says nothing about whether a protocol is optimized, that image quality is a separate question. What it never had was a way to answer that question at national scale, and thus dose remained the primary metric scored, from accreditation to Leapfrog's pediatric CT measure.
Read together, dose and noise benchmarks also surface the protocols already performing well. For most exam types, one was present in the data. Imalogix defined a reference protocol as the one achieving the lowest dose and noise among protocols sharing similar acquisition parameters across at least five scanners, then modeled universal adoption. The number of protocols exceeding the NRL fell by more than 40%, while median dose fell by roughly 3%. That small dose change is the substantive result: nearly as many protocols would need their dose raised to bring noise into range as would need it lowered, and a dose benchmark, blind to noise, cannot distinguish the two.
This logic underlies Harmonize, the Imalogix feature that benchmarks dose and image quality in tandem. Its reference levels are built from 10 million acquisitions scanned in 2025, and its noise reference levels rest on the same size-normalized, water-equivalent-diameter methodology as the updated U.S. dose reference levels published this year in Radiology, so that protocols serving large and small patients are judged on equal footing. NCRP Report 172 describes reference levels as dynamic values that should be reassessed as technology and clinical practice advance, and Imalogix will release new benchmarks annually, a cadence the platform’s scale makes possible.
Harmonize scores each protocol's dose and noise against those levels , and resolves each into one of four states: high dose, high noise, both, or neither. Where a protocol falls short, it surfaces the best-performing protocol already running on comparable scanners as a target for the next adjustment.
Optimization has always meant weighing dose against image quality. A noise benchmark makes the second half of that weighing visible alongside the first.
Click to read more.
