MRI motion artifacts cost about $115,000 per scanner, a year.
The reporting-side twin of that waste isn't measured the same way.

What the new study measured

Researchers from NYU Grossman School of Medicine, working with Siemens Healthineers, pulled sequence-level analytics from scanner log files across 85,300 MRI exams over a six-month period and published the results in the Journal of the American College of Radiology, as reported by Radiology Business and The Imaging Wire.

About 4.8% of all exams — roughly 4,000 — included at least one sequence repeated because of patient motion. Motion-driven repeats consumed about 115 additional scanner hours over the six months, which the study equates to roughly 268 lost MRI appointment slots — capacity that never went to a new patient.

The repeat rate wasn't uniform across exam types. Pediatric imaging had the highest rate, at 9.9% — unsurprising, given how hard it is to keep a young child still for several minutes — while breast MRI had the lowest, at 3.5%.

A repeat doesn't always fix the problem

The study went a step further than counting repeats — it asked whether they were worth it. In a brain MRI sub-analysis covering 77 patients, repeats generally did improve image quality and diagnostic confidence, in about 80% of cases. Read the flip side of that number: roughly one in five repeat sequences didn't improve anything, meaning the scanner time was spent twice for no clinical gain.

A smaller MRCP sub-analysis, covering 27 patients, found repeats were less successful still — quality and diagnostic confidence did not significantly improve overall. Some of the "fix" for motion waste is itself waste.

The number behind the number: $115,000 a year, per scanner

The dollar figure attached to MRI motion waste didn't come from the new study — it comes from an older, separate analysis that keeps getting cited because nothing has replaced it. In 2015, University of Washington researchers reviewed 192 clinical MR exams from a single week and modeled the cost of motion artifacts at roughly $592 per hour of lost scanner revenue — equivalent to about $115,000 per scanner per year (sensitivity range about $92,600–$139,000), a figure Radiology Business covered under the headline "$115,000 problem" and one later repeat-sequence research has continued to reference.

That 2015 study found significant motion artifacts on sequences in 7.5% of outpatient exams versus 29.4% of inpatient and emergency-department exams — and that roughly 20% of all exams needed at least one repeated sequence. The new 85,300-exam study's 4.8% repeat rate is markedly lower, which the authors attribute in part to newer acquisition technology — shorter protocols, parallel imaging, and better reconstruction — that makes scans less vulnerable to a fidgeting patient in the first place. Improvement, but not elimination: motion waste is smaller than a decade ago, not gone.

Acquisition-side waste vs. reporting-side waste

Motion repeats are easy to quantify because the unit of waste is obvious: a sequence either has to run again or it doesn't, and that shows up as minutes on a specific scanner. Reporting-side waste doesn't have that clean unit. A study that sits in a backlog, gets flagged for a second read, comes back for an addendum, or simply waits its turn on an overloaded worklist doesn't register as "115 hours on Scanner 3" — it registers as turnaround time and radiologist hours, spread across a department instead of a single machine. That makes it harder to isolate. It doesn't make it smaller.

Acquisition-side figures per the studies cited above. Reporting-side cells describe the state of measurement, not a competing dollar figure — no single study has priced both sides of the pipeline at once.

Fixing the scanner doesn't shrink the queue

Fixing acquisition-side waste is worth doing on its own terms — fewer repeats means less time in the bore and more available slots. But it doesn't reduce the total amount of reporting work; it just gets more exams to that stage faster. xAID has made a version of this point before, in the context of acquisition-speed AI: a 37-hospital system that cut MRI wait times by more than 60% with faster scanning didn't reduce the number of studies that eventually needed a radiologist's read — it moved the bottleneck downstream. Better motion handling would do the same: saved scanner hours become more completed exams, all of which still have to be read.

Capacity planning that stops at the scanner — much like equipment-side disruptions to imaging capacity more broadly — is only modeling half the pipeline. The other half is the queue of studies waiting to be read, and it deserves the same per-unit rigor motion research has applied to scanner minutes.

Where xAID fits — and where it doesn't

To be direct about scope: xAID's AI CT reporting has nothing to do with the acquisition-side problem this research describes — that's an MRI protocol and hardware question, and the right fixes are motion-robust sequences, better patient prep, and log-file analytics like the study above, not a report-drafting tool. xAID's footprint is the other half of the pipeline: once a CT study is acquired, its AI drafts a structured report, an in-house radiologist reviews every preliminary, and the result is delivered ready-to-sign for the reading radiologist. Fix the acquisition-side waste with better protocols and motion correction; let AI absorb the reporting-side surge that follows. Both halves of the pipeline lose money today — treating only one of them doesn't make the imaging center whole.

Frequently asked questions

Source: NYU Grossman School of Medicine / Siemens Healthineers study, Journal of the American College of Radiology (2026), as reported by Radiology Business and The Imaging Wire. Per-scanner cost estimate from Andre et al., Journal of the American College of Radiology (2015), also covered by Radiology Business. Figures are rounded as reported.