Claims that AI saves radiologists time are usually made from reader studies rather than from departments. This retrospective study measured it in practice, comparing 19,190 chest CT examinations before implementation of a commercial pulmonary nodule AI with 20,133 after, across 19,433 patients, using a survival model adjusted for reader, examination type, patient location and requesting specialty, clustered by radiologist.
Reporting was faster after implementation (adjusted hazard ratio 1.17, 95% CI 1.14-1.21), with adjusted median reporting time falling from 21.3 to 18.2 minutes, a 14.6% reduction. Exploratory modelling suggested that at this institution's volume of 20,000-22,000 chest CTs a year, this equated to roughly 0.5 of a full-time radiologist.
The heterogeneity is the more interesting finding. ECG-gated thoracic CT fell 41.1% and thoracic radiologists gained 25.0% — but emergency department examinations took 7.1% longer. The likely explanation is that in a setting where nodules are incidental and the clinical question is something else entirely, AI flags findings that must be reviewed and dismissed, adding work rather than removing it. That is a general lesson about deploying these tools: benefit concentrates where the AI's task matches the reader's task, and reverses where it does not.
- Adjusted median reporting time 21.3 to 18.2 minutes, a 14.6% reduction
- ECG-gated thoracic CT down 41.1%; thoracic radiologists 25.0% faster
- Emergency department examinations took 7.1% longer
- Roughly 0.5 full-time-equivalent radiologist saved at 20,000 scans a year
The statistics, in plain English
This is a before-and-after comparison rather than a randomised one, so anything else that changed between the two periods — staffing, workflow, case mix, familiarity with the reporting system — is mixed into the result. The adjustment handles measured differences, not unmeasured ones. The subgroup pattern is more persuasive than the headline number precisely because it is hard to explain by general drift: a secular trend would not slow emergency reporting while accelerating gated CT.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for radiology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free