What the survey found
The Society of Breast Imaging distributed an anonymous, IRB-exempt survey to its 2,264 members in May 2024. Only 162 responded — a 7.2% response rate — with 42.6% in private practice and 34.6% in academic settings. The results were published as "Assessing Artificial Intelligence in Breast Imaging: A Survey of Breast Radiologists' Insights on Adoption, Benefits, and Challenges" in the Journal of Breast Imaging in 2026, and covered by Radiology Business.
Adoption itself is now mainstream: 55.6% of respondents said they use AI-aided computer-aided detection (CAD) in their practice. Among those users, 71.0% reported improved work efficiency and 65.2% reported increased cancer detection. But only 39.1% reported a reduction in recall rates — the outcome that most directly affects how many patients get called back for additional imaging or biopsy.
When all respondents — users and non-users — rated AI's overall impact on their practice, the pattern repeated. 73.2% called the impact on workflow efficiency "great," and 65.8% said the same for patient care. But reduced burnout (47.0%), faster turnaround time (47.0%), and easing radiologist shortages (46.3%) all landed under half. AI-CAD also generated its highest false-positive rates on postsurgical scar (75.9%) and benign calcifications (69.3%), which the study's authors flagged as a limit on diagnostic utility.
Perception is outrunning the outcomes
Line the numbers up and a pattern appears: the more general and workflow-oriented the question, the higher the "yes." The more specific and outcome-oriented the question — did this actually reduce recalls, burnout, or turnaround time — the lower it drops.
| What was rated | % reporting benefit | Type of measure |
|---|---|---|
| Workflow efficiency | 73.2% | General perception |
| Improved work efficiency (AI-CAD users) | 71.0% | General perception |
| Patient care, overall | 65.8% | General perception |
| Increased cancer detection | 65.2% | Perceived diagnostic gain |
| Reduced burnout | 47.0% | Hard outcome |
| Faster turnaround time | 47.0% | Hard outcome |
| Easing radiologist shortages | 46.3% | Hard outcome |
| Reduced recall rates | 39.1% | Hard, patient-facing outcome |
None of this means breast imaging AI doesn't work — the same survey found majorities believe it enhances cancer detection and workflow. It means the confidence radiologists have in AI as a general workflow aid is running ahead of the harder, outcome-level proof: fewer callbacks, less burnout, shorter turnaround, measurable relief on staffing. Barriers reported alongside adoption reinforce the same story — cost (71.5%), software integration (62.0%), and lack of trust in the tools (63.3%) were all cited by a majority of respondents.
Why this isn't just a mammography problem
Breast imaging is where most AI detection tools have been deployed and studied the longest, which is precisely why this survey matters beyond mammography: it's the modality with the most mature AI market, and even there, radiologists say the evidence for hard outcomes lags the workflow story. That's a structural pattern worth watching wherever radiology AI is sold on efficiency claims — CT, MRI, or otherwise — not a one-off finding about breast screening.
A detection tool that flags a lesion doesn't, by itself, produce a shorter turnaround time, a lower recall rate, or less burnout — those depend on what happens after the flag: how the finding gets into a report, how much editing a radiologist has to do, and whether the tool's output is trustworthy enough to act on without re-verifying everything underneath it. That gap between "AI touched the case" and "the outcome measurably improved" is exactly what this survey is picking up.
What would make CT reporting AI different
Closing this kind of gap means measuring the same things radiologists say matter — turnaround time, report completeness, downstream recall or follow-up rates — rather than resting on perceived workflow benefit. It also means the AI has to produce something a radiologist can be accountable for, not a black-box flag layered on top of an existing report. AI CT reporting built around a full, structured draft report — reviewed in-house before it ever reaches the client's reading radiologist, who signs the final — is designed to generate that kind of measurable, ready-to-sign output, and to be evaluated on outcomes rather than on adoption numbers alone.
Frequently asked questions
Do radiologists report meaningful clinical benefits from breast imaging AI?
Not consistently. In a 2026 Society of Breast Imaging member survey published in the Journal of Breast Imaging, 55.6% of respondents used AI-aided computer-aided detection (CAD). Among those users, 71.0% reported improved work efficiency and 65.2% reported increased cancer detection, but only 39.1% reported a reduction in recall rates. When all respondents rated AI's overall impact, workflow efficiency (73.2%) and patient care (65.8%) scored highest, while reduced burnout (47.0%), faster turnaround time (47.0%), and easing radiologist shortages (46.3%) all fell under half.
What survey found this gap between AI adoption and clinical outcomes?
The Society of Breast Imaging distributed an anonymous, IRB-exempt survey to its 2,264 members in May 2024; 162 responded (a 7.2% response rate). The results were published as 'Assessing Artificial Intelligence in Breast Imaging: A Survey of Breast Radiologists' Insights on Adoption, Benefits, and Challenges' in the Journal of Breast Imaging in 2026.
Why does AI adoption in radiology outpace measurable outcomes?
The survey suggests a gap between perceived workflow impact and hard clinical outcomes. Respondents rated general impressions like workflow efficiency and patient care highly, but the specific, measurable outcomes tied to patient results and physician wellbeing — recall rates, burnout, turnaround time, and workforce shortages — were rated as a meaningful benefit by fewer than half. Barriers reported alongside adoption include cost (71.5%), software integration (62.0%), and lack of trust in the tools (63.3%).
What would close the evidence gap for radiology AI, including CT reporting?
Closing the gap means measuring the same outcome metrics radiologists say matter most — turnaround time, report completeness, downstream recall or follow-up rates — rather than relying on perceived workflow impact alone. It also means keeping a radiologist accountable for every report the AI touches: AI CT reporting built around a full structured draft, an in-house radiologist review of every preliminary, and the client's reading radiologist signing the final is designed to produce that kind of measurable, ready-to-sign output rather than an unaccountable black-box detection flag.
Source: Nenow J, Abdelrahman OA, Osman SOS, Couillard S, Parghi C, Zhang Z. "Assessing Artificial Intelligence in Breast Imaging: A Survey of Breast Radiologists' Insights on Adoption, Benefits, and Challenges." Journal of Breast Imaging (2026). DOI: 10.1093/jbi/wbaf079. As reported by Radiology Business. Figures are rounded as reported.