AI tools for breast cancer detection are moving into clinical use, but radiologists are not seeing the level of impact many had hoped for. A survey published in Clinical Imaging points to a clear gap between adoption and satisfaction: the technology is present, useful to some, and still far from decisive.
The survey covered 215 members of the Society of Breast Imaging. About half already use FDA-approved AI tools, while 11 percent plan to use them. Yet according to lead author Joud Almogati at UC San Diego Health, only a few consider AI a deciding factor.
Adoption Is Ahead Of Confidence
The survey suggests that breast cancer AI has passed an important threshold: it is no longer only a future idea. FDA-approved tools are already being used by a substantial share of surveyed breast imaging specialists, and another group intends to adopt them.
But adoption alone does not mean the tools have met professional expectations. Radiologists appear to be treating AI as a useful addition rather than a replacement for expert judgment. Most view AI mainly as a second opinion.
That distinction matters. A second opinion can help, but it does not necessarily transform the core decision-making process. If only a few radiologists see AI as a deciding factor, the technology remains supportive rather than central.
The Benefits Are Smaller Than Expected
The clearest gap in the survey is between what radiologists expected AI to improve and what they say it has improved so far. The tools help, but not as much as many anticipated.
- Only 35 percent report lower recall rates, while 59 percent expected them.
- Only 9 percent see fewer unnecessary biopsies, while 36 percent expected that.
- Only 29 percent report less burnout, while 56 percent expected relief.
Those numbers show a pattern. In each area, reported results fall well below expectations. The shortfall is especially sharp on unnecessary biopsies, where reported improvement is much lower than anticipated.
Recall rates, biopsies, and burnout are not abstract concerns. They are practical measures of whether a tool makes radiology work clearer, more efficient, or less burdensome. When gains are modest, radiologists may still use the tools while remaining cautious about their overall value.
Costs And Support Still Slow Wider Use
The survey also identifies the biggest barriers: costs and lack of institutional support. Those barriers can limit adoption even when a tool has regulatory approval and professional interest.
For a clinical team, an AI tool is not just software. It has to fit into a working environment, be paid for, and be supported by the institution using it. If those conditions are missing, even interested radiologists may hesitate.
This helps explain why the survey shows both interest and restraint. About half of surveyed members already use FDA-approved AI tools, and 11 percent plan to. At the same time, the reported benefits are uneven, and the decision to rely on AI remains limited.
The Hype Problem Around AI In Radiology
The article also points to a broader issue: the AI industry may have encouraged expectations that were too high. Prominent AI researchers predicted about ten years ago that radiologists would soon lose their jobs. Similar claims are making the rounds again, this time about knowledge work on computers.
That history matters because it frames how AI tools are judged. If a technology is presented as a near-term substitute for professional expertise, then incremental gains can look disappointing, even when the tool provides some assistance.
Nvidia CEO Jensen Huang has called this kind of prediction a "God complex" among the prophets of AI-driven job loss.
The survey results fit a more grounded view. AI tools for breast cancer detection are being used, and some radiologists report benefits. But the tools have not delivered the broad improvements many expected in recall rates, unnecessary biopsies, or burnout.
For now, the practical role of breast cancer AI appears to be narrower than the strongest predictions suggested. It is a second opinion for many radiologists, not a deciding factor for most. The future of these tools may depend less on sweeping claims and more on whether they can show clearer, measurable value in daily clinical work.