According to a recent survey published in "Clinical Imaging," researchers from the UC San Diego Health Center surveyed 215 members of the American Society of Breast Imaging and found that AI tools used for breast cancer detection did not meet the expectations of radiologists in clinical practice. Although nearly half of the respondents use FDA-approved AI diagnostic tools in their daily work, and another 11% plan to introduce them, most doctors only regard them as a "second opinion" for diagnosis, with very few considering them a decisive factor in decision-making.

AI in Healthcare

Data shows a significant gap between the actual effects of AI in improving efficiency and reducing workload and what doctors expected: only 35% of respondents reported a decrease in recall rates, far below the expected 59%; only 9% of doctors observed a reduction in unnecessary biopsies, while the expected proportion was 36%; meanwhile, only 29% of doctors felt a reduction in burnout, lower than the expected 56%. Currently, high costs and lack of institutional support remain the core bottlenecks limiting the widespread adoption of such AI tools.

As early as ten years ago, some well-known scholars in the industry predicted that radiologists would be quickly replaced by AI, and similar claims about computer-based jobs being replaced have recently resurfaced. NVIDIA CEO Jensen Huang criticized such views, calling them the "God complex" in AI-driven unemployment predictions. This study shows that the implementation of AI in high-end professional fields like healthcare is unlikely to happen overnight, and its core value in the short term remains in human-AI collaboration for diagnostic assistance, rather than fully replacing human experts.