放射科医生对人工智能及放射学未来的看法:来自美国全国性调查的见解
该研究对美国322名委员会认证的放射科医生进行横断面调查,评估其对人工智能(AI)在放射学中应用的态度及影响因素。82.9%的医生认为AI将显著影响放射学;年轻医生更乐观且更熟悉AI工具;年龄与乐观态度、性别与对AI角色的信心显著相关;急诊放射科医生中87.5%预计3-5年内采用AI,而儿科放射科医生中26.3%预计6-10年内采用;年轻医生在评估误导性AI输出时更易出错。结论认为AI整合具有变革潜力,但需结构化教育和监管框架。
要点
- 82.9%的放射科医生认为AI将显著影响放射学。
- 年轻医生(<40岁)对AI更乐观且更熟悉,但在评估误导性AI输出时更易出错。
- 急诊放射科医生中87.5%预计3-5年内采用AI,儿科放射科医生中26.3%预计6-10年内采用。
核心结论
- AI整合在放射学中具有变革潜力,但面临过度依赖、熟悉程度差异和亚专科差异等挑战,结构化教育和稳健监管框架对优化AI采用和降低风险至关重要。
研究设计与样本
- 横断面调查,使用Amazon Mechanical Turk(MTurk)和Qualtrics对322名委员会认证的放射科医生进行调查,收集人口学数据并评估对AI整合的态度,采用Pearson卡方检验分析人口学变量与认知的相关性。
诊断标准与关键数值
- 82.9%的放射科医生认为AI将显著影响放射学;年龄与乐观态度显著相关(χ2=47.551,P<.001);性别与对AI角色的信心显著相关(χ2=21.982,P<.001);急诊放射科医生中87.5%预计3-5年内采用AI,儿科放射科医生中26.3%预计6-10年内采用。
临床应用与阅片要点
- 年轻放射科医生在评估误导性AI输出时更易出错,凸显结构化培训项目的必要性。
- 作者认为需要针对人口学和亚专科需求定制结构化AI培训,并建立稳健监管框架以优化AI采用并降低风险。
局限性
- 原文未报告局限性
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
OBJECTIVES: To evaluate US radiologists' attitudes towards artificial intelligence (AI) in radiology, identify demographic factors influencing these perceptions, and analyse the potential challenges and opportunities AI integration presents in radiological practice.
METHODS: A cross-sectional survey of 322 board-certified radiologists was conducted using Amazon Mechanical Turk (MTurk) and Qualtrics. The survey collected demographic data (age, gender, experience, and subspecialty) and assessed attitudes towards AI integration in radiology. Pearson's chi-square tests were used to evaluate correlations between demographic variables and perceptions of AI's impact, confidence in its role, and anticipated adoption timelines.
RESULTS: The majority of radiologists (82.9%) indicated that AI would significantly impact radiology. Younger radiologists (<40 years) displayed higher optimism and greater familiarity with AI tools compared to their older counterparts. Statistical analysis revealed significant correlations between age and optimism (χ2 = 47.551, P < .001) and between gender and confidence in AI's role (χ2 = 21.982, P < .001). Subspecialty differences emerged, with 87.5% of emergency radiologists anticipating AI adoption within 3-5 years, whereas 26.3% of paediatric radiologists predicted adoption within 6-10 years. Notably, younger radiologists showed increased susceptibility to errors when evaluating misleading AI-generated outputs, underscoring the necessity for structured training programs.
CONCLUSIONS: The integration of AI in radiology holds transformative potential but poses challenges, including overreliance, varying familiarity levels, and subspecialty-specific disparities. Structured education and robust regulatory frameworks are critical to optimize AI's adoption and minimize associated risks.
ADVANCES IN KNOWLEDGE: This study highlights significant demographic variations in radiologists' attitudes towards AI and underscores the importance of targeted training and interventions to support effective AI integration. These findings add to the existing research by emphasizing the necessity for structured AI training tailored to demographic and subspecialty needs.
引用
- Alarifi M. Radiologists' views on artificial intelligence and the future of radiology: insights from a US National survey. Br J Radiol. 2026. PMID: 40971485. doi:10.1093/bjr/tqaf222
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。