PDF(Pubmed)
Abstract：
Objective Evaluating an artificial intelligence (AI) tool (AIATELLA, version 1.0; AIATELLA Oy, Helsinki, Finland) in interpreting cardiac magnetic resonance (CMR) imaging to produce measurements of the aortic root and valve by comparison of accuracy and efficiency with that of three National Health Service (NHS) cardiologists. Methods AI-derived aortic root and valve measurements were recorded alongside manual measurements from three experienced NHS consultant cardiologists (CCs) over three separate sites in the northeast part of the United Kingdom. The study utilised a comprehensive dataset of CMR images, with the intraclass correlation coefficient (ICC) being the primary measure of concordance between the AI and the cardiologist assessments. Patient imaging was anonymised and blinded at the point of transfer to a secure data server. Results The study demonstrates a high level of concordance between AI assessment of the aortic root and valve with NHS cardiologists (ICC of 0.98). Notably, the AI delivered results in 2.6 seconds (+/- 0.532) compared to a mean of 334.5 seconds (+/- 61.9) by the cardiologists, a statistically significant improvement in efficiency without compromising accuracy. Conclusion AI\'s accuracy and speed of analysis suggest that it could be a valuable tool in cardiac diagnostics, addressing the challenges of time-consuming and variable clinician-based assessments. This research reinforces AI\'s role in optimising the patient journey and improving the efficiency of the diagnostic pathway.
摘要：
客观评估人工智能(AI)工具(AIATELLA，版本1.0;AIATELLAOy,赫尔辛基,芬兰）通过与三位国家卫生服务（NHS）心脏病专家的准确性和效率进行比较，来解释心脏磁共振（CMR）成像以产生主动脉根部和瓣膜的测量值。方法在英国东北部的三个不同地点，记录了三名经验丰富的NHS顾问心脏病学家（CC）的人工测量结果，同时记录了AI衍生的主动脉根部和瓣膜测量结果。这项研究利用了一个全面的CMR图像数据集，组内相关系数（ICC）是AI和心脏病专家评估之间一致性的主要指标。患者成像是匿名的，并且在传输到安全数据服务器时是盲的。结果该研究表明，与NHS心脏病专家进行的主动脉根部和瓣膜的AI评估之间的一致性很高（ICC为0.98）。值得注意的是,AI在2.6秒内（+/-0.532）交付结果，而心脏病专家的平均值为334.5秒（+/-61.9），在不影响准确性的情况下，效率的统计显着提高。结论AI的准确性和分析速度表明它可能是心脏诊断的有价值的工具。解决耗时和可变的基于临床医生的评估的挑战。这项研究加强了AI在优化患者旅程和提高诊断途径效率方面的作用。
