人工智能在乳腺X线检查中的应用价值

Application value of artificial intelligence in mammography

目的探讨人工智能(AI)在乳腺X线检查中的应用价值。方法收集2020年1月至2021年12月于北京市密云区妇幼保健院行乳腺X线检查并能追踪到病理结果的122例妇女。分别进行医师阅片、AI阅片及医师在AI辅助下阅片,以病理结果为诊断金标准,根据诊断结果绘制ROC曲线,计算曲线下面积(AUC),评估诊断性能。结果医师诊断、AI诊断、医师在AI辅助下诊断的AUC分别为0.835、0.738、0.902;医师在AI辅助下诊断的AUC值大于独立诊断,差异有统计学意义(P<0.05);医师在AI辅助下诊断的灵敏度为90.0%,特异度为90.3%,均高于独立诊断。结论在乳腺X线诊断中,AI可提高医师的诊断效能,具有一定的临床意义及参考价值。

Objective To explore the application value of artificial intelligence(AI)in mammography.Methods From January 2020 to December 2021,a total of 122 women who underwent mammography in Beijing Miyun District Maternal and Child Health Hospital and whose pathological results could be traced were collected.Physician reading,AI reading and doctor reading with the aid of AI were performed respectively.The pathological results were used as the diagnos tic gold s tandard.ROC curves were drawn according to the diagnostic results,and the area under the curve(AUC)was calculated to evaluate the diagnostic performance.Results The AUC of physician diagnosis,AI diagnosis,and physician’s AI-assisted diagnosis were 0.835,0.738,and 0.902,respectively.The AUC value diagnosed by physicians assisted by AI was greater than that of independent diagnosis,and the difference was statistically significant(P<0.05).The sensitivity and specificity of physicians’diagnosis under the assistance of AI were 90.0%and 90.3%,which were higher than those of independent diagnosis.Conclusion In mammography diagnosis,AI can improve the diagnostic efficiency of physicians,and has certain clinical significance and reference value.