超声人工智能辅助诊断系统用于甲状腺髓样癌

Ultrasonic artificial intelligence-assisted diagnostic system for diagnosing medullary thyroid carcinoma

目的 以甲状腺乳头状癌(PTC)为对照,对比超声甲状腺人工智能(AI)辅助诊断系统(AI辅助诊断系统)与不同年资超声医师诊断甲状腺髓样癌(MTC)的效果。方法 纳入经病理证实的63枚MTC、70枚PTC和62枚良性结节。以AI辅助诊断系统分析并识别结节,将恶性概率值≥0.40者诊断为恶性结节;由高、中及初级职称医师各1名利用我国甲状腺影像报告和数据系统(C-TIRADS)对甲状腺结节进行分类;对比两种方法诊断MTC及PTC的效能。结果 AI辅助诊断系统诊断MTC和PTC的敏感度、特异度、阳性预测值、阴性预测值、准确率及曲线下面积(AUC)均低于3名医师;高、中级职称医师与AI辅助诊断系统诊断MTC和PTC的AUC差异均有统计学意义(P均<0.01),初级职称医师与AI辅助诊断系统AUC差异均无统计学意义(P均>0.05)。AI辅助诊断系统诊断MTC的敏感度、特异度、阳性预测值、阴性预测值、准确率及AUC均低于其诊断PTC,但AUC差异无统计学意义(P>0.05)。结论 超声甲状腺AI辅助诊断系统诊断MTC效能较高。

Objective To assess the effect of ultrasonic thyroid artificial intelligence(AI)-assisted diagnostic system(AI-assisted diagnostic system)for diagnosing medullary thyroid carcinoma(MTC)compared with different physicians and taken papillary thyroid carcinoma(PTC)as the controls.Methods Totally 63 MTC,70 PTC and 62 benign thyroid nodules confirmed by pathology were enrolled.AI-assisted diagnostic system was utilized to analyze thyroid nodules and identify the likelihood of malignancy,and the probability value threshold was set at≥0.40.All thyroid nodules were retrospectively reviewed and categorized by 3 physicians(1 senior physician,1 attending physician and 1 junior physician)according to Chinese thyroid imaging reporting and data system(C-TIRADS).The efficacy of AI-assisted diagnostic system and physicians for diagnosing MTC and PTC were evaluated.Results AI-assisted diagnostic system showed lower sensitivity,specificity,positive predictive value,negative predictive value,accuracy,and area under the curve(AUC)for diagnosing MTC and PTC compared with physicians.Significant differences of AUC were found between senior physician and AI-assisted diagnostic system,as well as between attending physician and AI-assisted diagnostic system for diagnosing MTC and PTC(all P<0.01),while no significant difference of AUC was between junior physicians and AI-assisted diagnostic system(both P>0.05).The sensitivity,specificity,positive predictive value,negative predictive value,accuracy and AUC for AI-assisted diagnostic system for diagnosing MTC were all lower than those for diagnosing PTC,but the AUC was not significantly different(P>0.05).Conclusion Ultrasonic thyroid AI-assisted diagnostic system had relatively high value for diagnosing MTC.