AI辅助多模态超声联合血清促甲状腺激素对甲状腺结节定性的诊断效能

Qualitative efficacy of AI-assisted multimodal ultrasound combined with serum thyroid stimulating hormone in guiding thyroid nodules

目的探究甲状腺结节定性诊断中人工智能(AI)辅助多模态超声联合血清促甲状腺激素(TSH)的指导效能,以期为甲状腺结节临床诊治提供参考。方法选取2021年1月—2022年6月张家口市第一医院就诊的甲状腺结节性疾病患者100例,采用AI常规超声、超声造影、实时剪切波弹性成像(SWE)检查,以病理结果为金标准,将患者分为良性组(23例)与恶性组(77例)。检测2组血清TSH水平,采用ROC曲线分析TSH对甲状腺结节定性的预测价值,并评估AI辅助多模态超声联合TSH对甲状腺结节定性的诊断效能。结果检查发现113个结节,其中26个良性,87个恶性。AI常规超声+超声造影+SWE联合诊断的灵敏度为89.66%(78/87),比较单一及两两联合诊断高(P<0.05)。良性组血清TSH水平比恶性组低(P<0.05)。ROC曲线显示,血清TSH诊断的AUC、灵敏度、特异度、截断值分别为0.800、74.71%、84.62%、4.44 mIU/L。AI辅助多模态超声联合血清TSH诊断的准确度、灵敏度、阴性预测值分别为93.81%(106/113)、97.70%(85/87)、91.30%(21/23),比较单一诊断高(P<0.05)。结论在甲状腺结节定性诊断中AI辅助多模态超声联合血清TSH的指导效能确切,应用价值较高。

Objective To explore the guiding efficacy of artificial intelligence(AI)assisted multimodal ultrasound com-bined with serum thyroid stimulating hormone(TSH)in qualitative diagnosis of thyroid nodules,in order to provide refer-ence for clinical diagnosis and treatment of thyroid nodules.Methods A total of 100 patients with thyroid nodular dis-ease treated in Zhangjiakou First Hospital from January 2021 to June 2022 were selected and examined by AI routine ul-trasound,contra-ultrasound,and real-time shear wave elastography(SWE).The patients were divided into a benign group(23 cases)and a malignant group(77 cases)with pathological results as the gold standard.Serum TSH levels were detected in the two groups.The receiver operating characteristic curve(ROC)was used to analyze the predictive value of TSH in the qualitative diagnosis of thyroid nodules,and the diagnostic efficacy of AI-assisted multimodal ultra-sound combined with TSH in the qualitative diagnosis of thyroid nodules was evaluated.Results A total of 113 nodules were found by pathological findings,including 26 benign nodules and 87 malignant nodules.The sensitivity of AI conven-tional ultrasound+CEUS+SWE combined diagnosis was 89.66%(78/87),which was higher than that of single and pairwise combined diagnosis(P<0.05).The serum TSH level in the benign group was lower than that in the malignant group(P<0.05).The ROC curve showed that the AUC(95%CI),sensitivity,specificity,and cut-off values of serum TSH diagnosis were 0.800,74.71%,84.62%,and 4.44 mIU/L,respectively.The accuracy,sensitivity,and negative predictive value of AI-assisted multimodal ultrasound combined with serum TSH diagnosis were 93.81%(106/113),97.70%(85/87),and 91.30%(21/23),respectively,which were higher than that of single diagnosis(P<0.05).Conclusion AI-assisted multimodal ultrasound combined with serum TSH is effective and valuable in the qualitative di-agnosis of thyroid nodules.