人工智能在甲状腺结节良恶性中的诊断价值

目的探讨超声S-Detect技术在甲状腺结节良恶性中的诊断价值。方法由3名超声医师对98例患者共136个甲状腺结节的超声图像进行良恶性评估后与S-Detect技术判别结果进行对比。以病理学结果或ATA指南为诊断标准进行对照,分析不同年资医师及S-Detect技术的诊断效能和评价者之间的一致性。结果136个甲状腺结节中良性病变81个,恶性病变55个。S-Detect技术诊断灵敏度高于低、中年资医师(P<0.05),特异度低于不同年资医师(P<0.05),S-Detect诊断准确度为65.44%。S-Detect与高年资医师的诊断一致性中等(Kappa=0.439,P<0.05)。与S-Detect联合诊断可明显提高超声医师的诊断效能(P<0.05)。结论S-Detect技术在甲状腺结节良恶性诊断中的灵敏度较高。与S-Detect技术联合诊断可明显提高超声医师的诊断效能,具有很好的临床应用前景。S-Detect技术与高年资医师诊断一致性相对较好,未来有助于超声图像的标准化判读。

Diagnostic Value of Artificial Intelligence in Benign and Malignant Thyroid Nodules

Objective To evaluate the diagnostic value of S-Detect technique in benign and malignant thyroid nodules.Methods Three ultrasound doctors evaluated 136 thyroid nodules in 98 patients and compared them with S-Detect technique.Finally,the results of pathology or ATA guidelines were used as diagnostic criteria.To analyze the consistency of diagnostic performance and evaluators of different seniority doctors and S-Detect techniques.Results There were 81 benign lesions and 55 malignant lesions in 136 thyroid nodules.The diagnostic sensitivity of S-Detect was higher than that of low-and middle-aged doctors(P<0.05),and the specificity was lower than that of ultrasound doctors(P<0.05).The accuracy of S-Detect was 65.44%.The consistency of S-Detect with senior doctors was good(Kappa=0.439,P<0.05).Combined with S-Detect diagnosis can significantly improve the diagnostic performance of ultrasound doctors(P<0.05).Conclusions S-Detect technique is highly sensitive in the diagnosis of benign and malignant thyroid nodules.S-Detect technique is more effective than younger and middle-aged ultrasound doctors.The combined diagnosis with S-Detect technique can obviously improve the diagnostic efficiency of ultrasound doctors,and has a good clinical application prospect.S-Detect technology has good consistency with the diagnosis of senior doctors,which will be helpful to standardized interpretation of ultrasound images in the future.