Kim G R, Lee E, Kim H R, Yoon J H, Park V Y, Kwak J Y
From the Department of Radiology (G.R.K., J.H.Y., V.Y.P., J.Y.K.), Severance Hospital, Research Institute of Radiological Science, Center for Clinical Imaging Data Science, Yonsei University College of Medicine, Seoul, Korea.
Department of Computational Science and Engineering (E.L.), Yonsei University, Seoul, Korea.
AJNR Am J Neuroradiol. 2021 Aug;42(8):1513-1519. doi: 10.3174/ajnr.A7149. Epub 2021 May 13.
Comparison of the diagnostic performance for thyroid cancer on ultrasound between a convolutional neural network and visual assessment by radiologists has been inconsistent. Thus, we aimed to evaluate the diagnostic performance of the convolutional neural network compared with the American College of Radiology Thyroid Imaging Reporting and Data System (TI-RADS) for the diagnosis of thyroid cancer using ultrasound images.
From March 2019 to September 2019, seven hundred sixty thyroid nodules (≥10 mm) in 757 patients were diagnosed as benign or malignant through fine-needle aspiration, core needle biopsy, or an operation. Experienced radiologists assessed the sonographic descriptors of the nodules, and 1 of 5 American College of Radiology TI-RADS categories was assigned. The convolutional neural network provided malignancy risk percentages for nodules based on sonographic images. Sensitivity, specificity, accuracy, positive predictive value, and negative predictive value were calculated with cutoff values using the Youden index and compared between the convolutional neural network and the American College of Radiology TI-RADS. Areas under the receiver operating characteristic curve were also compared.
Of 760 nodules, 176 (23.2%) were malignant. At an optimal threshold derived from the Youden index, sensitivity and negative predictive values were higher with the convolutional neural network than with the American College of Radiology TI-RADS (81.8% versus 73.9%, = .009; 94.0% versus 92.2%, = .046). Specificity, accuracy, and positive predictive values were lower with the convolutional neural network than with the American College of Radiology TI-RADS (86.1% versus 93.7%, < .001; 85.1% versus 89.1%, = .003; and 64.0% versus 77.8%, < .001). The area under the curve of the convolutional neural network was higher than that of the American College of Radiology TI-RADS (0.917 versus 0.891, = .017).
The convolutional neural network provided diagnostic performance comparable with that of the American College of Radiology TI-RADS categories assigned by experienced radiologists.
卷积神经网络与放射科医生的视觉评估在甲状腺癌超声诊断性能方面的比较结果并不一致。因此，我们旨在评估卷积神经网络与美国放射学会甲状腺影像报告和数据系统（TI-RADS）相比，在使用超声图像诊断甲状腺癌时的诊断性能。
2019年3月至2019年9月，通过细针穿刺、粗针活检或手术，对757例患者的760个甲状腺结节（≥10mm）进行了良恶性诊断。经验丰富的放射科医生评估了结节的超声特征，并指定了美国放射学会TI-RADS 5类中的1类。卷积神经网络根据超声图像提供结节的恶性风险百分比。使用约登指数计算截断值下的敏感性、特异性、准确性、阳性预测值和阴性预测值，并在卷积神经网络和美国放射学会TI-RADS之间进行比较。还比较了受试者操作特征曲线下的面积。
760个结节中，176个（23.2%）为恶性。在由约登指数得出的最佳阈值下，卷积神经网络的敏感性和阴性预测值高于美国放射学会TI-RADS（81.8%对73.9%，P = 0.009；94.0%对92.2%，P = 0.046）。卷积神经网络的特异性、准确性和阳性预测值低于美国放射学会TI-RADS（86.1%对93.7%，P < 0.001；85.1%对89.1%，P = 0.003；64.0%对77.8%，P < 0.001）。卷积神经网络的曲线下面积高于美国放射学会TI-RADS（0.917对0.891，P = 0.017）。
卷积神经网络提供的诊断性能与经验丰富的放射科医生指定的美国放射学会TI-RADS类别相当。
