【摘要】 目的 探讨MobileNetV2模型在甲状腺结节微波消融术后消融灶检测中的辅助诊断价值。方法 收集患甲状腺结节的患者200例和行甲状腺结节微波消融治疗的患者200例，在测试集上评估经训练集训练的MobileNetV2模型性能，并与医师组分类比较。结果 在测试集上，MobileNetV2模型诊断消融灶的准确度为0.90%,灵敏度为91%,特异度为89%,ROC曲线下面积(AUC)为0.90,大致与高年资医师相同，高于中、初级医师，差异有统计学意义。结论 MobileNetV2模型可以提高甲状腺结节微波消融灶检出的准确度，在消融术后随访中改善超声诊断流程。更多还原

【Abstract】 Objective To explore the diagnostic value of MobileNetV2 model in the ablated lesion detection after thyroidal nodule microwave ablation. Methods We collected two hundred patients with thyroid nodules and two hundred patients who underwent microwave ablation of thyroid nodule. The performance of the MobileNetV2 model trained in the training set was evaluated on the test set and compared with that of the physician group. Results On the test set, the diagnostic accuracy, sensitivity, specificity and area under roc the ROC curve of MobileNetV2 model were 0.90%, 91%, 89% and 0.90 respectively, which were roughly the same as those of senior doctors, and higher than those of middle rank and junior doctors, and all these differences were statistically significant. Conclusions The MobileNetV2 model can effectively improve the accuracy of ablated lesion detection after thyroidal nodule microwave ablation, and optimize the procedure of ultrasonic diagnosis during the follow-up after microwave ablation.更多还原