融合先验知识特征的甲状腺乳头状癌中央区淋巴结转移人工智能诊断模型

Objective: To establish an artificial intelligent diagnosis model of CLNM in patients with PTC based on thyroid ultrasound images, using artificial intelligence technology and fusing prior knowledge. Methods: The clinical data and ultrasound images of 774 patients with PTC who met the research standards in the partner hospitals from January 2018 to December 2019 were collected and sorted, and the corresponding data were marked to construct a new data set for this study. Using deep learning methods such as convolutional neural network, residual learning, and spatial pooling, an intelligent diagnosis model of CLNM in patients with PTC was constructed, and prior knowledge features were incorporated into it. Besides, a comprehensive evaluation of the model performance is carried out. Results: In the test set, the accuracyArtificial Intelligence Diagnosis Model of Central Compartment Lymph Node Metastasis in Papillary Thyroid Carcinoma Fusing Prior Knowledge Features, sensitivity and specificity of this model in predicting CLNM in patients with PTC were 80.62%, 90.24% and 63.83%. Conclusion: The artificial intelligence diagnostic model constructed in this study can be used to diagnose CLNM in patients with PTC. It is especially suitable for initial screening due to its high sensitivity, and can also provide a basis for decision-making of clinical treatment plans.