Multilayer Perceptron Predicting Cervical Lymph Node Metastasis for Papillary Thyroid Carcinoma

Abstract

Background: Lymph node metastasis is related to thyroid cancer recurrence; hence, early identification and prediction of cervical lymph node metastasis (CLNM) in thyroid cancer are essential. Materials and methods: Ultrasound characteristics and patients’ clinical information for 478 thyroid nodules from 383 patients were collected, and a multilayer perceptron (MLP) was used to train and test the veracity to predict CLNM and form a network model. Sixty new patients with papillary thyroid carcinoma (PTC) were evaluated with the MLP neural network model. The metastasis status of these patients was then compared with the pathological results. The prediction of metastasis by the MLP and by multiple regression was compared. Results: Calcification, age, sex, and maximum diameter were important predictive factors of CLNM by the MLP, and the area under the receiver operating characteristic curve was 0.715. No significant differences were found between the MLP and multiple regression in predicting CLNM. The average confidence of the model used in these new patients in predicting metastasis with PTC was 68.9%. Conclusion: Ultrasound images from thyroid nodule characteristics and patients’ clinical information can be used as predictive factors of CLNM by the MLP method to a certain extent.

Keywords

• cancer
• lymph node
• metastasis
• predict
• thyroid