Radiomics based on magnetic resonance imaging for preoperative prediction of lymph node metastasis in head and neck cancer: Machine learning study
Yuepeng Wang, Yuepeng Wang, Taihui Yu, Zehong Yang, Yuwei Zhou, Ziqin Kang, Yan Wang, Yan Wang, Zhiquan Huang
Abstract
BACKGROUND: In this study, we use machine learning techniques to develop an efficient preoperative magnetic resonance imaging (MRI) radiomics approach for evaluation of cervical lymph node (CLN) status. METHODS: After collecting all patients' MRI images, we used CLN radiomic features, the apparent diffusion coefficients (ADC) from diffusion-weighted imaging (DWI), and lymph node short diameter of the CLN to build MRI model to predict the status of the CLN. RESULTS: One hundred and twenty cases met inclusion criteria. The MRI model including the radiomic features, ADC, and lymph node size of the CLN achieved better performance for CLN status prediction with the area under the receiver operating characteristic (ROC) curve (AUC) of 0.83. CONCLUSIONS: The multiomic signature of MRI radiomics, ADC, and lymph node size of CLNs has high predictive value for the status of CLNs. This model has provided scientific value to the surgeon regarding cervical lymph nodes before surgery.