Multimodal MRI deep learning for predicting central lymph node metastasis in papillary thyroid cancer

Motivation: In papillary thyroid cancer (PTC), accurately predicting central lymph node metastasis (CLNM) remains challenging, often leading to unnecessary central lymph node dissection (CLND). Goal(s): Our goal was to develop a deep learning (DL) model using multimodal MRI to predict CLNM in PTC patients and compare it with classical machine learning (ML) models. Approach: MRI images from 105 PTC patients were obtained. The DL model was trained on 80% of the data and tested on the remainder. Results: The DL-Fusion model achieved an AUC of 0.891, outperforming all other ML models. Impact: The high predictive accuracy of the DL-Fusion model can enhance clinical decision-making, allowing for more precise assessment of CLNM and reducing the frequency of unnecessary CLND in PTC patients.