The Deep Learning-Based Prediction Method for Cervical Lymph Node Metastasis of Thyroid Cancer Based on Multimodal Data

Thyroid cancer is a common malignant tumor worldwide, and there is a risk of misdiagnosis or over-treatment for small-sized thyroid cancer. The key to eliminating this risk lies in predicting cervical lymph node metastasis. The traditional manual prediction methods are not ideal, so this paper proposes an intelligent prediction method based on deep learning. This method is based on common clinical examination data, including ultrasound images, ultrasound image text descriptions, and blood test data. A deep convolutional neural network model is proposed using data from these three modalities as inputs. The model integrates multimodal data features through mean cross-attention and extracts fine-grained classification information from the fused feature map through a bilinear CNN network, demonstrating strong feature discrimination ability. Through collecting data and conducting various experiments, the experimental results show that the model proposed in this paper has good performance in predicting cervical lymph node metastasis, is better than the manual prediction, and has broad application prospects. Future research will continue to strive for data collection, increasing the amount of collected data and improving multi-center representativeness, and enhancing the diagnostic and predictive capabilities of the network model through technological improvements.