Lymph nodes are vital immune organs in the human body, widely distributed throughout the lymphatic system and involved in various immune responses. Lymph node diseases can be categorized into benign and malignant based on the degree of lesion. Traditional methods for classifying lymph node diseases rely heavily on radiological features, which may not be the best choice for specific tasks. Additionally, the use of only deep learning methods can lead to overfitting due to limited training data. This study proposes a new fusion method that combines radiomics features with deep learning features, using a gated multimodal fusion mechanism to optimize the feature integration process. This method not only considers the independence of each feature but also dynamically adjusts the weight of each feature in the final decision through a gating mechanism, effectively compensating for the limitations of single methods. Our fusion framework provides a flexible and effective approach for utilizing information from different sources to achieve more accurate and robust classification of lymph node ultrasound images. Experimental results show that the model achieved an AUC of 93.23% in the task of benign and malignant classification of lymph node ultrasound images, outperforming traditional radiological methods, standalone deep learning models, and other hybrid methods, significantly enhancing the accuracy and reliability of the classifications.