Automatic Classification of Thyroid Nodules in Ultrasound Images using MobileNetV2-MSVM Hybrid Model

Although many studies have been conducted on automatic classification methods and the detection of abnormalities in thyroid nodule ultrasound data, this issue remains a topic of interest and continues to be researched, developed, and applied in practice. In this paper, we introduce an automatic thyroid nodule classification method in ultrasound images using MobileNetV2-Multi Support Vector Machine (MSVM) Hybrid Model. This proposed method supports doctors and radiologists in diagnosing thyroid-related diseases with high accuracy. This method was tested with 1730 clinical ultrasound image data and achieved an impressive accuracy of 97.12%, a sensitivity of 98.25%, a specificity of 98.25% and an F-measure of 91.80%. These results indicate that the proposed method has strong potential for practical applications in thyroid nodule classification.