The thyroid is a little gland in the neck that resembles a butterfly and makes hormones that are vital for controlling a number of metabolic functions. Common thyroid disorders include hypothyroidism, hyperthyroidism, and thyroid nodules, which are abnormal growths or masses inside the thyroid. Thyroid abnormalities, which can have a major influence on metabolism and general health, are commonly detected and classified by thyroid ultrasonography (USG) imaging. Deep learning (DL), in particular, has made it possible to identify and analyze patterns in clinical images because it can extract hierarchical feature representations from images without the requirement for annotated data. This is made possible by advancements in artificial intelligence (AI). Reducing needless fine needle aspiration (FNA) procedures depends on accurately identifying malignant thyroid nodules and differentiating them from benign ones. In this work, DL is used for feature extraction in order to locate thyroid nodules in USG pictures. ResNet-18 and VGG-19, two pre-trained DL models, were optimized for thyroid USG image classification. Digital Database of Thyroid Ultrasound Images (DDTI), a gold standard dataset, was used to train and evaluate both models. According to the results, ResNet-18 outperformed VGG-19 in classification accuracy, with classification accuracies of 97.13% and 90.31%, respectively.