Thyroid nodules pose a significant clinical challenge, necessitating accurate and efficient diagnosis to differentiate between benign and malignant cases. With advancements in deep learning, Convolutional Neural Networks (CNNs) have demonstrated remarkable performance in medical image classification. This study develops a CNN-based model for thyroid nodule classification using the DDTI Thyroid Ultrasound Images dataset. Three CNN architectures were implemented and evaluated based on accuracy and training loss. The best-performing model achieved 94.7% classification accuracy, demonstrating the effectiveness of deep learning for thyroid ultrasound analysis. The findings highlight the potential of CNN-based classification in medical imaging, offering a reliable tool for assisting radiologists in early detection and diagnosis of thyroid conditions.