The increasing incidence of thyroid cancer cases and the challenge of unreliable false positive diagnostic rates in expert-reviewed ultrasound images emphasize the critical need for precise tumor diagnosis. Convolutional Neural Networks (CNNs), a state-of-the-art deep learning technique, exhibit remarkable capabilities in addressing computer vision challenges. This study introduces a specialized AlexNet CNN model designed for the detection of thyroid cancer in pre-processed ultrasound (US) images. The experimental outcomes reveal that the proposed model achieves an accuracy of 0.5, sensitivity of 0.3846, specificity of 0.7143, Positive Predictive Value (PPV) of 0.7143, and Negative Predictive Value (NPV) of 0.385.