ABSTRACT
The thyroid gland, a little butterfly-shaped gland at the front of the neck, generates hormones which govern metabolism. Thyroid problems are most typically detected and classified via ultrasound (US) imaging. US imaging has become one of the most important contributions for analyzing thyroid disorders due to its safety, accessibility, non-invasiveness and cost-effectiveness. Machine learning (ML) advances, especially deep learning (DL) is proving to be beneficial in recognizing and quantifying patterns in clinical images. At the heart of these advancements is DL algorithms’ ability to extract hierarchical feature representations directly from images, eliminating the requirement for constructed features. This study describes the evolution of ML, the concepts of DL algorithms, and an overview of successful applications, including clinical picture segmentation for US imaging of thyroid-related illnesses. Finally, certain research difficulties are mentioned along with future enhancements.
