Deep Learning for thyroid Nodule Detection and Classification

Thyroid cancer is the most common endocrine malignancy, particularly affecting women, with ultrasound being the primary diagnostic tool. Although the TI-RADS system has improved specificity in classifying thyroid nodules, subjectivity among radiologists leads to inconsistent judgments and unnecessary biopsies. Deep learning has shown potential to overcome these challenges by matching or exceeding expert diagnostic accuracy. This project aims to systematically select and build a deep-learning pipeline that accurately localizes thyroid nodules in ultrasound images and classifies their malignancy risk.