Diagnostic Performance of Artificial Intelligence for Interpreting Thyroid Cancer in Ultrasound images

Thyroid ultrasonography is mainly used for the detection and characterization of thyroid nodules. However, there some limitation since diagnostic performance remains highly subjective depends on radiologist experiences. Therefore, artificial intelligence (AI) was expected to improve ultrasound. To evaluate AI differentiating malignant benign nodules compare it with that an experienced a third-year radiology resident, 648 patients 650 nodules, who underwent ultrasound guided-FNA biopsy had decisive diagnosis from FNA cytology at Siriraj Hospital between January 2014 June 2020, were enrolled. Although specificity accuracy slightly higher in than resident (specificity 78.85% vs. 67.31% 69.23%; 78.46% 70.77% 70.77%, respectively), showed comparable sensitivity (p=0.187-0.855).