Feature-Based Discrimination of Thyroid Cancer on Ultrasound Images

Abstract

In this study, to improve the diagnostic accuracy of thyroid cancer, we evaluate the performance of feature-based models on ultrasound image for discrimination. A automated segmentation method was used to characterize microcalcifications, masses and nodules. The model performed discrimination of thyroid cancer based on texture and morphological features extracted from ultrasound images. Performances were compared to benchmark models. Our feature-based model achieved a discriminatory accuracy compared to other methods. Experimental results on three scenarios demonstrate the effectiveness of the proposed model. Especially for samples that are very difficult for doctors to judge, the model showed good performance. This may have clinical value for early detection and treatment of thyroid cancer.