Texture Analysis on Thyroid Ultrasound Images for the Classification of Hashimoto Thyroiditis

Abstract

The biopsy using Fine Needle Aspiration (FNA) is a major procedure/testing method which has been regularly recommended, at a time, exactly when a thyroid nodule is suspected or else identified. The FNA will usually reveal if a nodule is benign or malignant. Histopathology is also sometimes recommended. Another regular test is the ultrasound. Yet, the ultrasound cannot recognize or distinguish the thyroid disorders. Hashimoto thyroiditis is the most widely recognized kind of inflammation of the thyroid gland. The motto of this work is to identify the Hashimoto’s thyroiditis disorder using only ultrasonogram images without going for any painful examination. In this paper, features are studied using the Neighborhood Gray Tone Difference Matrix (NGTDM), Statistical Feature Matrix (SFM), and Laws’ texture energy measures methods. The salient features from the above procedures are helpful to identify and in separating the two types of ultrasonic thyroid images as normal and Hashimoto’s thyroiditis. The student two-tailed unpaired T-test method is employed to classify the two groups. A major difference between the two groups (p < 0.001) was observed. The results are correlated with the histopathology results. The results prove that the Hashimoto thyroiditis can be identified using the ultrasound images.