Thyroid nodules segmentation methods in clinical ultrasound images: A review

Abstract

Thyroid is an extremely essential butterfly shaped organ which is positioned in front of neck. For finding thyroid disease, ultrasound imaging is considered a best for identification of thyroid nodules because of its inexpensive and trouble-free nature. For diagnosis of thyroid disorder, it is very important to separate thyroid nodules from thyroid gland. For the same in biomedical image processing, image segmentation and classification followed by image preprocessing is very imperative steps. The key goal of this work is to categorized of thyroid nodules, thyroid gland segmentation methods. This paper explained different segmentation methods like counter and shaped based, region based and machine learning and deep learning strategies for thyroid gland and thyroid nodules. In addition, grouping of thyroid knobs is classified by traditional and convolution neural network for different feature extraction methods. All these methods are evaluated and compared by different quality metrics. After careful assessment of plenty of paper we did comparative and comprehensive analysis of segmentation methods with deep learning. Segmentation by deep learning approach gives better performance but it required more marked data base. In thyroid nodules segmentation region-based methods and contour and shape-based methods gives better performance. This paper concludes segmentation methods and their performance which is based on image attributes applications etc. The performance is assets by accuracy, selectivity and specificity and mean overlap parameters. Moreover, we concentrate on several advantages and disadvantages of current thyroid nodules ultrasound image segmentation of thyroid nodules and thyroid gland methods for 2Dimentionsl Ultrasound image, which would be useful for future studies such as identification of specific thyroid disease.