Thyroid Nodules Classification Of Segmented Image Using Geometric And Moment Features

Abstract

            Ultrasound image is one of the most inexpensive modalities because of its low cost and safety. This research focuses on classification of thyroid nodules by using deep learning segmentation approach. Semantic segmentation plays on important role in today's robotics connected applications. Most of the research nowadays on semantic segmentation fulfilled the way of increasing accuracy of all the models of segmentation with highly computation cost. The US image is segmented using U-Net segmentation and it is coupled with Mobile net deep learning segmentation for high accurate and sensitive feature extraction based upon Geometric features and Moment features. This selected feature is used for further classification. The classifiers used in this research work are Support vector machine and Random forest classifiers and the performance and the parameters are considered for future comparison.