Classification of ultrasonographic thyroid tumor images to tirads categories via texture analysis methods

Thyroid tumors are constanly encountered with using medical imaging techniques or processes. Ultrasonography, fine needle aspiration and routine controls on patients are most used methods to determine, thyroid tumor is benign or malign. Three techinques which are mentoined increments doctors work load and consumes time of doctors and patients. From previous research, majority of thyroid tumors are benign and to reduce work load, there is a risk stratification system which is called TIRADS. There are many types of TIRADS risk system and American College of Radiology Thyroid Imaging and Reporting Data System (ACR-TIRADS) is one of them. However, drawback of this risk strafication system is being subjective which creates subjective judgement about tumor and while doctors classifying tumors, they need to examine multiple features to obtain TIRADS class so that is time wasting. In this study, we worked on thyroid tumors with texture anaylsis and classification algorithms to take off doctors work load and help them to achieve more relevant clinical diagnosis. As the results of this study we achieved accuracy up to %94.7, sensitivity up to %94.5, specificity up to %95.4 and precision up to %96.8 respectively.