A Proposed Algorithm for the Detection of Thyroid Cancer based on Image Processing

Abstract

Introduction: Today's world allows digital images to be downloaded and stored. Computer image processing systems have been developed to make these actions faster and more accurate, especially in medical engineering. Many diseases, including brain tumors, cancers are identified and estimated by benefiting of the image processing. Objective: In this paper, a proposed algorithm for detection the thyroid cancer is presented. Material and Methods: To this end, a few images of thyroid cancer are collected, and then these images are converted to jpeg format. After that, these images are called and converted into a matrix using MATLAB software. The obtained matrix is normalized and then, the noise of these images are removed. Next, the highlighted areas are defined by a threshold value. By summing up this highlighted value and negation of the original images, the cancerous area is determined. Results: The results of the five images illustrate that the accuracy of the algorithm for the detection of thyroid cancer is 93.5% that show an improvement of 3-7 % than other works. Furthermore, sensitivity, specificity, and F-score, Mathew Correlation Coefficient (MCC) are 62%, 93.2%, 92.2%, and 86.4%, respectively, for the proposed method. Conclusions: The method is a simple way and owns an acceptable accuracy that can be used for the detection of thyroid cancer in a portable computer. As the images show the cancerous parts have lighter pixels than non-cancerous parts that the feature is used for separating the cancerous part of the thyroid gland from non-cancerous parts.