Computer aided diagnosis for thyroid cancer system based on internal and external characteristics

Background and aims: Thyroid cancer is one of the fastest growing cancers worldwide. Thyroid ultrasound images are diagnosed based on several characteristics to determine the malignancy of the nodule. The characteristics are divided into two, i.e. external characteristics and internal characteristics. A computer-aided diagnosis (CADx) is necessary to assist radiologists in analysing these characteristics more objectively. Methods: Firstly, a pre-processing step was applied to remove label and reduce speckle noise by applying adaptive median filter followed by bilateral filter. Secondly, active contour and morphology operation were applied to segment the nodules. Subsequently, geometric and texture features were extracted. In the final step, multilayer perceptron was used to classify internal characteristics while support vector machine was used for classifying external characteristics. Results: The sensitivity, specificity and accuracy of nodule classification based on analysis of external characteristics were 100%, 95.45% and 97.78%, respectively, whereas classification results based on internal characteristics were 95.35%, 90.91% and 94.44%, respectively. Conclusion: A computer-aided diagnosis (CADx) for thyroid cancer system based on analysis of external and internal characteristics has been developed. The sensitivity, specificity and level of accuracy for both characteristics showed that the proposed system was reliable to assist radiologist in classifying thyroid nodules. (c) 2019 The Authors. Production and hosting by Elsevier B.V. on behalf of King Saud University. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Automated summary

A computer-aided diagnosis system for thyroid cancer has been developed, which analyzes internal and external characteristics to classify nodules, showing that the system is reliable in assisting radiologists with classification.