@article { author = {Gharbali, Akbar}, title = {Computerize classification of Benign and malignant thyroid nodules by ultrasound imaging}, journal = {Iranian Journal of Medical Physics}, volume = {15}, number = {Special Issue-12th. Iranian Congress of Medical Physics}, pages = {172-172}, year = {2018}, publisher = {Mashhad University of Medical Sciences}, issn = {2345-3672}, eissn = {2345-3672}, doi = {10.22038/ijmp.2018.12680}, abstract = {Introduction: Early detection and treatment of thyroid nodules increase the cure rate and provide optimal treatment. Ultrasound is the chosen imaging technique for assessment of thyroid nodules. Confirmation of the diagnosis usually demands repeated fine needle aspiration biopsy (FNAB). So, current management, has morbidity and non zero mortality. The goal of the present study is to explore diagnostic potential of automatic texture analysis (TA) methods in differentiation benign and malignant thyroid nodules by ultrasound imaging in order to help for reliable diagnosis and monitoring of the thyroid nodules with no need more biopsy. Materials and Methods: The database consist of 70 thyroid patients (26 benign and 44 malignant). They under- went thyroid sonography which were reported by radiologist and proven by the biopsy. One ultrasound image per patient was loaded in Mazda Software version 4.6 for automatic texture analysis. Regions of interests (ROIs) were defined within the abnormal part of the thyroid nodules ultrasound images. Gray levels intensity within a ROIs normalized and then up to 270 multi scale texture features parameters per ROIs per normalization schemes were computed via statistical methods employed in Mazda software. From the statistical point of view, all calculated texture features parameters are not useful for texture analysis. So, the features based on F: maximum Fisher coefficient and P: minimum probability of classification error and average correlation coefficients (POE+ACC) eliminated to 10 best and most effective features per normalization schemes. We analyze this feature under two standardization states standard (S) and nonstandard (nS) with Linear Discriminate Analysis (LDA) and None Linear Discriminate Analysis (NDA). The first nearest neighbor (1NN) and artificial neural network (A-NN) classifier were performed for features obtained via LDA and NDA respectively to differential diagnosis benign versus malignant thyroid nodules. The confusion matrix applied between visual and automated texture analysis results to find out sensitivity and specificity of the applied texture analysis methods under different applied options. The Receiver operating characteristic (ROC) curve analysis used for comparison discrimination performance of the employed texture analysis methods.Results: The results demonstrated the influence of the normalization, reduction and standardization on the effectiveness of the obtained features in discrimination tasks by LDA and NDA texture analysis.In comparison with LDA, The selected subset features represent the highest discrimination performance for NDA in distinguishing benign from malignant thyroid with sensitivity of 94%, specificity of 100%, and Az value of 0.97. It means that via our method 97% patient with thyroid nodules lesions can be diagnostic without doing clinical biopsy. For LDA, this discrimination result has sensitivity of 85.7%, specificity of 87.5%, and Az value of 0.86.Conclusion: Our results indicate Computer aided diagnosis can provide useful information to help radiologists in classification of benign and malignant thyroid nodules.}, keywords = {Ultrasound,Thyroid Nodules,Computer Aided Diagnosis Texture Analysis}, url = {https://ijmp.mums.ac.ir/article_12680.html}, eprint = {} }