Non-invasive Thyroid Detection Using Thermal Imaging Technique

Abstract

Thyroid disorder is due to variation in TSH hormones which helps the human body by controlling metabolism. Thyroid hormones are generally generated in response to a different hormone which is released by pituitary gland. There are four main types of thyroid disorders: Hyperthyroidism (too much thyroid hormone), Hypothyroidism (too little thyroid hormone), Benign (non-cancerous). A thyroid disorders is an abnormal growth of cells within the thyroid gland or inside the throat, which can be cancerous (malign). Thermal distribution in human body is a natural indicator of abnormalities. Thermal imaging is a non – invasive screening method for monitoring the distribution of body temperature. In the view of study we propose classification of Thyroid abnormalities using thermal image. The proposed technique is based on the following computational methods; the median filter for pre-processing, Active Contour segmentation is used to segment the selected Region of Interest and then the features are extracted. Feature extraction was done from the taken images and correlation was estimated between the normal subjects and the thyroid subjects. Various classifiers have been tested for their accuracy and some classifiers such as Multilayer Perceptron, Bayes Net and KNN showed high accuracy. The accuracy estimated by these classifiers was tested in Weka tool and their ROC curves with AUC scores have been derived. These classifiers can be used for feature classification and be efficiently used for the performance of non-invasive thyroid diagnosis.