An EM-MPM algorithmic approach to detect and classify thyroid dysfunction in medical thermal images

Abstract

In this paper, a non-invasive method to diagnose thyroid using thermal imaging process is proposed. Heat distribution in an object is referred as thermography it is utilised in medical analysis as the human body emits certain amount of heat. The proposed technique is based on the following computational methods expectation maximisation - maximise of the posterior marginal algorithm (EM-MPM) for segmenting the thyroid region, grey-level co-occurrence matrix (GLCM) for feature extraction and support vector machine (SVM) for classifying abnormalities. The experiment was carried out of 40 thermal images of which ten were normal and 30 abnormal (hyper and hypo) from real human thyroid region thermal image. The accuracy of proposed system is 97.5% which is significantly good. As a result domain user are able to analyses the prediction given by the proposed system for decision support tool.