Classification of thyroid diseases using machine learning frameworks

Abstract

Thyroid diseases are instigated due to disparity in production of hormones –TSH, T4 and T3. Most of the patients of thyroid dysfunction go untreated due to late detection or no detection at all. Machine learning based models for detection of thyroid diseases offer a significant assistance to healthcare. The medical history of the patient supplies the features required by machine learning based classification and prediction models for thyroid dysfunction. The aim of this research paper is to acquire a classification model based on machine learning techniques for assessment of euthyroidism, hyperthyroidism, and hypothyroidism among males, females, and children. Different machine learning classification algorithms such as naïve bayes, decision tree, random forest and logistic regression are used for classification of real data. The accuracy of each of the techniques has established using metrics like precision, recall, specificity and sensitivity. A thyroid dataset has been retrieved from two hospitals in Haryana from January 2020 to July 2020 to train the proposed model. The dataset comprises of medical history of 539 thyroid patients including children, men, and women of various ages. Out of 539 patients screened, 163 have irregular TSH, 138 have prevalence of elevated TSH with 376 having minimal TSH elevation.