Artificial Neural Network for Thyroid Disease Diagnosis

Thyroid is one of the major glands in the human body. It is responsible for various hormones secretion like Triiodothyronine (T3) and Thyroxin (T4). Indeed, they stimulate our metabolism and act on various organs. So, its malfunction can cause various disorders, like hypothyroidism, hyperthyroidism and thyroid nodules as small tumors in the gland, which can be developed into cancers. This complex situation involves early discovery and updated monitoring of patients’ health status. The stunning advancements of modern computational technologies led by Artificial Intelligence (AI) gave birth to significant areas such as Machine Learning (ML) applications. ML increasingly contributes to the improvement of various fields, including the domain of medicine and healthcare. The main objective is to improve diagnosis through the classification of acute health crisis and the prediction of symptoms or health problem. In this study, the proposed intelligent model mainly aims at performing an accurate predictive diagnosis on the people’s chance of obtaining thyroid disease. It can be useful to support endocrinologists’ examination at an early stage, which can reduce time and costs and enhance patients’ well-being. The proposed predictive model is based on artificial neural networks (ANN) by considering patients’ clinical and laboratory dataset (T3, T4, TSH…). Many techniques were applied such as forward and backward propagation algorithms to minimize prediction error, as well as k-fold cross validation method to split the dataset. As a result, the analysis and the evaluation of the experimental results show a very good performance of the suggested model (93.58% of accuracy).