Demystifying Hypothyroidism Detection with Extreme Gradient Boosting and Explainable AI

Hypothyroidism is a prevalent disease of thyroid glands in human. Thyroid is an endocrine gland in vertebrates known to control physiological metabolism by producing few specific hormones - thyroxine (T4) and triiodothyronine (T3) hormones. Hyopthyroidism can be defined as the absence of thyroid hormones T3 and T4 in bloodstream. Pituitary gland produces a hormone named Thyroid-stimulating hormone (TSH) which stimulates the thyroid to produce T43 and T4. Hypothyroidism causes human to gain weight, exhaustion, infertility, cardiovascular illness, dyslipidaemia, etc. Hypothyroidism affects around 5% of the world’s population, with another 5% being undiagnosed. If not treated, hypothyroidism can be life threatening. In this paper, we have addressed this issue by proposing a framework and using state-of-the-art machine learning algorithms: XGBoost, AdaBosst, and CatBoost and represented a comparative analysis among the algorithms with different quantitative performance evaluation metrics. In our work, we have showed XGBoost outperformed the other two models with an accuracy of 99.87%, while AdaBoost had 99.73% accuracy, and CatBoost showed 99.75% accuracy on our test sets. We have further used Explainable Artificial Intelligence (XAI) architectures: LIME and SHAP, to interpret the model’s decision in a comprehensive manner to address the ‘Black Box’ issue of machine learning algorithms.