TDTD: Thyroid disease type diagnostics

Recently; medical data mining has become one of the well-established research areas of machine learning and AI base techniques have been used to solve the complex classification problem of thyroid disease. Due to the existence of non-palpable nodules it is very hard to detect the structural changes of thyroid disease by assessing the thyroid functional changes. For instance at structural level “Euthyroid” is normal thyroid hormonal functional state but this would be involved in initial structural changes such as goiter, cold nodule, MNG (multiple nodule goiter) and cancer (Grave's Disease and so on). The ideal system should not only identify all the thyroid disease types but also recommend state of structural levels; otherwise it would be converted into serious disease, such like cancer. In-order to mitigate such problems, this paper proposes a framework TDTD (Thyroid Disease Types Diagnostics) that aims to assist the physicians during the diagnostic process of thyroid diseases in a very structured and transparent manner. Proposed system TDTD presents a novel method MDC (Medical data cleaning) for filling of missing values in medical datasets by building classifier based upon the Bayesian isotonic regression algorithm because missing values of medical data (i.e. blood tests) are different in nature and they could not be filled with normal procedures. In second phase two classifiers are trained to classify the functional and structural levels of thyroid disease at granular level using multi and binary SVM (support vector machine) algorithms, in final phase performance and evaluation is approximated using confusion matrix, precision and recall measures.