Diagnosis is an important task in medical science because of its criticality, efficiency and accuracy in determining whether or not a patient has a particular disease. This shall further decide the most suitable line of treatment. There has been a large increase in the number of thyroid cases over the past few years. Since thyroid has a complex relation with metabolism and body weight, it is extremely important to diagnose thyroid disease as early as possible. This paper presents an exhaustive survey of work done in the past with respect to semi-automated medical diagnosis in general and thyroid disease diagnosis in particular. Medical Diagnosis encompassed the use of classifiers like Fuzzy Neural Networks, k-Nearest Neighbor and Decision Tree, while the latter included the use of Computer-Aided Diagnosis, different Neural Networks and Support Vector Machine. Amongst these, the impact of Feature Selection using Particle Swarm Optimization and Ant Colony Optimization on classification was also surveyed. © 2015 IEEE.