Automatic classification of migraine and tension-type headaches using machine learning methods

Migraines and tension-type headaches (TTH) are the more prevalent types of primary headaches. Usually, the diagnosis of these types of headaches is imprecise. Meanwhile, emotional and cognitive dysfunctions are associated with these disorders and have potential diagnostic power. This paper aims to classify patients with TTH and migraines using machine learning methods. It uses psychological questionnaires and demographic data as features for the classification. The Boruta algorithm is applied to select the most relevant features, and various machine learning models are compared to find the best one. In this study, 79 patients with migraine headaches and 81 with TTH referring to the two neurology clinics in Isfahan City, Iran, from 2019 to 2020 completed the cognitive emotion regulation questionnaire and perseverative thinking questionnaire. Results show that random forest (RF) and naïve Bayes achieve the highest accuracy, sensitivity, and specificity in distinguishing TTH and migraine. In fact, RF and naïve Bayes outperformed other classifiers in training and testing sets with a slight difference. The TTH and migraine were classified with 98% accuracy, 100% sensitivity, and 96% specificity in the testing set. Applying machine learning algorithms to distinguish TTH and migraine using these disorders’ psychological roots helps significantly in the more accurate diagnosis and treatment of these two disorders.