Abstract
Headache disorders rank among the most disabling neurological conditions worldwide, yet clinical differentiation between migraine, tension-type headache and cluster headache continues to pose diagnostic challenges even for experienced neurologists. The overlap in symptom profiles, reliance on patient recall, and time constraints in outpatient settings collectively contribute to diagnostic delays that may extend over several years — particularly in the case of cluster headache. Over the past decade, machine learning methods have been applied to structured symptom questionnaires, free-text diary entries and wearable physiological signals with the aim of automating headache subtype classification according to ICHD-3 criteria. A review of 28 relevant studies published between 2018 and 2024 reveals that ensemble classifiers such as random forests and gradient boosting achieve three-class diagnostic accuracy in the range of 82–93%, considerably outperforming single-symptom rule-based screening. NLP models trained on headache diary corpora attain migraine detection sensitivity of 86–94%. Multimodal wearable approaches combining photoplethysmography and actigraphy reach episode-level classification accuracy of 76–88%. These figures carry particular relevance for neurology outpatient practice in Uzbekistan, where headache complaints account for roughly one quarter of referrals and access to specialist headache clinics remains limited to the largest urban centres.
References
This work is licensed under a Creative Commons Attribution 4.0 International License.