To search, Click below search items.


All Published Papers Search Service


Diagnosis of Headache using Artificial Neural Networks


Karina Borges Mendes, Ronald Moura Fiuza, Maria Teresinha Arns Steiner


Vol. 10  No. 7  pp. 172-178


In this paper, we evaluate the use of artificial neural networks in order to predict the diagnosis of headaches. We intend to create an auxiliary tool for headache diagnosis to be used in primary level of Health Care System. It could enhance access of patients to headache diagnosis and treatment. Data were collected from a structured questionnaire which was applied to 2,177 patients with headache, in Neurological Clinic of Joinville, Santa Catarina, Brazil, from 2002 to 2006. Artificial neural networks were trained to reach the diagnosis performed by a single neurologist, who uses the criteria of the International Classification of Headaches Disorders. We tested them varying their activation function, initial weights and number of elements in input, hidden and output layers. We verified the feasibility of their use to predict the diagnosis of migraine without aura, migraine with aura, tension type headache and medication-overuse headache. The best results were obtained using binary coordinates as input vectors (information of the questionnaire) and one single neuron the output (diagnosis). Sensitivity and specificity of artificial neural networks were respectively 0.93 and 0.91 for tension-type headache identification, 0.99 and 0.94 for migraine without aura, 1.0 and 0.98 for migraine with aura and 1.0 and 0.96 for medication-overuse headache. In this way, we conclude that artificial neural networks can be used as a tool to support the diagnosis of common forms of headache and can therefore cooperate for greater access to the health system.


Artificial Neural Network, Pattern Recognition, Primary Health Care, Headache