Feature extraction by quick reduction algorithm: Assessing the neurovascular pattern of migraine sufferers from NIRS signals

Abstract

A migraine is a neurological disorder that can be caused by many factors, including genetic mutations, life-style, cardiac defects, endocrine pathologies, and neurovascular impairments. In addition to these health problems, an association between some types of migraines and increased cardiovascular risk has emerged in the past 10 years. Moreover, researchers have demonstrated an association between migraines and impaired cerebrovascular reactivity. It is possible to observe carbon dioxide dysregulation in some migraineurs, while others show a markedly decreased vasomotor reactivity to external stimuli. Therefore, the assessment of the cerebrovascular pattern of migraineurs is important both for the onset of a personalized therapy and for follow-up care. Near-infrared spectroscopy is a widely used tool for the non-invasive monitoring of brain oxygenation. It can be used to track hemodynamic changes during external stimulation (i.e. vaso-active maneuvers such as hypercapnia or hyperventilation). Unfortunately, near-infrared spectroscopy (NIRS) signals acquired during vaso-active maneuvers are non-stationary and require a time–frequency processing approach. To fully describe the cerebrovascular patterns of migraineurs, we extracted several parameters from the NIRS signals. Using these parameters, we compiled a dataset in which complexity was very high and the clinical/physiological information was impossible to track.