AI in Neuroscience: Development of Methods to make Personalized Predictions for Migraine and Stroke from E-Health Sensor Data
The research of this PhD project can be subdivided into two main disease areas: migraine and stroke. For both we will be investigating how artificial intelligence (AI) and machine learning (ML) techniques can be used to study these afflictions, their (early) detection, and their potential treatment.
- Hermes Spaink
Migraine is a prevalent, multifactorial brain disease, characterized by recurrent attacks of invalidating headaches with or without aura. Although the pathophysiology is relatively well studied, less is known about specific triggers that lead to migraine attacks. Many patients indicate weather to be a possible trigger for their attacks. We will investigate the specific role of weather conditions in migraine both on the group and the individual level and aim to build patient-specific predictive models. Next, we will extend the predictive models with other data types including touchscreen usage (tappigraphy) data. Studies have shown that smartphone usage behaviour may be a useful proxy for neurobehavioral changes (reflecting changes in brain activity) related to neurological and psychological states. We will use AI to find patterns in tappigraphy data of migraine patients and associate these to migraine onset using headache diaries. We will investigate various models and configurations including explanatory models to detect the underlying relevant pat-terns and factors in the data.
Stroke is caused when blood supply to part of your brain is interrupted or reduced, preventing brain tissue from getting oxygen and nutrients. Of note, stroke and migraine are comorbid disorders with to certain extent overlapping disease mechanisms (foremost with respect to the occurrence of spreading depolarizations). With respect to stroke, the research for this PhD project will be around “The Stroke Box”. Although there have been major advances in (personalized) stroke treatment, the need for primary stroke prevention remains essential. For stroke, atrial fibrillation (AF) and hypertension are important risk markers, but AF is often missed in the cardiovascular risk management of the general practitioner. Recent studies have shown telemonitoring of patients can greatly improve healthcare and reduce costs. Continuing on the existing ‘box framework’ already applied in various areas of the LUMC, we will adapt the box so it can be used for patients who suffered from a stroke. The stroke box will hence serve as a patient-centred eHealth approach for improving post-stroke care by earlier detection, prediction and prevention of recurrent strokes.