Researchers develop new method for early prediction of crises in chronic diseases

Researchers from Universidad Politécnica de Madrid and Universidad Complutense de Madrid have developed a methodology for early prediction of crises in chronic diseases, such as migraines.

The study carried out by researchers from Center for Computational Simulation, at Universidad Politécnica de Madrid, and researchers from Universidad Complutense de Madrid manage to increase the time for early detection of migraines up to 40 minutes using non-intrusive wireless body sensor network. In most cases, this time is enough to anticipate the drug intake and thus preventing or lessening the pain effects. The developed methodology could be used for other chronic diseases.

Migraine is a neurological disease that affects around 15% of the European population and generates large costs to public and private health care systems. The prediction of this type of event will allow doctors to act and mitigate pain according to the pharmacokinetics of current treatments.

This pilot study aims to early detect the appearance of migraines using non-intrusive wireless body sensor network (WBSN). Thus, researchers used a commercial ambulatory monitoring device for controlling the biometric variables of skin surface temperature, sweating, heart rate and oxygen saturation. Preliminary studies showed the feasibility of these predictive modeling techniques in migraines. The researchers, who work along with the headache unit from Hospital Universitario de la Princesa de Madrid, have developed techniques for prediction improvement that increase up to 10 minutes the models obtained so far.

The use of sensor networks is increasingly frequent, but they can still have errors. The real ambulatory monitoring is subject to sensor losses, data failure, disconnections and so on. The researchers presented a robust model selection strategy based on the state of the sensors of the monitoring equipment and according to the desired criteria in terms of the prediction quality.

Researchers say "by using the methodology proposed, the prediction can be adjusted to a compromise between the conservative (quality prediction) and the daring (prioritizing the time of advance and increasing the uncertainty), all this depending on the feasibility of the sensor at every moment". The results of this pilot study suggest that these models could be adapted to the characteristics of each patient.

This new methodology could be applied to other chronic diseases with symptomatic crises in which a prediction of event would allow doctors to take decisions that mitigate their effects, for instance the ambulatory monitoring of patients admitted for stroke at risk of having another one.