Development of a pressure control algorithm for treatment of sleep-disordered breathing

Abstract

Sleep-disordered breathing is a family of pathologies that are caused by a total or partial collapse of the upper airways. They are characterized by abnormal respiratory patterns and insufficient ventilation during sleep. Those patterns can be detected based on the mandibular movements measured by a distance sensor, the Jawac. Nowadays, positive airway pressure therapy is considered as the most effective one by providing a pneumatic splint and preventing the collapse.

This thesis aims at developing a CPAP pressure control algorithm and implementing it on a microcontroller. Clinical tests are then carried to assess the effectiveness of the algorithm. Improvements of the algorithm are then made after the analysis of the results of the tests.

The decisions of the control algorithm are based the Jawac signal only. The three criteria to which the algorithm responds are a large opening of the patient's mouth, respiratory effort characterized by oscillations at breathing frequency and salient jaw movements related to arousals. One test has shown that these rules are pertinent and that the system has good performances (i.e. no apnea remains). However, the major constraint was that the algorithm performs well only on sleeping patients. Therefore, a detection of the sleep and wake states, also based on the Jawac signal, is added to the control algorithm to ensure that the algorithm runs only when the patient is asleep. Another limitation of the system is its insufficient inertia. Hence, solutions to avoid increasing or decreasing pressure too fast are suggested. For the increasing part, a precedence is set to the opening rule. For the decreasing, several strategies are considered: using a dynamic lower bound on pressure to prevent from decreasing at specific moments, stretching the delay before decreasing pressure to slow down the process, and adding rules to keep the pressure constant on the top of the increasing and decreasing ones. All these solutions have been simulated and make sense from a physiological point of view.

The analysis of the Jawac signal in the frequency domain, using the wavelet transform, allows to detect regular oscillations and salient jaw movements. The results are similar as with the temporal analysis of the signal. Therefore, this approach is not interesting since it requires more computation power. A non-linear model of sleep apnea, where the pharynx is modelled as a collapsible tube, is also studied in order to better understand its dynamics. The main results are that sleep apnea can be classified in two categories: gradual and catastrophic (sudden) collapse and that apneas correspond to a fixed points of the model.