When spontaneous breathing is no longer sufficient to maintain alveolar ventilation and subsequent gas exchanges, noninvasive mechanical ventilation may be applied to reduce the ventilatory work of breathing. It has become a standard procedure to relieve patients with acute or chronic respiratory failure. Noninvasive ventilation is commonly applied during nocturnal session. Consequently, patient-ventilator interactions may affect the quality of sleep. Noninvasive ventilation is performed using two-level of pressure ventilator, the higher level corresponding to the positive inspiratory pressure (PIP) and the lower lever to the positive expiratory pressure (PEP).
In routine measurements, polysomnographies in the sleep laboratory were performed using a pneumotachograph proximal to the patient’s interface, allowing to record airflow and proximal airway pressures. Tracheal sounds via a suprasternal microphone, thoracic and abdominal motion, actimetry, SpaO2 and heart rate where also measured. In this study, the analysis was mainly based in nonvasive measurements, namely the airflow and the airway pressure.
In previous studies, we distinguished four groups of patient-ventilator interactions [1] that can be defined as follows :
Optimal mechanical interactions (no asynchronies, no non intentional leak) (Group O) ;
No asynchronies but with siginificant non intentional leaks
(Group OL) ;
With a significant rate of asynchrony events but without non intentional leak
(Group A) ;
With a significant rate of asynchrony events and with non intentional leak (Group AL).
Experimental data For each patient, the data file contains three columns, the first one correspond to the hypnogram scored (by 30 s windows) from 1 to 6 (1 REM Sleep, 2=Stage 4, 3=Stage 3, 4=Stage 2, 5=Stage 1 and 6=Awake). The two other columns correspond to the two EEG we recorded with a sampling rate at 64 Hz. According the International 10–20 system, these two EEG correspond to C3-A2 and C4-A1 on the head, respectively. A high pass filter at 0.5 Hz was applied to the data. A review on technique from nonlinear dynamical systems theory to automatically score EEG [2].
[1] H. Rabarimanantsoa-Jamous, Synchronization in mechanical ventilation with patients presenting chronic respiratory failure, Ph’D Thesis, University of Rouen, 2008 - Rina’s thesis
[2] K. Susmakova & A. Krakovska, Discrimination ability of individual measures used in sleep stages classification, Artificial Intelligence in Medicine, 44 (3), 261-277, 2008 - Full text.