Distinguishing the strength of hypoxic stimulus in intermittent hypoxia

The intermittent hypoxia (IH) phenomenon is a subject of intensive examinations. In this study we examined whether it could be possible to distinguish the strength of the hypoxic stimulus given in IH cycles on the basis of observed changes in the breathing pattern. We investigated the ventilatory responses to five hypoxic-normoxic cycles (1 min hypoxia/3 min normoxia) in rabbits. Two different hypoxic stimuli were given: gas mixtures of 14% and 11% O2 in N2, each one in a separate run of IH. Ventilatory features: frequency (f), tidal volume (VT) and minute ventilation (VE) were analyzed using the algorithms of the pattern recognition theory. The probability of wrong classification was used as a criterion for the recognition quality evaluation. This probability can be estimated experimentally by calculating the percentage of misclassifications, i.e., an error rate (Er). When the features were analyzed alone, the VT offered the lowest misclassification rate of 19.3% and 10.3% in the stimulus and normoxic periods, respectively. However, using the single features measured during the stimulus and recovery phases allowed to decrease the error rate more than 2-fold, achieving 4.3% for VT. The best results were obtained for both phases associated with single cycles of IH. Suitable feature selection procedures enabled reducing the global misclassification rate to 0.7%. In conclusion, the pattern recognition approach may potentially be useful for controlling the stimulus strength in intermittent hypoxic training.