Toward prediction of physiological state signals in sleep apnea

摘要: A recurrent connectionist model is described to predict dynamic respiratory state in the apneic sleeping patient. The time-domain of nonlinear time-lagged interactions between heart rate, respiration, and oxygen saturation was developed implicitly embed dynamics respiration cardiovascular control systems. Multiple future time scales were enforced on network during training explore limits prediction horizon produce a global representation trajectory. Predicted results are presented terms invariant geometric statistics (largest Lyapunov exponent /spl lambda//sub L/ correlation dimension D/sub c/). error 13%, while c/ within 9% true series value. magnitude these errors may fall experimental noise levels. This methodology eventually be useful continuous positive airway pressure (CPAP) therapy devices, lead increased patient compliance with this therapy.