The utility of neural network in the diagnosis of Cheyne-Stokes respiration

The aim of this study was to design a diagnostic model to identify patients with Cheyne-Stokes respiration (CSR-CSA) based on indices of oximetric spectral analysis. A retrospective analysis of oximetric recordings of 213 sleep studies conducted over a one-year period at a Veterans Affairs medical facility was performed. A probabilistic neural network (PNN) was developed from salient features of the oximetric spectral analysis, desaturation events and the delta index. A fivefold cross-validation was used to assess the accuracy of the neural network in identifying CSR-CSA. When compared to overnight polysomnography, the PNN achieved a sensitivity of 100% (95% confidence interval [CI] 85%-100%) and a specificity of 99% (95% 97%-100%) with a corresponding area under the curve of 99% (95% CI 99%-100%). When combined with overnight pulse oximetry, PNN offers an accurate and easily applicable tool to detect CSR-CSA.