COPD severity classification using principal component and cluster analysis on HRV parameters

The application of principal component analysis and cluster analysis (PCA-CA) using heart rate variability (HRV) parameters to identify the most severe chronic obstructive pulmonary disease (COPD) subjects in a mixture of normal and COPD population is discussed. These parameters were obtained from real physiological data and cross-spectral analysis (i.e. the coherence and partial coherence between heart rate, blood pressure and respiration signals). Results demonstrated that these two groups could be differentiated with greater than 99.0% accuracy. Furthermore, differences on the same HRV parameters between all four severity levels of COPD subjects were also investigated. These groups were differentiated with over 88.0% accuracy. In analyzing the studied data of the COPD population, the technique correctly characterized 8.5% of COPD group as severe COPD. It was concluded that the PCA-CA technique identified the combination of parameters that can classify disease severity (COPD) as well as differences between normal and COPD subjects in a mixed population. The PCA-CA technique could perhaps also be used to classify other diseases non-invasively.