Heart rate variability: An index of brain processing in vegetative state? An artificial intelligence, data mining study

ObjectiveBrain processing at varying levels of functional complexity has been documented in vegetative state. In this study, data mining procedures are applied to identify significant changes in heart rate variability (an emerging objective descriptor of autonomic correlates of brain activation) in response to complex auditory stimuli with emotional value (music).MethodsThe heart rate of subjects in vegetative state from brain damage (n = 6) or spontaneous hemorrhage (n = 3) and 16 healthy controls was recorded while they passively listened to four pre-selected music samples by different authors (mean recording time: 3 m and 36 s ± 24 s). The parametric and non-parametric frequency spectra were computed on the heart rate, spectra were compared within/across subjects and music authors, and the spectra descriptors were entered into a 1-R rules data mining procedure (WEKA software Leave One Out and Ten Fold Cross validation). The procedure independently classified the heart rate spectral patterns of both patients and controls and the emotions reported by healthy subjects as “positive” or “negative”.ResultsIn both healthy controls and vegetative state subjects, the power spectra while passively listening to music differed from baseline when compared irrespective of the music authorship and from each other when compared across music samples. Data mining sorted the nu_LF (normalized parameter unit of the spectrum low frequency range) as the significant descriptor of heart rate variability in the conditions of the study. The nu_LF classification of the healthy controls’ HRV changes in response to music replicated that based on subjective reports with 75–93.7% accuracy.ConclusionsAlthough preliminary, these findings suggest that autonomic changes with possible emotional value can be induced by complex stimuli also in vegetative state, with implications on the residual responsiveness of these subjects.SignificanceHeart rate variability descriptors and data mining methods appear applicable to investigate brain function in the absence of consciousness.