Mutual information analysis of spinal and supraspinal responses to graded electrical stimulations

F.G. ARGUISSAIN; J.A. BIURRUN MANRESA; C.D. MØRCH; O.K. ANDERSEN

San Diego

Conferencia; Neuroscience 2013; 2013

Aim: to determine the amount of information carried by single-trial response features extracted from electromyographic (EMG) and electroencephalographic (EEG) signals in response to nociceptive and non-nociceptive electrical stimulation.Methods: sixteen male subjects participated in a single experimental session, in which they received repeated electrical stimuli to the arch of the foot in order to elicit the nociceptive withdrawal reflex (NWR). Six stimulation intensities (0.5x, 0.75x, 1.0x, 1.25x, 1.5x and 2.0x) multiples of the NWR threshold (i.e. the minimum intensity necessary to elicit the reflex) where applied. NWR features (root-mean-squared amplitude, interval peak z-score, latency and duration) were quantified from EMG signals recorded from the tibialis anterior muscle. EEG features were defined as the amplitudes and latencies extracted from the characteristic peaks (N1, P1, N2 and P2) of the somatosensory-evoked potentials recorded at the vertex. The Information Theory framework was applied to quantify the amount of mutual information (MI) contained in these features with respect to the stimulus. MI is expressed in bits, where zero can be interpreted as a completely random stimulus-response relation and the theoretical maximal value is given by the entropy or self-information of the stimulus (2.58 bits in the present study). Mutual information quantities were calculated on an individual basis and for a subset of combinations of these features.Results: median MI values of all selected response features from both modalities were informative in regard to the stimulus when considered individually. Significant differences were found among the median MI values of the response features (Kruskall-Wallis test, P≤0.001). Specifically, the information carried by the EMG features (0.39 bits in average) was significantly higher than information contained in the EEG features (0.11 bits in average) (Dunn?s test, p0.05). A preliminary analysis showed that the combination of response features might be more informative than their individual values; however, an overall redundant effect was observed. In most of the combinations, the observed redundancy was small, which might indicate that the stimulus-related information carried by each feature is mainly independent from each other.Conclusion: these results provide evidence that supports the importance of assessing and interpreting both SEPs and NWR when studying the nociceptive system. The current study presents an alternative approach that allows the quantification of stimulus-signal and signal-signal relationships at single-trial level without linearity and gaussianity constraints.