Rok Istenic

Multiscale entropy-based approach to automated surface EMG classification of neuromuscular disorders

We introduce a novel method for an automatic classification of subjects to those with or without neuromuscular disorders. This method is based on multiscale entropy of recorded surface electromyograms (sEMGs) and support vector classification. The method was evaluated on a single-channel experimental sEMGs recorded from biceps brachii muscle of nine healthy subjects, nine subjects with muscular and nine subjects with neuronal disorders, at 10%, 30%, 50%, 70% and 100% of maximal voluntary contraction force. Leave-one-out cross-validation was performed, deploying binary (healthy/patient) and three-class classification (healthy/myopathic/neuropathic). In the case of binary classification, subjects were distinguished with 81.5% accuracy (77.8% sensitivity at 83.3% specificity). At three-class classification, the accuracy decreased to 70.4% (myopathies were recognized with a sensitivity of 55.6% at specificity 88.9%, neuropathies with a sensitivity of 66.7% at specificity 83.3%). The proposed method is suitable for fast and non-invasive discrimination of healthy and neuromuscular patient groups, but it fails to recognize the type of pathology.

EMG Based Muscle Force Estimation using Motor Unit Twitch Model and Convolution Kernel Compensation

In this paper we introduce a new method for muscle force estimation from multi-channel surface electromyograms. The method combines a motor unit twitch model with motor unit innervation pulse trains, which are estimated from multi-channel surface electromyograms. The motor unit twitches are then aligned to the innervation pulse trains and summed up to obtain the total muscle force. The method was tested on real surface EMG signals acquired during force ramp contractions of abductor pollicis brevis muscle in 8 male subjects. With 22 ± 5 (mean ± std. dev.) motor units identified per subject, the force estimation error of our method was 16 ± 4 % RMS. These results were compared to the method which uses the EMG amplitude processing to estimate muscle force. The results of our new concept proved to be completely comparable to those of EMG amplitude processing.

Estimation of the Number of Signal Sources in Compound Signals Using Activity Index Variance

This work addresses the problem of estimating the number of signal sources in convolutive signal mixtures. Our approach is based on the blind source separation. The first step builds a measure of the global activity of all active sources. This measure is known as activity index, and is based on the inverse of correlation matrix of the observed signals. The next step analyses the activity index variance. Our experiments have shown that the number of sources can be estimated from the activity index variance. A set of multi-channel synthetic signal mixtures with different number of active sources was generated for the evaluation. The preliminary results using synthetic signals show that our approach is robust and effectively estimates the number of signal sources.

Surface EMG pre-processing techniques for the detection of common input to motor neuron populations

In this paper we compared four surface EMG preprocessing techniques to improve the detection of common input to two motor neuron populations. We proposed multichannel approach named Activity index and its improvement higherorder Activity index. Both methods were compared to raw and rectified EMG. Techniques were evaluated on simulated EMG signals of two motor neuron populations and EMG-EMG coherence was used as a measure. Higher-order Activity index performed better than original Activity index, coherence values were in the range of rectified EMG.

On detection of pathological tremor in electroencephalograms

This study examines the feasibility of online detection of tremor-related component in noninvasively acquired multichannel electroencephalographic (EEG) signals. In particular, performances of different feature extraction techniques, ranging from time-frequency and time-scale analysis to blind source separation of EEG signals are mutually compared and their suitability for online tremor detection in EEG discussed. The results on EEG signals from five tremor-affected patients demonstrate that, under constraint of high frequency resolution, the time-frequency analysis combined with support vector machine classifiers offers acceptable accuracy in tremor detections (sensitivity > 90 % at specificity of 90 % or above). The other tested approaches either fail to reliably identify the presence of the tremor-related EEG component or suffer from large inter-subject and/or inter-trail variability.

The influence of source overlapping on correlation matrix and activity index

In this paper, the influence of overlapping of pulse signal sources on their correlation matrix and activity index is studied. The activity index is defined as a Mahalanobis distance of signal observations. Influences of source overlapping on activity index were simulated for a different number of overlapping sources and degrees of their overlapping. The findings lead to an improved model of activity index, which can support a more reliable estimation of the number of active sources in convolutive mixtures of pulse sources.