Izvorni znanstveni članak

Effects of supervised practice on the accuracy of observers for manual segmentation of simulated electromyograms

Arthur de Sa Ferreira ; Laboratory of Computational Simulation and Modeling in Rehabilitation, Postgraduate Program of Rehabilitation Science, Centro Universitário Augusto Motta/UNISUAM, Rio de Janeiro, Brazil
Fernando Silva Guimaraes ; Laboratory of Cardiovascular and Respiratory Performance, Postgraduate Program of Rehabilitation Science, Centro Universitário Augusto Motta/UNISUAM, Rio de Janeiro, Brazil
Regina Coeli Souza e Silva ; Laboratory of Computational Simulation and Modeling in Rehabilitation, Postgraduate Program of Rehabilitation Science, Centro Universitário Augusto Motta/UNISUAM, Rio de Janeiro, Brazil
Manuel Armando Ribeiro Magalhaes ; Laboratory of Cardiovascular and Respiratory Performance, Postgraduate Program of Rehabilitation Science, Centro Universitário Augusto Motta/UNISUAM, Rio de Janeiro, Brazil

Sažetak

Visual interpretation of electromyograms is common, but its accuracy is unknown. This study compared the accuracy curves of inexperienced observers in detecting muscular contractions from variable, simulated
surface electromyogram signals. Accuracy was assessed both without feedback (unsupervised practice) and with feedback (supervised practice) to determine whether a training effect existed. Six observers performed manual segmentation in 300 simulated waveforms using a phenomenological model with a variable number of contractions (n=1, 2 or 3), smooth changes in amplitude, marked on-off timing, and a variable signal-to-noise ratio (0-39 dB). Segmentation was organized in two one-day sessions with 15 blocks of 20 signals each for the unsupervised and supervised practices, respectively. Supervised practice was provided by an immediate visual feedback on the manual segmentation. The accuracy curve showed no significant linear regressions for either unsupervised (R2=.104, p=.241) or supervised practices (R2=.153, p=.150). No significant difference in accuracy was observed between the unsupervised and supervised practices (85% [77; 99] and 88% [73; 97], respectively; p=.295). Unsupervised practice yielded low accuracy for one muscular contraction (AUC=.43;
cut-off=12.8 dB) and increased with supervised practice (AUC=.63; cut-off=9.5 dB). Unsupervised practice resulted in high accuracy for two contractions (AUC=.88; cut-off=6.9 dB) and was similar to the supervised practice (AUC=.81; cut-off=6.3 dB). Supervised practice using visual feedback improved the accuracy of inexperienced observers in the segmentation of one muscular contraction in simulated electromyograms and did not influence the accuracy of two muscular contractions.

Ključne riječi

electromyography; contraction detection; computer simulation; muscle activity

Hrčak ID:

131881

URI

https://hrcak.srce.hr/131881

Datum izdavanja:

30.12.2014.

Posjeta: 1.088 *