Analysis of large array surface myoelectric potentials for the low back muscles

An algorithm was developed and tested for the ability to differentiate between the spatial distribution of large arrays of acute and normal recordings of surface electromyographic (EMG) data from subjects with and without low back pain (LBP). The surface EMG data from 62 channels were statistically analyzed and the spatial distribution of the root mean square (RMS) values were used in a multivariate quadratic discriminant model to classify the healthy and acute LBP subjects. The surface EMG distribution from the low back of 161 healthy and 44 acute LBP subjects were collected in three minimum stress postural positions including standing, 20 degrees of lumbar flexion and standing with arms extended forward holding 1.36 kg (3 lb) of weight in each hand. The best results obtained from the 'flexion' group of experiments correctly reclassified 95.5% (42 of 44) of the acute subjects and 99.4% (160 of 161) of the healthy. The success rate of this reclassification were found to be superior to reported patient classifications based on smaller set of electrode pairs using fewer subjects. The results indicated a potential of the model for clinical patient classification.