Naive Bayesian learning for small training samples: Application on chronic Low Back Pain diagnostic with sEMG sensors

This paper presents a method to classify chronic Low Back Pain subject by analyzing the muscular fatigue. A new signal characteristic is introduced: the spatial distribution of the Median Frequency slope. Because of the high number of sensors relative to the number of subjects tested, the classification method used is the Naive Bayesian classifier. The low back muscular fatigue is measured by the use of a matrix of 60 surface electromyography sensors. A total of 65 subjects comprising 43 with chronic low back pain condition and 22 healthy have perform a Sorenson test to produce fatigue on lumbar erector spinae muscles. A success rate of almost 70% cross-validated by a leave-one-out method is reported. The statistical significance of this success rate is evaluated by a permutation method.