Abstract
Introduction: It has previously been reported that EMG variables recorded from the lumbar spinal muscles may be recorded reproducibly, are able to discriminate low back pain subjects from normal volunteers and are predictive of future back pain. At present, however, an experienced operator is required to acquire the signals and to determine the value of some variables. This has hindered the transfer of the technique from the laboratory to the clinical setting.
Methods: The EMG signal is subjected to a Fast Fourier Transform and a power spectrum is produced. An Expert System has been developed to examine this power spectrum. In accordance to a rule base several variables are generated including the half width. The error analysis can detect a number of possible errors of recording that can affect test results and unusual traces are flagged for further consideration. In some defined cases a correction is automatically applied.
Results: The Standard error between tshe manually generated half width and the automatically calculated value is 30%. Using the automated system 5% of subjects were found to change classification from normal to at risk. The sensitivity and specificity of detecting recording errors was 0.5 and 0.4 respectively. Work is ongoing.
Conclusions: The new system has reduced data set analysis from days to minutes, thus many different methods of analysis can be compared and contrasted readily. The automatic calculation of half width and other variables has brought clinical usage one step closer, and allow EMG analysis to provide a useful tool for monitoring treatment and measuring outcome.
Correspondence should be addressed to: Mr John O’ Dowd, SBPR, c/o BOA, The Royal College of Surgeons, 35–43 Lincoln’s Inn Fields, London WC2A 3PE.
