Colonic slow-wave analysis. Limitations of usefulness of fast Fourier transform (FFT)

The fast Fourier transform (FFT) has been used to determine frequency components of colonic slow-wave activity. We studied the effect of (1) recorder filter characteristics, (2) number of data points and, (3) data window overlap technique and ingestion of a 1000-kcal meal on the resulting power spectrum. Human rectosigmoid slow-wave activity was recorded in nine normal subjects and stored on FM tape for computer analysis. The dynograph filter characteristics were tested using square wave signals, and derived compensation factors were applied to the FFT before viewing. The dynograph filter, when set to optimize visualization of slow waves, attenuates low frequencies nonlinearly. Failure to compensate for the dynograph filter results in inaccurate detection of slow-wave frequencies. FFT of 1-min data gives a different power spectrum than an FFT of 4 min data, indicating a rapidly changing waveform. FFT's of 1 min of data when examined over time fail to demonstrate a consistent frequency spectrum, confirming this conclusion. The lower frequencies in the normal human rectosigmoid are present at the greatest power. These studies indicate that the colon has slow waves of irregular frequencies, in contrast to the stomach or small intestine. No change in the dominant frequency was seen following the ingestion of a 1000-kcal meal.