Impact of sampling rate for time domain analysis of continuous intracranial pressure (ICP) signals

Abstract

Time domain analysis of the intracranial pressure (ICP) waveform is critically dependent on the exact reproduction of the ICP waveform. This study explored how the sampling rate of the ICP signal affects the time domain analysis. It was also assessed through this study how upsampling (interpolation) improves the time domain analysis. From the hospital database, a set of 55 ICP waveforms were retrieved from 48 patients (28 children and 20 adults). First, the ICP signals originally sampled at 200 or 100 Hz were compared with the ICP signals downsampled to 5, 10, 20, 25, 50 Hz (and 100 Hz). Second, the original ICP signals were compared with ICP signals upsampled (i.e. interpolated) to 100 Hz (from 5, 10, 20, 25 or 50 Hz). For each ICP recording the output of time domain analysis was the average value and the quantitative distribution of mean ICP wave amplitudes determined every six second (6 s) time window. The total material incorporated a total of 373,371 6 s time windows. Downsampling revealed that the time domain analysis could be most faithfully applied to ICP signals sampled at 50 Hz or above, while ICP signals sampled at 25 Hz deviated more from the original signal than we would accept from a clinical perspective. The use of interpolation gave better representation of the peaks, and should be applied to all ICP signals sampled at lower rate than 100 Hz.