Objective: We aimed to compare the invasive (iICP) and a non-invasive intracranial pressure (nICP) monitoring methods in patients with traumatic brain injury, based on the similarities of the signals' power spectral densities.
Materials and methods: We recorded the intracranial pressure of seven patients with traumatic brain injury admitted to Hospital São João, Portugal, using two different methods: a standard intraparenchymal (iICP) and a new nICP method based on mechanical extensometers. The similarity between the two monitoring signals was inferred from the Euclidean distance between the non-linear projection in a lower dimensional space (ISOMAP) of the windowed power spectral densities of the respective signals. About 337 h of acquisitions were used out of a total of 608 h. The only data exclusion criterion was the absence of any of the signals of interest.
Results: The averaged distance between iICP and nICP, and between arterial blood pressure (ABP) and nICP projections in the embedded space are statistically different for all seven patients analysed (Mann-Whitney U, p < 0.05).
Conclusions: The similarity between the iICP and nICP monitoring methods was higher than the similarity between the nICP and the recordings of the radial ABP for all seven patients. Despite the possible differences between the shape of the ABP waveform at radial and parietal arteries, the results indicate-based on the similarities of iICP and nICP as functions of time-that the nICP method can be applied as an alternative method for ICP monitoring.
Keywords: Arterial blood pressure; Dimensionality reduction; Fourier transform; Intracranial pressure; Non-invasive.