To create an algorithm to detect rheoencephalographic (REG) pulse wave second peak
increase, which may detect an increase in intracranial pressure (ICP). REG was measured
in 19 healthy volunteers during control and a 15-degree head-down tilt (HDT), which
caused an increase in ICP. We developed an algorithm for automatically calculating
P1 and P2 from the REG pulse waveform. The result was compared to manual measurements
during control and HDT positions. The automatic determination of the peaks’ time instants
was considered a series of two-class decision problems in ±15-ms-wide sliding decision
windows. We achieved an accuracy of 0.9826, a sensitivity of 0.7727, and a specificity
of 0.9902. We used the correlation coefficient between manual measurements and automated
data and the normalized mean absolute error (NMAE) metric to characterize the precision
of peak amplitude value estimation. We achieved a high correlation (> 0.8) in 92%
of all tests, and NMAE < 0.3 in 96%. The remaining cases were analyzed using Bland-Altman
plots to uncover the main causes of differences. We tested the hypothesis of increasing
P2 peak amplitude during HDT compared to the control position. In the female group,
the number of significant increases was 4 out of 6 (67%), both in manual and automatic
measurements. In the case of the male group’s manual measurements, 12 of 13 (92%),
and the automatic calculation gave 10 of 13 (77%). The significance is to switch from
invasive ICP to noninvasive REG to have the same information for decision-making at
the bedside to save the lives of neurocritical care patients. Future REG correlation
studies suggested using REG in neurocritical care monitoring, space research, and
military medical practice.
