Wimer, Jeremiah R.; Westman-Forbes, Amaya K.; Wu, Jingxian; Sexton, Kevin W.; Jensen, Hanna K.; Saunders, Robert; Jensen, Morten O. (2026).听.听Biomedical Signal Processing and Control, 126, 110985.听
Accurately assessing whether a patient has enough circulating blood and body fluid is especially important in children, but existing methods can be invasive or difficult to perform. This pilot study explored whether聽peripheral venous pressure (PVP) waveforms鈥攕ignals recorded from a vein鈥攃ould be used to detect a patient’s聽intravascular volume status聽(the amount of fluid circulating in the blood vessels). The researchers analyzed PVP waveforms from 18 pediatric patients and used a modified聽integral pulse frequency modulation (IPFM)聽model to separate meaningful physiological signals, such as heartbeats, heart rate variability, and breathing patterns, from noise. These processed signals were then used to train machine learning models to classify patients as聽resuscitated聽(adequately restored fluid levels) or聽hypovolemic聽(having abnormally low blood volume). Models trained using the IPFM-extracted heartbeat signal achieved nearly 100% testing accuracy, outperforming models trained on the processed PVP waveforms (95.8%) or the original PVP data (90.1%). The findings suggest that the IPFM approach can improve the detection of dehydration-related changes by removing information that is not relevant to volume status. However, because the study included only a small number of patients, larger studies are needed to determine whether the method performs as well in broader clinical populations. The authors also note that because dehydration and blood loss produce similar changes in PVP waveforms, this approach may also be useful for detecting hemorrhage.
Fig. 2.听Filtered PVP signal聽聽and its slow-changing moving average聽.