Estimating intracranial pressure with optically-measured cerebral blood flow cardiac waveforms in pigs

Optical measurements of pulsatile heart-beat fluctuations in cerebral blood flow (CBF) have shown promise for non-invasive estimation of intracranial pressure (ICP). Herein, we trained distinct random forest models on concurrent absolute and normalized CBF waveform measurements in a pig model of hydrocephalus (n=18). The accuracy of both models was assessed against invasive ICP measurements. The use of absolute CBF waveforms resulted in the best agreement between estimated and invasive ICP, but the limits of agreement of ±20 mmHg were large. In future work, improved filtering of noise, enlarged training dataset, and the incorporation of additional features may improve accuracy.