A Deep Learning Framework for Deriving Non-Invasive Intracranial Pressure Waveforms from Transcranial Doppler.
Murad MegjhaniKalijah TerilliBennett WeinermanDaniel NametzSoon Bin KwonAngela VelazquezShivani GhoshalDavid J RohSachin AgarwalE Sander ConnollyJan ClaassenSoojin ParkPublished in: Annals of neurology (2023)
Increased intracranial pressure (ICP) causes disability and mortality in the neuro-intensive care population. Current methods for monitoring ICP are invasive. We designed a deep learning framework using a domain adversarial neural network to estimate noninvasive ICP, from blood pressure, electrocardiogram, and cerebral blood flow velocity. Our model had a mean of median absolute error (MAE) of 3.88+/-3.26 mmHg for the Domain adversarial neural network and 3.94+/-1.71 mmHg for the Domain Adversarial Transformers. Compared to non-linear approaches such as support vector regression, this was 26.7% and 25.7% lower. Our proposed framework provides more accurate non-invasive ICP estimates than currently available. This article is protected by copyright. All rights reserved.