Non-Invasive Measurement of Intracranial Pressure? Accurate, Finds Johns Hopkins University Team

Disadvantages of the Invasive Procedure

Elevated ICP can be life-threatening, and accurate monitoring is crucial for patient prognosis. Standard procedures, however, require placing sensors in close contact with brain tissue or even within the brain ventricles. These carry numerous risks.

For example, external ventricular drainage (EVD) risks include catheter misplacement, which occurs in approximately 15% of cases, the danger of infection of the brain membranes or brain (about 6% of cases), and bleeding (in 12% of cases). Surgical expertise and specialized equipment, which are not always available, are required.

ICP Relationship to Physiological Waveforms

The scientists tested the hypothesis that severe brain injury, and especially elevations in ICP, are accompanied by pathological changes in systemic cardiocirculatory function (for example, due to dysregulation in the autonomic nervous system), and that extracranial physiological processes can be studied to better understand brain activity and ICP severity.

In the study described below, they explored the relationship between the ICP waveform and three physiological waveforms routinely measured in intensive care units: arterial blood pressure (ABP), photoplethysmography (PPG), and electrocardiography (ECG). The researchers assumed that by using deep learning methods, they could calculate, and thus obtain non-invasively, ICP waveforms.

For the purposes of the study, data in the MIMIC-III database were anonymized. They evaluated 600 hours of simultaneously obtained data on ICP, ABP, ECG, and PPG progressions from 10 patients admitted to the ICU with critical brain disorders. The data were divided into non-overlapping 10-second windows, and they attempted to recreate the ICP curve from ABP, ECG, and PPG curves using deep learning (DL) methods. They assessed the predictive performance of six different DL models (RNN, GRU, LSTM, TCN, VNET, Transformer), both in iterations with one and multiple patients.

Safer, Gentler, More Accessible

A key finding is that the best models predicted ICP with an average error (MAE) of approximately 5 mmHg and limited bias. The researchers found that the PPG signal was important for predicting ICP. The MAE (± standard deviation [SD]) of the best-performing models was 1.34 (± 0.59) mmHg (iteration with one patient), respectively, 5.10 (± 0.11) mmHg (iteration with multiple patients).

The study authors stated that overall, the results supported their hypothesis that information about the course of the ICP curve is encoded in the shapes of the ECG, PPG, and ABP curves.

"This research provides compelling evidence that it is possible to accurately calculate a surrogate ICP waveform from extracranial physiological data. With further validation, our algorithms, which allow real-time monitoring and are non-invasive, could be integrated into the clinical workflow for monitoring ICP, representing a significant advance over current invasive approaches," they stated.

Preliminary results support the feasibility and accuracy of the used procedure, being more accurate than most other non-invasive techniques and on par with invasive techniques. With refinement and further validation, this method could represent a safer, gentler, and more accessible alternative to invasive ICP measurement, enabling rapid initiation of treatment for severe brain injuries, additionally with lower resource demands.

Editorial Team, Medscope.pro

Sources:

1. Nair S. S., Guo A., Boen J. et al. A deep learning approach for generating intracranial pressure waveforms from extracranial signals routinely measured in the intensive care unit. Comput Biol Med 2024 Jul; 177: 108677, doi: 10.1016/j.compbiomed.2024.108677.
2. New research identifies less invasive method for examining brain activity following traumatic brain injury. Johns Hopkins Medicine, 2024 Jul 23. Available at: www.hopkinsmedicine.org/news/newsroom/news-releases/2024/07/new-research-identifies-less-invasive-method-for-examining-brain-activity-following-traumatic-brain-injury