Noninvasive Intracranial Pressure Estimation Using Subspace System Identification and Machine Learning Algorithms: A...

Area of Science:

• Biomedical Engineering
• Critical Care Medicine
• Machine Learning

Background:

• Accurate noninvasive estimation of intracranial pressure (ICP) is critical in managing patients with acute brain injury.
• Current methods for ICP monitoring are invasive, posing risks and limitations in clinical settings.

Purpose of the Study:

• To develop and evaluate a novel machine learning algorithm for noninvasive estimation of mean intracranial pressure (ICP).
• To integrate system identification and ranking-constrained optimization for enhanced ICP estimation accuracy.

Main Methods:

• A machine learning framework utilizing subspace system identification to model cerebral hemodynamics.
• Integration of arterial blood pressure (ABP), cerebral blood velocity (CBv), and R-wave to R-wave interval (R-R interval) signals for ICP simulation.
• Convex optimization with ranking constraints to learn a mapping function for estimating ICP from noninvasive signal features.

Main Results:

• The proposed algorithm demonstrated feasibility in estimating mean ICP from noninvasive signals.
• Approximately 31.88% of testing data achieved estimation errors within 2 mmHg.
• An additional 34.07% of testing data had estimation errors between 2 mmHg and 6 mmHg.

Conclusions:

• The study presents a proof-of-concept for noninvasive ICP estimation in patients with acute brain injury.
• Further development, larger datasets, and prospective validation are needed to improve accuracy and clinical utility.