Machine learning-driven microwave system for noninvasive monitoring of intracranial pressure

Intracranial pressure (ICP) refers to the pressure inside the skull. It is influenced by the complex interactions between the volume of brain tissue, cerebrospinal fluid (CSF), and blood. Maintaining a normal ICP is crucial for normal brain function, as elevated ICP can restrict blood flow to the brain, potentially resulting in severe health issues. Because of this, there is significant interest in non-invasive methods for monitoring ICP. In this paper, a Machine Learning (ML) driven, non-invasive, and quantitative microwave method and setup are proposed for ICP monitoring in human subjects. The proposed method is independent of the type of microwave sensors and is carefully devised for accurate measurements based on two-level feature extraction, including advanced signal attributes. Six thin, small, lightweight microwave sensors are evaluated with different placement strategies for accurate ICP monitoring. The proposed method was tested on a realistic human phantom model developed exclusively for this study. The phantom model corresponds to the dielectric properties and hydrodynamics of a human head. A unique data set creation module and Ordered Selection Scheme (OSS) are also proposed to ensure real-time operation with a lightweight ML algorithm. In addition, the quantitative method is devised using weighted regression on signal attributes selected from OSS. It is deduced from numerous trials that the proposed microwave system can even detect minute changes in ICP, and its response is analogous to pressure values measured by invasive sensors used as a ground-truth device. The proposed microwave-based setup is potentially suitable for wearable applications, enabling safe and prolonged usage.