Fully Automated and AI-Assisted Optical Fiber Sensing System for Multiplexed and Continuous Brain Monitoring

Abstract

Continuous and comprehensive brain monitoring is crucial for timely identification of changes or deterioration in brain function, enabling prompt intervention and personalized treatments. However, existing brain monitoring systems struggle to offer continuous and accurate monitoring of multiple brain biomarkers simultaneously. This study introduces a multiplexed optical fiber sensing system for continuous and simultaneous monitoring of six cerebrospinal fluid (CSF) biomarkers using tip-functionalized optical fibers and computational algorithms. Optimized machine learning models are developed and integrated for real-time spectra analysis, allowing for precise and continuous readout of biomarker concentrations. The developed machine learning-assisted fiber optic sensing system exhibits high sensitivity (0.04, 0.38, 0.67, 2.62, 0.0064, 0.33 I/I₀ change per units of temperature, dissolved oxygen, glucose, pH, Na⁺, Ca²⁺, respectively), reversibility, and selectivity toward target biomarkers with a total diameter less than 2.5 mm. By monitoring brain metabolic and ionic dynamics, this system accurately identified brain physiology deterioration and recovery using ex vivo traumatic brain injury models. Additionally, the system successfully tracked biomarker fluctuations in clinical CSF samples with high accuracy (R² > 0.93), demonstrating excellent sensitivity and selectivity in reflecting disease progression in real time. These findings underscore the enormous potential of automated and multiplexed optical fiber sensing systems for intraoperative and postoperative monitoring of brain physiologies.