Frequency-Difference Electrical Impedance Tomography for Stroke Monitoring: Effects of Model Accuracy and Reconstruction Methods

Abstract

Stroke covers a number of critical abnormalities in cerebral hemodynamics that require prompt diagnosis and continuous monitoring to improve patient outcomes. Electrical impedance tomography (EIT) is an imaging technology that enables dynamic cerebral monitoring through electrical conductivity contrast. Traditionally, monitoring with EIT relies on time-difference measurements (hence, tdEIT), requiring prior measurement data from a healthy state, which is typically unavailable in clinical practice. Frequency-difference EIT (fdEIT) offers an alternative approach that eliminates this requirement. However, fdEIT presents greater challenges in image reconstruction because conductivity variations with frequency change occur for all head tissues, making reconstruction model accuracy more critical than in tdEIT. In this study, the effects of model accuracy, conductivity change restrictions, and noise tolerance are investigated for three fdEIT reconstruction methods. Simulations of intra-cerebral hemorrhage reveal that, unlike tdEIT, accurate model geometry and the inclusion of cerebrospinal fluid are necessary, and restricting conductivity change in the skull region is highly recommended. When accurate head geometry is available, the reference-weighted fdEIT method demonstrates markedly superior performance compared with standard fdEIT, and comparable to an ideal tdEIT case. On the other hand, when accurate head geometry is not available, the standard fdEIT method is recommended, though with limited performance. Regardless of the reconstruction method, ensuring a measurement signal-to-noise ratio of at least 50 dB is essential for reliable imaging.