Imaging hemorrhagic stroke with magnetic induction tomography: realistic simulation and evaluation.

Magnetic induction tomography (MIT) is a noncontact method for detecting the internal conductivity distribution of an object. This technology has the potential to be used in the biomedical area to check bio-impedance change inside the human body, for example to detect hemorrhage in the human brain. In this study the hemorrhagic stroke detectability with a 16-channel MIT system operating at 10 MHz was evaluated. Since the conductivity distribution is changed by the hemorrhagic stroke as well as the squeezed brain tissue around the stroke, deformation of the brain tissue is also considered and simulated with the help of a FEM-based linear bio-mechanical model in this paper. To simulate the raw measurement data as realistically as possible, the noise estimated from the experimental MIT system with hypothesis testing methods at 95% confidence level is added to the simulated measurements. Stroke images of 600 noisy samples for each detection assignment are reconstructed by the one-step Tikhonov-regularized inverse eddy current solution. Under the statistical framework, the detection failure is in control of a high false negative rate which represents a large artifact visualized in the reconstruction domain. The qualitative detectability of 18 detecting assignments, with three hemorrhagic positions (shallow, medial and center of the cerebrum) and two volume values (10 ml and 20 ml), overlaid by noise with three levels (standard deviation of phase change at 5 × 10−3°, 2.5 × 10−3°, 10 × 10−3°), are investigated. These detecting assignments are compared with each other to find out which volumes of deformed spherical hemorrhagic stroke can be detected by the modeled MIT system.