Intracranial hemorrhage alters scalp potential distribution in bioimpedance cerebral monitoring: Preliminary results from FEM simulation on a realistic head model and human subjects

Abstract

Purpose: Current diagnostic neuroimaging for detection of intracranial hemorrhage (ICH) is limited to fixed scanners requiring patient transport and extensive infrastructure support. ICH diagnosis would therefore benefit from a portable diagnostic technology, such as electrical bioimpedance (EBI). Through simulations and patient observation, the authors assessed the influence of unilateral ICH hematomas on quasisymmetric scalp potential distributions in order to establish the feasibility of EBI technology as a potential tool for early diagnosis.

Methods: Finite element method (FEM) simulations and experimental left-right hemispheric scalp potential differences of healthy and damaged brains were compared with respect to the asymmetry caused by ICH lesions on quasisymmetric scalp potential distributions. In numerical simulations, this asymmetry was measured at 25 kHz and visualized on the scalp as the normalized potential difference between the healthy and ICH damaged models. Proof-of-concept simulations were extended in a pilot study of experimental scalp potential measurements recorded between 0 and 50 kHz with the authors' custom-made bioimpedance spectrometer. Mean left-right scalp potential differences recorded from the frontal, central, and parietal brain regions of ten healthy control and six patients suffering from acute/subacute ICH were compared. The observed differences were measured at the 5% level of significance using the two-sample Welch t-test.

Results: The 3D-anatomically accurate FEM simulations showed that the normalized scalp potential difference between the damaged and healthy brain models is zero everywhere on the head surface, except in the vicinity of the lesion, where it can vary up to 5%. The authors' preliminary experimental results also confirmed that the left-right scalp potential difference in patients with ICH (e.g., 64 mV) is significantly larger than in healthy subjects (e.g., 20.8 mV; P < 0.05).

Conclusions: Realistic, proof-of-concept simulations confirmed that ICH affects quasisymmetric scalp potential distributions. Pilot clinical observations with the authors' custom-made bioimpedance spectrometer also showed higher left-right potential differences in the presence of ICH, similar to those of their simulations, that may help to distinguish healthy subjects from ICH patients. Although these pilot clinical observations are in agreement with the computer simulations, the small sample size of this study lacks statistical power to exclude the influence of other possible confounders such as age, sex, and electrode positioning. The agreement with previously published simulation-based and clinical results, however, suggests that EBI technology may be potentially useful for ICH detection.

FIG. 1.

Numerical head model showing the following different anatomical structures: (1) CSF ventricles, (2) white matter, (3) gray matter, (4) brain blood vessels, (5) cerebellum, (6) adipose, (7) mastoid bones, (8) spinal cord, (9) arteries, (10) CSF, (11) vertebral column, (12) skull, (13) bone facial, (14) nerves, (15) air sinus, (16) humor, (17) retina, (18) orbital fat, (19) nose, (20) ears, (21) subcutaneous tissue, (22) head muscle, and (23) skin (Ref. 38).

FIG. 2.

Diagram of the custom-made EIS system and the electrode positions for scalp potential recordings.

FIG. 3.

Circuit diagram of the transconductance amplifier used to deliver the white noise probing current.

FIG. 4.

LabVIEW program window offering real-time options: (top left) noise variance, (top right) raw voltage traces, (bottom left) resistance, and (bottom right) reactance.

FIG. 5.

(a) shows portable spectrometer and its size relative to a CT scanner at MGH in the background. (b) Spectrometer connector channels. (c) Example of a measurement in the MGH neurology department.

FIG. 6.

Schematic describing the left–right potential asymmetry analysis and expected observations for a healthy subject and a patient suffering from ICH: (a) shows the locations of current electrode pair delivering Gaussian white noise to the head as well as the scalp potential measurement sites. (b) shows expected Gaussian potential distribution at each electrode of the symmetrically paired electrodes on frontal (F8–F7), central (C4–C3), and parietal (P8–P7) areas of the scalp of the healthy and ICH damaged heads. (c) shows the absolute potential difference of the frontal, central, and parietal electrode pairs. (d) shows mean of the absolute potential difference at each electrode pair of healthy and patient as well as the electrode pair with maximum potential difference. It is expected that the left–right potential asymmetry caused by hematoma is larger than the asymmetry caused due to the anatomical asymmetries in the healthy head.

FIG. 7.

FEM numerical simulation solutions in the midsagittal plane of the healthy numerical head model after probing with a sinusoidal current of 140 μA at 25 kHz. (a) Faraday’s law of induction. (b) Gauss law. (c) Ampere’s law. (d) Gauss’s law for magnetism.

FIG. 8.

(Left) Current density map in the midcoronal view of the healthy head model where CSF shows the highest current density (dark red), while the mastoid bones (dark blue), vertebral column, and skull have the lowest current density distribution (green). (Right) A pie chart showing the fraction (in percentage) of the total 140 μA sinusoidal current present in different anatomical structures in the healthy head model, where almost 50% of the input current flows through the CSF and gray and white matter, with CSF having the largest portion of the injected current.

FIG. 9.

Scalp potential distributions on the surface of the numerical head model. (a) The healthy numerical head model and equipotential line distributions on the scalp with three pairs of electrodes located on the frontal (F7–F8), central (C3–C4), and parietal (P7–P8) areas of the scalp according to the 10–20 international EEG electrode placement system and symmetric relative to the midsagittal plane of the numerical head. The equipotential lines and potential gradients are symmetric relative to the midsagittal plane of the model and the three electrode pairs. (b) Numerical head model (left) with a spherical hemorrhagic lesion 20 mm in radius in the frontal left hemisphere close to F7 electrode. The hemorrhagic lesion creates a different potential gradient at F7 compared to the healthy control and its counterpart on the right hemisphere, F8; however, the other two electrode pairs in a greater distance from the lesion remain symmetric.

FIG. 10.

Simulated normalized potential difference of the healthy model and three numerical ICH-damaged models with a hematoma of 10.4 cm³ in (a) frontal, (b) central, and (c) parietal. In all cases, the scalp potential differences between the healthy and damage models are zero everywhere except in the vicinity of the lesion.

FIG. 11.

Contact impedances of electrodes Fpz, F7, C3, P7, F8, C4, and P8 for healthy control and ICH damaged patients at four different frequencies (100 Hz, 1 kHz, 10 kHz, and 100 kHz). Each box plot shows the median, and the edges of the box are the 25th and 75th percentiles: the whiskers extend to the most extreme data points.

FIG. 12.

Mean absolute voltage differences for frontal (F8–F7), central (C4–C3), and parietal (P8–P7) electrode pairs of healthy and patient subjects with brain injury (ICH): (a) mean absolute voltage differences of all three electrode pairs; (b) mean ± 1SD of each group in plot (a); (c) electrode pair with maximum voltage difference for each group; and (d) mean ± 1SD of each group in plot (c).