You are viewing the site in preview mode

Skip to main content
Download PDF

A novel method for detecting intracranial pressure changes by monitoring cerebral perfusion via electrical impedance tomography

Fluids and Barriers of the CNS volume 22, Article number: 10 (2025) Cite this article

Abstract

Background

Acute and critical neurological diseases are often accompanied with elevated intracranial pressure (ICP), leading to insufficient cerebral perfusion, which may cause severe secondary lesion. Existing ICP monitoring techniques often fail to effectively meet the demand for real-time noninvasive ICP monitoring and warning. This study aimed to explore the use of electrical impedance tomography (EIT) to provide real-time early warning of elevated ICP by observing cerebral perfusion.

Methods

An intracranial hypertension model was prepared by injecting autologous un-anticoagulated blood into the brain parenchyma of twelve Landrace swine. Invasive ICP monitoring was used as a control method, and a high-precision EIT system was used to acquire and analyze the changing patterns of cerebral perfusion EIT image parameters with respect to ICP. Four EIT parameters related to cerebral perfusion were extracted from the images, and their potential application in detecting ICP elevation was analyzed.

Results

When ICP increased, all EIT perfusion parameters decreased significantly (P < 0.05). When the subjects were in a state of intracranial hypertension (ICP > 22 mmHg), the correlation between EIT perfusion parameters and ICP was more significant (P < 0.01), with correlation coefficients ranging from −0.72 to −0.83. We tested the objects when they were in baseline ICP and in ICP of 15–40 mmHg. Under both circumstances, ROC curve analysis showed that the comprehensive model of perfusion parameters based on the random forest algorithm had a sensitivity and specificity of more than 90% and an area under the curve (AUC) of more than 0.9 for detecting ICP increments of both 5 and 10 mmHg.

Conclusion

This study demonstrates the feasibility of using perfusion EIT to detect ICP increases in real time, which may provide a new method for real-time non-invasive monitoring of patients with increased ICP.

Introduction

Neurological acute diseases, such as trauma, stroke, and brain tumors, are often accompanied by elevated intracranial pressure (ICP) [1]. Persistently elevated ICP may reduce cerebral perfusion, leading to ischemia and hypoxia of brain tissues, which can ultimately lead to severe neurological damage and brain death, if left untreated [2, 3]. Early monitoring of elevated ICP is critical for the proper treatment of patients in neurocritical care and for the  improvement of prognosis [4]. Therefore, real-time monitoring of elevated ICP to prevent inadequate cerebral perfusion is of great clinical value in the diagnosis and treatment of craniocerebral diseases.

Existing ICP monitoring methods are mainly categorized into invasive ICP monitoring and noninvasive ICP monitoring [5]. Invasive ICP monitoring is widely recognized as the gold standard for ICP monitoring. However, this method may lead to complications such as intracranial hemorrhage and infection during placement of the pressure transducer [6]. Therefore, a variety of noninvasive techniques for ICP monitoring such as Transcranial Doppler [7], Optic Nerve Sheath Diameter [8], and Magnetic Resonance Imaging [9] have been developed, to indirectly obtain ICP information. However, these methods require specialized personnel in clinical applications. Also, they are not handy to use, and are not suitable for prolonged noninvasive real-time monitoring. As a result, there is an urgent clinical need for a noninvasive and harmless bedside ICP monitoring method with results that are easy to interpret, as well as realizing the monitoring and warning of ICP changes effectively.

Electrical impedance tomography (EIT) is a non-invasive functional imaging technique that measures boundary voltages on body surfaces to estimate the spatial distribution of tissue resistivity in vivo [10]. Different tissues have different impedance levels, which can change under physiological and pathological conditions. EIT can therefore detect physiological and pathological changes in internal impedance [11]. EIT has been investigated in a number of scenarios, such as imaging of lung respiratory function and perfusion [12], functional brain imaging [13], and functional imaging of abdominal organs [14]. Yang et al. investigated the correlation between EIT imaging results and ICP changes during mannitol dehydration, which showed that the changes in EIT reconstruction values caused by the dehydration effect are strongly correlated with the changes in ICP of patients [15]. Yang et al.’s study demonstrated the unique advantages of the EIT for ICP monitoring, which, as a completely noninvasive and harmless monitoring approach, achieve long-term monitoring of intracranial conditions with high temporal resolution. It provided a basis for the application of EIT to noninvasive ICP monitoring. However, it was limited to only one case of mannitol dehydration therapy and focused on the basal impedance properties of brain tissue. Besides, the results might be affected by the nature of the brain disease, injury, or changes in physiological status, possibly damaging the specificity of the results.

Continuously elevated ICP may disrupt the cerebral autoregulation, causing a decrease in cerebral blood flow (CBF), which in turn may lead to insufficient blood supply to brain tissues, resulting in poor perfusion [16, 17]. This phenomenon suggests that we can indirectly assess ICP changes by monitoring the state of cerebral perfusion [18], thus realizing real-time warning of the adverse consequences of elevated ICP. Given that blood is a highly conductive substance, alterations in its volume can have a considerable impact on the impedance value. It can be theoretically demonstrated that it is feasible to reconstruct the image of blood volume changes caused by cerebral blood flow pulsations through EIT. Therefore, our team carried out a series of exploratory research work on cerebral perfusion imaging around this mechanism in the early stage [19,20,21]. We successfully extracted the dynamic cerebral perfusion signal, and verified the feasibility of using EIT to characterize the perfusion process. These results provide a basis for EIT to monitor the cerebral perfusion status reflecting ICP changes.

In this study, we constructed an intracranial hypertension swine model and used invasive ICP monitoring as a control method to analyze the changing pattern of cerebral perfusion parameters of EIT with ICP. We aimed to ascertain the feasibility of EIT in reflecting the ICP changes through the monitoring of cerebral perfusion.

Materials and methods

Animal preparation

Twelve healthy Landrace swine weighing 18.5 ± 3 kg were used for the experiment. The experimental protocol and procedures were approved by the Animal Ethics Committee of the Air Force Medical University (Ethical permission number: IACUC-20241299). Anesthesia was induced in all swine with a dose of 3 mg/kg of Zoletil ®50, and the swine were extubated and anesthesia was maintained with isoflurane of concentration of 1–3%.

The femoral artery was punctured with an indwelling needle to establish access for blood sampling. After arterial blood pressure, oxygen saturation, and end-expiratory carbon dioxide were monitored, the scalp of each Landrace swine was incised longitudinally, the skull was separated and exposed, the periosteum was scraped, and a window with a sagittal length of approximately 60 mm and a coronal length of approximately 60 mm was formed. Two holes were drilled using cranial drills, one at 15 mm to the right of the sagittal suture, 15 anterior to the coronal suture for the placement of ICP probe and the other at 15 mm on the right side of the sagittal suture, 10 mm on the posterior side of the coronal suture for the placement of a 22 G indwelling needle, which was inserted into the hole to a depth of 18 mm. In addition, 16 sterilized 1.2-mm dental root canal stumps (Xihu Biomaterials, Hangzhou, Zhejiang Province, China, square-head type) were drilled into the margins of the skull at equal intervals (Fig. 1A), none of which penetrated the dura mater. The above cranial holes were sealed with bone wax to minimize the possibility of cerebrospinal fluid leakage.

Fig. 1
figure 1

Schematic diagram of the experimental method and procedure. A 16 copper electrode positions (yellow), indwelling needle position, and ICP probe position. B Diagram of the experimental scenario. C Experiment time. EIT data acquisition was started at baseline after about 10 min of blood injection, with an interval of about 10 min between injections, for a total of 5 injections. D Schematic diagram of the ARV curves and the extraction of perfusion parameters

Full size image

ICP monitor and EIT system

ICP data was acquired by an ICP monitor (SOPHYSA Inc., Orsay, France) with a frame rate of 2 frames per second (Fig. 1B).

EIT data was obtained using EC-100 PRO (UTRON Technology Co., Ltd., Hangzhou, China, Fig. 1B), a high-speed and high-precision system developed by our team [19, 20]. The system operates over at a frequency range of 10–250 kHz, an output current range of 10–1250 μA, and a signal-to-noise ratio greater than 90 dB. For the experiments, the frame rate of data acquisition was set to 40 frames per second and the excitation frequency was set to50kHz.

Experimental protocols

The intracranial hypertension model was prepared by injecting autologous un-anticoagulated blood into the brain parenchyma of swine in batches. After connecting the electrodes to the EIT system, baseline EIT data was collected and ICP data acquisition was started simultaneously. After approximately 10 min, 2 ml of blood was withdrawn through an indwelling needle preplaced in the femoral artery and then injected into the parenchymal region of the brain through an indwelling needle preplaced intracranially at a uniform rate within 1 min. This process was repeated after approximately 10 min for a total of 5 times, and its time sequence is illustrated in Fig. 1C. Blood pressure was maintained constant throughout the procedure using norepinephrine. When the surgery was completed, the animals were put to death under deep anesthesia, the cranium was opened with a cranial drill and a bone-biting forceps, and the brain tissue was removed intact for observation of the location of the hematoma.

EIT image reconstruction and perfusion parameter extraction

The original boundary voltage signals collected by the EIT system were first passed through a bandpass filter with a passband frequency of 1–5 Hz to extract the voltage signals reflecting the dynamic blood perfusion. Then the circular finite element model conforming to the actual electrode distribution location was used for the forward problem calculation to obtain the potential at each node. Finally, the inverse problem calculation was carried out by the damped least squares method [22] to obtain the reconstructed image reflecting the dynamic changes of blood flow, as shown in Eq. (1).

$$\Delta \rho = \left( {J^{T} J + \lambda R} \right)^{ - 1} J^{T} \Delta V$$
(1)

Δρ is the change in resistivity distribution between the current frame and the reference frame, ΔV is the normalized boundary voltage change between the current frame and the reference frame, J is the Jacobi matrix, λ is the regularization parameter, and R is the regularization matrix.

The average reconstruction value (ARV) associated with the whole brain in the EIT image sequence was calculated by Eq. (2).

$$ARV = \frac{{\left( {\mathop \sum \nolimits_{i = 1}^{N} \Delta \rho_{i} } \right)}}{N}$$
(2)

Δρi denotes the reconstruction resistivity of the i-th pixel in the EIT image; N denotes the total number of pixels in the region of interest.

ARV reflects dynamic changes in resistivity caused by pulsatile changes in the congested state of brain tissue. Since there is a significant difference in resistivity between blood and brain tissue, when the heart contracts, blood is pumped into the cranial cavity, resulting in a decrease in resistivity. This process is represented in the curve (Fig. 1D) as a drop from the peak (ARVp) to the trough (ARVv), termed the descending branch, with the corresponding time interval denoted as Td. Subsequently, during cardiac diastole, the blood gradually flows out of the cranial cavity, with a consequent rise in resistivity from ARVv back to the next peak (ARVp′) defined as the ascending branch, with a time interval denoted as Ta. A complete cerebral perfusion cycle is defined as a process that begins at ARVp, passes through ARVv, and reaches ARVp′. In order to provide a more detailed description of the perfusion state, a series of parameters were extracted for each vascular pulse cycle. These included the mean perfusion velocity (MV), the systolic wave height (Hs), the inflow volume velocity (IV), and the angle between the ascending branch and the baseline (Aab).

The MV is used to quantify the rate of blood volume change in the vessels. It was calculated by Eq. (3):

$$MV= \frac{\sum_{i=1}^{N}\left({ARV}_{p}-{ARV}_{i}\right)}{{T}_{a}+{T}_{d}}$$
(3)

ARVi denotes the average reconstruction value of the ith EIT image frame; N denotes the total number of EIT image frames acquired in the current perfusion cycle.

The Hs primarily reflects the filling degree of the cerebral vessels, which is indicative of the degree of pulsatile blood flow supply. The IV is defined as the ratio of Hs to Ta, which serves to reflect the correlation between the two variables. The Hs and IV were calculated by Eqs. (4) and (5), respectively:

$${H}_{s}={ARV}_{p}{-ARV}_{v}$$
(4)
$$IV= \frac{{H}_{s}}{{T}_{d}}$$
(5)

The Adb is a parameter that reflects the elasticity of the cerebral blood vessels. It is calculated using Eq. (6):

$${A}_{db}=arccos\left(\frac{{X}_{p}\left({X}_{v}-{X}_{p}\right)}{\sqrt{{{X}_{p}}^{2}}\bullet \sqrt{{\left({X}_{v}-{X}_{p}\right)}^{2}+{\left({Y}_{v}-{Y}_{p}\right)}^{2}}}\right)$$
(6)

The baseline is defined as the vector starting from \({\text{ARV}}_{\text{p}}\) whose direction is parallel to the x-axis. \({\text{X}}_{\text{p}}{,\text{Y}}_{\text{p}}\) are the x and y coordinates of \({\text{ARV}}_{\text{p}}\); \({\text{X}}_{\text{v}},{\text{Y}}_{\text{v}}\) are the x and y coordinates of \({\text{ARV}}_{\text{v}}\).

Data analysis and statistical methods

We analyzed our experiment results using SPSS 27.0 statistical software (SPSS Inc, Chicago, IL, USA) and MATLAB (MathWorks, R2022a, USA.). The level of significance for statistical analysis was 0.05. Data was tested for normality using the Shapiro–Wilk test. The ICP values were categorized into the following four groups according to the ICP grading criteria: 0–15, 15–20, 20–40, and >40 mmHg. To examine the mean perfusion parameters across samples at varying ICP levels, Friedman test was employed to investigate the variability of these parameters with ICP. Subsequently, the Bonferroni’s post-hoc test was utilized to conduct pairwise comparisons. PASS 15.0 was used to calculate the average statistical power of the four parameters based on the current sample size. The pixel values of functional EIT (fEIT) are not mere impedance changes, but valuable functional information that influences clinical decisions. A sequence of EIT images was calculated for each set of 5 perfusion cycles, from which the waveform of each pixel was extracted to calculate the average perfusion parameters, and these perfusion parameters were presented as images to generate fEIT images.

Based on the International Traumatic Brain Injury Foundation Intervention criteria for the treatment of intracranial hypertension [23], we calculated the Spearman correlation coefficients between the perfusion parameters and ICP in the two intervals defined by an ICP value of 22 mmHg, and then assessed the differences in the correlation coefficients between the two intervals using a paired t-test.

To assess the diagnostic efficacy of EIT parameters in detecting ICP elevation, the mean MV, Hs, IV, and Adb at baseline were first determined for each sample as reference values, and the normalized change in these parameters relative to baseline (ΔMV, ΔHs, ΔIV, ΔAdb), as well as the change in ICP (ΔICP), were calculated. With 500 randomly selected sample data per animal and a total of 6000 sample data for ROC curve analysis, we evaluated the efficacy of each independent EIT parameter change in detecting ICP elevations >5 and 10 mmHg. Finally, AUC, sensitivity and specificity were calculated from the ROC curves.

Finally, we also modelled the ICP elevation change detection using the random forest algorithm. The study dataset was composed of 6000 samples of ΔMV, ΔHs, ΔIV, ΔAdb, and Y. ΔMV, ΔHs, ΔIV, and ΔAdb are the normalized rate of change of MV, Hs, IV, and Adb, respectively, with respect to the reference value. Y = {0,1} is the label (where 0 and 1 each indicate that the ΔICP is less than and greater than a certain threshold value). The dataset was then randomly divided into training (70%) and test (30%) datasets. The training set was utilized to train the random forest model. With the help of the test set, we comprehensively evaluated the detection efficiency of the model by calculating the AUC, sensitivity and specificity. The reference values chosen were normal ICP values and intracranial hypertension values (between 15 and 40 mmHg) because patients are often already in a state of intracranial hypertension at the time of initial monitoring, making it critical to explore the ability of the random forest algorithm to detect elevated ICP for different reference values.

Results

Variation of EIT signal with ICP

In Fig. 2A, exemplary changes in EIT transmission impedance, ICP, and mean arterial pressure were presented throughout the course of the experiment, in which five injections of blood were performed, each of which resulted in a rapid rise in transmission impedance. At the same time, ICP decreased slowly after the rapid rise. The mean arterial pressure remained stable throughout the experiment without significant fluctuations. Figure 2B, on the other hand, shows the changes in ICP and ARV before and after a single blood injection. It can be observed that after the blood injection, ICP increased rapidly and then decreased slowly. At the same time, the amplitude of ARV decreased rapidly and then gradually increased. In order to investigate the effect of ICP changes on the perfusion EIT images, time points corresponding to three different ICP levels (i.e., t1, t2, and t3 in Fig. 2) were selected for the EIT images. Each of these time points represented a cardiac cycle. Using the beginning of systole as the reference frame, differential imaging of the cardiac cycle at these three time points was performed. The results are shown in Fig. 2C. By observing the dynamic cerebral perfusion EIT images at the selected time points, we saw that the cerebral perfusion in the edge region of the images gradually decreased with the gradual increase in ICP. After removing the pulsation signal from the EIT using low-pass filtering, the base impedance was determined. Figure 2D shows the results of EIT-based impedance imaging for moments t1, t2, and t3, using t1 as the reference moment. The consecutive EIT images show that the blue region in the lower left corner represented the blood injection region. The impedance of this region gradually increased with the number of blood injections.

Fig. 2
figure 2

A Raw measurement data of EIT and ICP. The upper part indicates the average transmission impedance from 256 channels. The red arrow shows the blood injection time point. The middle part shows the raw and filtered data of ICP. The lower part shows the mean arterial pressure data. B Comparison of ARV and ICP during blood injection. C The average transmission impedance of the perfusion cycle corresponding to t1, t2 and t3 is shown in the upper part (in the red box). The corresponding EIT image of the perfusion cycle is shown in the lower part (sequence interval of the images: 25 ms). t1, t2 and t3 correspond to ICPs of approximately 10, 25, and 40 mmHg, respectively. D Reconstructed EIT base impedance image with t1 as reference data

Full size image

Effect of ICP changes on EIT perfusion parameters

Figure 3 shows the mean perfusion parameters of all samples at different ICP levels. The Shapiro–Wilk test showed that the data obeyed a normal distribution (P > 0.05). The application of Friedman test revealed significant differences (P < 0.001) in perfusion parameters at different ICP levels. Subsequently, a Bonferroni post-hoc test was conducted to perform pairwise comparisons, and the results are presented in Fig. 3. The average statistical power was 0.94.The fEIT images in the lower part of the Fig. 3 similarly reveal a trend of a gradual decrease in perfusion parameters in the peripheral regions of the image with increasing ICP.

Fig. 3
figure 3

Comparison of mean perfusion parameters for all samples at different ICP levels. The lower image shows the parameter distribution of one sample for one perfusion cycle at 0–15, 15–20, 20–40, and >40 mmHg, where the value of a pixel is the perfusion parameter value of that pixel. A mean perfusion velocity (MV); B systolic wave height (Hs); C inflow volume velocity (IV); D angle between descending branch and baseline (Adb). Ns no significant difference; *P < 0.05; **P < 0.01; ***P < 0.001

Full size image

Table 1 shows the mean Spearman correlation coefficients and standard deviations between ICP and each perfusion parameter. It is obvious from the table that the correlation was significantly higher in the group above 22 mmHg than in the group below 22 mmHg (P < 0.01). This suggests that at higher ICP levels, there is a stronger correlation between ICP and cerebral perfusion, i.e., the effect of ICP on CBF is more significant: as ICP rises further, CBF decreases more significantly.

Table 1 Correlation coefficient between ICP and perfusion parameters
Full size table

Evaluation of the detection efficacy of EIT perfusion parameters in the detection of elevated ICP

The ability of ΔMV, ΔHs, ΔIV, and ΔAdb as single variables to detect ICP elevation greater than 5 and 10 mmHg was assessed by ROC curve analysis (Fig. 4A, B). The AUCs of these variables for detecting ICP elevation greater than 5 mmHg are 0.80, 0.74, 0.74, and 0.72, respectively, and the AUCs for detecting ICP elevation greater than 10 mmHg are 0.86, 0.76, 0.73, and 0.73, respectively. The specific values of the AUC, sensitivity, and specificity for each of the parameters are presented in Table 2. This suggests that independent EIT perfusion parameters have some accuracy in the diagnosis of elevated ICP.

Fig. 4
figure 4

ROC curves of four independent perfusion parameters and random forest model for detecting ICP elevation relative to baseline. A Perfusion parameter for detecting ICP elevation greater than 5 mmHg; B perfusion parameter for detecting ICP elevation greater than 10 mmHg; C random forest model for detecting ROC curves of ICP elevation greater than 5 mmHg; D random forest model for detecting ROC curves of ICP elevation >10 mmHg

Full size image
Table 2 Evaluation indexes of different parameters in the diagnosis of ΔICP > 5 mmHg and ΔICP > 10 mmHg
Full size table

In this study, a random forest model was used to detect ICP elevation, which achieved an AUC of 0.95, 95% sensitivity with 96% specificity for an ICP elevation of 5 mmHg (Fig. 4C). In detecting ICP elevation at 10 mmHg (Fig. 4D), the AUC was 0.96, with sensitivity and specificity of 95 and 97%. Figure 5 demonstrates the performance of the random forest model for detecting ICP elevations over 5 and 10 mmHg at different reference values. Specifically, the model performed well in terms of AUC, sensitivity and specificity, with values exceeding 0.92. Notably, these performance metrics show an increasing trend as the reference value increases.

Fig. 5
figure 5

Detection efficacy of the random forest model for detecting ICP elevation when changing reference values. A Detection of ICP elevation greater than 5 mmHg. B Detection of ICP elevation greater than 10 mmHg

Full size image

Discussion

Persistently elevated ICP can result in cerebral blood supply deficiency, a significant contributor to secondary brain injury [24]. ICP monitoring has become an important part of brain monitoring after brain injury [25]. However, existing ICP monitoring techniques have various limitations, which make it difficult to effectively realize real-time noninvasive ICP monitoring [26]. EIT has outstanding advantages in terms of high temporal resolution, noninvasiveness, and bedside monitoring, making it a promising approach to overcome the limitations of the existing techniques. Previous studies [15], only focused on the relationship between ICP and EIT basal impedance, even though basal impedance is affected by many factors and has poor specificity. Therefore, based on the effect of ICP on CBF [27], this study attempted to indirectly assess the changes in ICP by monitoring the cerebral perfusion status through EIT, opening new perspectives and methods for ICP monitoring. In this study, we explored the potential application of dynamic cerebral perfusion EIT in real-time monitoring of ICP changes by constructing an intracranial hypertension model in swine. Through EIT reconstructed images, we visualized the cerebral perfusion changes under different ICP levels. Subsequently, we extracted the parameters related to blood pulsatility from ARV and analyzed the patterns of these parameters with intracranial pressure levels. In addition, we explored the potential of EIT perfusion parameters in detecting elevated ICP.

Changes in cerebral perfusion status with ICP can be observed by EIT

In Fig. 2A, the changes of ICP and basal impedance with the progress of the experiment can be observed: the ICP increased rapidly during each blood injection, and decreased rapidly and then slowly after the end of the blood injection. This change in ICP became more and more obvious as the experiment progressed, and the baseline ICP gradually increased from 10 mmHg to about 20 mmHg. This indicates that the blood injection will lead to a significant increase in ICP, and after the blood injection is completed, due to the influence of the animals’ own ICP compensatory mechanism, their ICP will gradually fall back, but the overall trend is still rising. The basal impedance of the brain increased rapidly after each blood injection [28], but did not decrease significantly after the completion of blood injection, and its baseline gradually increased with the increase of blood injection. Meanwhile, the relationship between ICP and the dynamic perfusion signal is more direct: as can be seen from the dynamic perfusion signal (ARV) in Fig. 2B, the amplitude of the ARV decreased with the increase of ICP and gradually rebounded when ICP decreased. In a further study, we analyzed the potential application of utilizing dynamic CBF perfusion EIT image parameters in ICP monitoring. We observed in dynamic cerebral perfusion EIT images (Fig. 2C) that intracranial hypertension (25–40 mmHg) resulted in a significant decrease in peripheral cerebral perfusion relative to normal ICP levels (10 mmHg), and that the higher the ICP, the more pronounced the decrease. This result suggests that elevated ICP directly leads to a decrease in CBF in these regions and that this change can be detected by the EIT.

In further analysis, we used a number of perfusion parameters aiming to quantitatively describe the perfusion status of CBF. By comparing the cerebral perfusion parameters at different ICP levels (Fig. 3), it can be seen that all these parameters decreased significantly with increasing ICP, further confirming the sensitivity and effectiveness of EIT in monitoring ICP fluctuations. The fEIT images of each parameter shown at the bottom of Fig. 3 present in detail the trends of each cerebral perfusion parameter at different ICP levels, while providing information about the perfusion parameters of the global intracranial region. Notably, the significant differences in the perfusion parameters in the comparison of ICP levels suggest that the correlation of these parameters with ICP changes may vary at different stages. In order to study the correlation between ICP and perfusion parameters in more depth, we divided the dataset into two intervals according to whether the ICP exceeded the therapeutic threshold of 22 mmHg [23] and analyzed the relationship between ICP and perfusion parameters within these two intervals separately. The results showed that in the interval where the ICP exceeded 22 mmHg, a significant negative correlation between ICP and perfusion parameters was demonstrated, with a correlation coefficient of approximately −0.8. In the interval where the ICP did not exceed this threshold, this negative correlation was weaker, with a correlation coefficient of approximately −0.2. In addition, the difference of the correlation between the two intervals was statistically significant (P < 0.01), suggesting the stronger association between ICP and perfusion parameters in the state of intracranial hypertension. This trend might be due to the fact that the cerebral autoregulation mechanism was able to maintain the stability of CBF during the initial rise in ICP (<22 mmHg) [24, 29]. However, once ICP reached a certain threshold, this self-regulatory ability was limited and cerebral perfusion decreased significantly as ICP continued to rise [30]. This suggest that the EIT is more appropriate for monitoring the changing status of ICP in patients with intracranial hypertension.

EIT perfusion parameters promising for ICP change volume assessment

By performing ROC curve analysis on a single parameter, we revealed its potential value in detecting ICP rise. Moreover, by evaluating the ability of EIT to detect ICP elevations greater than 10 mmHg, superior results were obtained compared to a previous study by Yang [15]. In addition, we explored the performance of the EIT in detecting ICP rises greater than 5 mmHg, which likewise demonstrated its good detecting ability. In addition, a random forest model was developed to detect ICP elevation, and its AUC, sensitivity and specificity for detecting ICP elevation greater than 10 mmHg were 0.96, 95 and 97%. To further examine the ability of the random forest model to detect ICP at different intracranial pressure levels, 15–40 mmHg were selected as baseline reference values in this study. The efficiency of the model in detecting ICP changes with increases of 5 and 10 mmHg relative to these baseline values was evaluated. The model’s AUC, sensitivity, and specificity exceeded 0.92 at all selected reference values. The results show that the EIT parameters are capable of detecting ICP elevation, which lays a solid foundation for the development and optimization of related diagnostic techniques in the future.

Limitations of ICP monitoring based on EIT basal impedance

It is noteworthy that in our study, the plateau period of both cerebral basal impedance and ICP increased with increasing blood injection, showing a positive correlation, which is different from the negative correlation between basal impedance and ICP obtained by Yang et al. [15]. This discrepancy may be due to the fact that in the study of Yang et al., the dehydration process utilized the high osmotic pressure properties of the dehydrating agent to allow water molecules in the extracellular fluid in the vicinity of the microvessels to enter the vasculature [31], which led to a reduction in the cellular interstitial component and decreased its electrical conductivity, causing the brain’s basal impedance to rise with decreasing ICP. In contrast, after injection of un-anticoagulated blood into the brain parenchyma, the increase in the contents of the cranial cavity caused an increase in ICP, which was compensated for by the absorption of cerebrospinal fluid (CSF) [32]. Since the resistivity of blood is higher than that of CSF, this led to the phenomenon of cerebral basal impedance rising with ICP in this study [33]. This phenomenon suggests that the relationship between the basal impedance of brain tissue and ICP may vary in different clinical applications. Therefore, the method of ICP monitoring based on dynamic perfusion impedance parameters has significant potential because the relationship between dynamic perfusion impedance parameters and ICP is more direct in this method, which is important for the application of EIT in monitoring ICP.

Advantages and limitations of perfusion EIT for monitoring ICP

In view of the inherent risks associated with invasive ICP measurement, a wide range of non-invasive methodologies have been investigated [34, 35]. These methods typically assess physiological alterations that are indirectly associated with ICP, and can be classified into four principal categories: fluid dynamics, ophthalmology, ear, and electrophysiology [26, 36]. In this study, we employed EIT to assess cerebral perfusion, thereby enabling us to estimate changes in ICP. Consequently, EIT can be classified as one of the fluid dynamics methods. This method is similar to TCD and Near-Infrared Spectroscopy (NIRS) [37, 38], both of which estimate ICP levels by measuring changes in parameters related to CBF. However, the application of NIRS for ICP measurement is constrained by the availability of NIRS equipment, with approximately 50% of records failing to meet the requisite reliability standards [26, 39]. In contrast to TCD, which can only detect specific large blood vessels [40], EIT can provide comprehensive information on the perfusion status of the entire brain. Furthermore, EIT is able to monitor continuously for extended periods of time, which is a capability that TCD and the majority of other non-invasive monitoring techniques lack [24]. It can be reasonably concluded that EIT represents an efficacious bedside monitoring instrument, capable of furnishing timely alerts regarding elevated ICP. This functionality enables prompt intervention and treatment by medical personnel, thereby reducing the necessity for invasive intracranial pressure monitoring techniques and minimizing patient risk and discomfort.

There are several limitations to this study due to the conditions. (1) During prolonged monitoring, although the observed cerebral perfusion parameters were directly affected by changes in ICP, they were theoretically also affected by blood pressure fluctuations. Given the stability of blood pressure could not be fully controlled during the experiment, this somewhat limited the feasibility and reliability of the EIT-ICP correlation analysis. Future studies should consider the implementation of simultaneous blood pressure continuous monitoring, which will allow for more precise adjustment and interpretation of effects caused by blood pressure fluctuations during data analysis. (2) In the results of this study, we focused on evaluating the ability of the EIT to detect elevated ICP without delving into the absolute value of ICP. This decision was made primarily to account for the effects of interindividual variability, which might potentially blur the accurate assessment of the absolute value of ICP. Therefore, future studies should focus on developing and optimizing EIT parameters that reflect individual ICP characteristics to reduce the impact of individual variability. (3) Although this study demonstrated that EIT perfusion parameters can be used for the monitoring of ICP changes, the small sample size might limit the statistical power of the study and the broad applicability of the results. Therefore, it is still necessary to promote the clinical application of related methods by larger sample sizes in future studies. (4) Although this study has demonstrated the potential of EIT to monitor ICP through animal experiments, several practical issues remain to be resolved before it can be applied in clinical practice. For instance, in animal experiments, electrodes are penetrated through the skull to enhance signal stability; however, in clinical practice, non-invasive scalp electrodes are typically preferred. Furthermore, other medical devices in the ICU environment may generate electromagnetic interference, which could impair the quality of the EIT signal. Although some shielding methods are currently available to reduce the impact of external interference, these factors still need to be further studied and optimized in order to better promote the clinical application of EIT. An intracranial hypertension animal model was constructed by injecting blood into a specific area, which may not fully reflect the specific conditions of various complex areas in a clinical environment. It is therefore evident that, in order to facilitate a more accurate comprehension and implementation of these findings, the collection of additional clinical data is imperative for subsequent in-depth analysis.

Conclusion

In this study, we constructed an intracranial hypertension swine model. A high-performance EIT system recently developed by our team was utilized to collect information about weak impedance caused by the CBF. The EIT parameters reflecting the changes of the cerebral perfusion status were extracted, and possible effects of ICP elevations on cerebral perfusion impedance were analyzed. This is the first time to confirm the feasibility of EIT to achieve ICP elevation early warnings by monitoring the changes of the CBF status. The results demonstrate that EIT could be a promising bedside monitoring tool for real-time non-invasive monitoring and early warning of ICP changes, which may have great potential for clinical application.

Availability of data and materials

The datasets analyzed during the current study are available from the corresponding author on reasonable request.

Abbreviations

ICP:

Intracranial pressure

EIT:

Electrical impedance tomography

CBF:

Cerebral blood flow

ROC:

Receiver operating characteristic

AUC:

Area under the curve

ARV:

Average reconstruction value

fEIT:

Functional electrical impedance tomography

References

  1. Canac N, Jalaleddini K, Thorpe SG, Thibeault CM, Hamilton RB. Review: pathophysiology of intracranial hypertension and noninvasive intracranial pressure monitoring. Fluids Barriers CNS. 2020;17:21.

    Article  Google Scholar 

  2. Claassen J, Thijssen DHJ, Panerai RB, Faraci FM. Regulation of cerebral blood flow in humans: physiology and clinical implications of autoregulation. Physiol Rev. 2021;101:1487–559.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  3. Steiner LA, Andrews PJ. Monitoring the injured brain: ICP and CBF. Br J Anaesth. 2006;97:26–38.

    Article  PubMed  CAS  Google Scholar 

  4. Robba C, Graziano F, Rebora P, Elli F, Giussani C, Oddo M, Meyfroidt G, Helbok R, Taccone FS, Prisco L, Vincent JL, Suarez JI, Stocchetti N, Citerio G. Intracranial pressure monitoring in patients with acute brain injury in the intensive care unit (SYNAPSE-ICU): an international, prospective observational cohort study. Lancet Neurol. 2021;20:548–58.

    Article  PubMed  CAS  Google Scholar 

  5. Hawryluk GWJ, Citerio G, Hutchinson P, Kolias A, Meyfroidt G, Robba C, Stocchetti N, Chesnut R. Intracranial pressure: current perspectives on physiology and monitoring. Intensive Care Med. 2022;48:1471–81.

    Article  PubMed  Google Scholar 

  6. Heldt T, Zoerle T, Teichmann D, Stocchetti N. Intracranial pressure and intracranial elastance monitoring in neurocritical care. Annu Rev Biomed Eng. 2019;21:523–49.

    Article  PubMed  CAS  Google Scholar 

  7. Fernando SM, Tran A, Cheng W, Rochwerg B, Taljaard M, Kyeremanteng K, English SW, Sekhon MS, Griesdale DEG, Dowlatshahi D, McCredie VA, Wijdicks EFM, Almenawer SA, Inaba K, Rajajee V, Perry JJ. Diagnosis of elevated intracranial pressure in critically ill adults: systematic review and meta-analysis. BMJ. 2019;366: l4225.

    Article  PubMed  PubMed Central  Google Scholar 

  8. Li J, Wan C. Non-invasive detection of intracranial pressure related to the optic nerve. Quant Imaging Med Surg. 2021;11:2823–36.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  9. Barkatullah AF, Leishangthem L, Moss HE. MRI findings as markers of idiopathic intracranial hypertension. Curr Opin Neurol. 2021;34:75–83.

    Article  PubMed  PubMed Central  Google Scholar 

  10. Brown BH. Electrical impedance tomography (EIT): a review. J Med Eng Technol. 2003;27:97–108.

    Article  PubMed  CAS  Google Scholar 

  11. Pettigrew RI, Peterson KP, Heetderks W, Seto B. The National Institute of Biomedical Imaging and Bioengineering marks its first five years. Acad Radiol. 2007;14:1448–54.

    Article  PubMed  Google Scholar 

  12. Frerichs I, Amato MB, van Kaam AH, Tingay DG, Zhao Z, Grychtol B, Bodenstein M, Gagnon H, Böhm SH, Teschner E, Stenqvist O, Mauri T, Torsani V, Camporota L, Schibler A, Wolf GK, Gommers D, Leonhardt S, Adler A. Chest electrical impedance tomography examination, data analysis, terminology, clinical use and recommendations: consensus statement of the translational EIT development study group. Thorax. 2017;72:83–93.

    Article  PubMed  Google Scholar 

  13. Ke XY, Hou W, Huang Q, Hou X, Bao XY, Kong WX, Li CX, Qiu YQ, Hu SY, Dong LH. Advances in electrical impedance tomography-based brain imaging. Mil Med Res. 2022;9:10.

    PubMed  PubMed Central  Google Scholar 

  14. Sadleir RJ, Fox RA. Detection and quantification of intraperitoneal fluid using electrical impedance tomography. IEEE Trans Biomed Eng. 2001;48:484–91.

    Article  PubMed  CAS  Google Scholar 

  15. Yang B, Li B, Xu C, Hu S, Dai M, Xia J, Luo P, Shi X, Zhao Z, Dong X, Fei Z, Fu F. Comparison of electrical impedance tomography and intracranial pressure during dehydration treatment of cerebral edema. Neuroimage Clin. 2019;23: 101909.

    Article  PubMed  PubMed Central  Google Scholar 

  16. Bergman L, Cluver C, Carlberg N, Belfort M, Tolcher MC, Panerai RB, Van Veen T. Impaired cerebral autoregulation, cerebral perfusion pressure, and intracranial pressure in eclampsia. Am J Obstet Gynecol. 2022;226:287–8.

    Article  PubMed  Google Scholar 

  17. Conzen C, Becker K, Albanna W, Weiss M, Bach A, Lushina N, Steimers A, Pinkernell S, Clusmann H, Lindauer U, Schubert GA. The acute phase of experimental subarachnoid hemorrhage: intracranial pressure dynamics and their effect on cerebral blood flow and autoregulation. Transl Stroke Res. 2019;10:566–82.

    Article  PubMed  CAS  Google Scholar 

  18. Martínez-Palacios K, Vásquez-García S, Fariyike OA, Robba C, Rubiano AM. Non-invasive methods for intracranial pressure monitoring in traumatic brain injury using transcranial doppler: a scoping review. J Neurotrauma. 2024;41(11–12):1282–98.

    Article  PubMed  Google Scholar 

  19. Li WC, Xia JY, Zhang G, Ma H, Liu BY, Yang L, Zhou YM, Dong XZ, Fu F, Shi XT. Fast high-precision electrical impedance tomography system for real-time perfusion imaging. Ieee Access. 2019;7:61570–80.

    Article  Google Scholar 

  20. Shi XT, Li WC, You FS, Huo XY, Xu CH, Ji ZY, Liu RG, Liu BY, Li YD, Fu F, Dong XZ. High-precision electrical impedance tomography data acquisition system for brain imaging. IEEE Sens J. 2018;18:5974–84.

    Article  Google Scholar 

  21. Yan X, Wang Y, Li W, Zhu M, Wang W, Xu C, Li K, Liu B, Shi X. A preliminary study on the application of electrical impedance tomography based on cerebral perfusion monitoring to intracranial pressure changes. Front Neurosci. 2024;18:1390977. https://doiorg.publicaciones.saludcastillayleon.es/10.3389/fnins.2024.1390977.

    Article  PubMed  PubMed Central  Google Scholar 

  22. Xu C, Dai M, You F, Shi X, Fu F, Liu R, Dong X. An optimized strategy for real-time hemorrhage monitoring with electrical impedance tomography. Physiol Meas. 2011;32:585–98.

    Article  PubMed  Google Scholar 

  23. Carney N, Totten AM, O’Reilly C, Ullman JS, Hawryluk GW, Bell MJ, Bratton SL, Chesnut R, Harris OA, Kissoon N, Rubiano AM, Shutter L, Tasker RC, Vavilala MS, Wilberger J, Wright DW, Ghajar J. Guidelines for the management of severe traumatic brain injury, Fourth edition. Neurosurgery. 2017;80:6–15.

    Article  PubMed  Google Scholar 

  24. Evensen KB, Eide PK. Measuring intracranial pressure by invasive, less invasive or non-invasive means: limitations and avenues for improvement. Fluids Barriers CNS. 2020;17:34.

    Article  PubMed  PubMed Central  Google Scholar 

  25. Dattilo M. Noninvasive methods to monitor intracranial pressure. Curr Opin Neurol. 2023;36:1–9.

    Article  PubMed  Google Scholar 

  26. Zhang X, Medow JE, Iskandar BJ, Wang F, Shokoueinejad M, Koueik J, Webster JG. Invasive and noninvasive means of measuring intracranial pressure: a review. Physiol Meas. 2017;38:R143-r182.

    Article  PubMed  Google Scholar 

  27. Zadka Y, Doron O, Rosenthal G, Barnea O. Mechanisms of reduced cerebral blood flow in cerebral edema and elevated intracranial pressure. J Appl Physiol. 2023;1985(134):444–54.

    Article  Google Scholar 

  28. Dai M, Liu XC, Li HT, Xu CH, Yang B, Wang H, Shi XT, Dong XZ, Fu F. EIT imaging of intracranial hemorrhage in rabbit models is influenced by the intactness of cranium. Biomed Res Int. 2018;2018:1321862.

    Article  PubMed  PubMed Central  Google Scholar 

  29. Megjhani M, Weiss M, Ford J, Terilli K, Kastenholz N, Nametz D, Kwon SB, Velazquez A, Agarwal S, Roh DJ, Conzen-Dilger C, Albanna W, Veldeman M, Connolly ES Jr, Claassen J, Aries M, Schubert GA, Park S. Optimal cerebral perfusion pressure and brain tissue oxygen in aneurysmal subarachnoid hemorrhage. Stroke. 2023;54:189–97.

    Article  PubMed  CAS  Google Scholar 

  30. Donnelly J, Czosnyka M, Harland S, Varsos GV, Cardim D, Robba C, Liu X, Ainslie PN, Smielewski P. Cerebral haemodynamics during experimental intracranial hypertension. J Cereb Blood Flow Metab. 2017;37:694–705.

    Article  PubMed  Google Scholar 

  31. Qureshi AI, Suarez JI. Use of hypertonic saline solutions in treatment of cerebral edema and intracranial hypertension. Crit Care Med. 2000;28:3301–13.

    Article  PubMed  CAS  Google Scholar 

  32. Kasprowicz M, Lalou DA, Czosnyka M, Garnett M, Czosnyka Z. Intracranial pressure, its components and cerebrospinal fluid pressure-volume compensation. Acta Neurol Scand. 2016;134:168–80.

    Article  PubMed  CAS  Google Scholar 

  33. Tang T, Weiss MD, Borum P, Turovets S, Tucker D, Sadleir R. In vivo quantification of intraventricular hemorrhage in a neonatal piglet model using an EEG-layout based electrical impedance tomography array. Physiol Meas. 2016;37:751–64.

    Article  PubMed  PubMed Central  Google Scholar 

  34. Changa AR, Czeisler BM, Lord AS. Management of elevated intracranial pressure: a review. Curr Neurol Neurosci Rep. 2019;19:99.

    Article  PubMed  Google Scholar 

  35. Rajajee V. Noninvasive intracranial pressure monitoring: are we there yet? Neurocrit Care. 2024;41:332–8.

    Article  PubMed  PubMed Central  Google Scholar 

  36. Canac N, Jalaleddini K, Thorpe SG, Thibeault CM, Hamilton RB. Review: pathophysiology of intracranial hypertension and noninvasive intracranial pressure monitoring. Fluids Barriers CNS. 2020;17:40.

    Article  PubMed  PubMed Central  Google Scholar 

  37. Arslan D, Yıldızdaş D, Horoz Ö, Aslan N, İncecik F. Evaluation of the relationship between NIRS (near-infrared spectroscopy) and optic nerve sheath diameter measurement in children with increased intracranial pressure: a pilot study. Ital J Pediatr. 2021;47:88.

    Article  PubMed  PubMed Central  Google Scholar 

  38. Bertuetti R, Gritti P, Pelosi P, Robba C. How to use cerebral ultrasound in the ICU. Minerva Anestesiol. 2020;86:327–40.

    Article  PubMed  Google Scholar 

  39. Zweifel C, Castellani G, Czosnyka M, Helmy A, Manktelow A, Carrera E, Brady KM, Hutchinson PJ, Menon DK, Pickard JD, Smielewski P. Noninvasive monitoring of cerebrovascular reactivity with near infrared spectroscopy in head-injured patients. J Neurotrauma. 2010;27:1951–8.

    Article  PubMed  Google Scholar 

  40. Saqqur M, Zygun D, Demchuk A. Role of transcranial Doppler in neurocritical care. Crit Care Med. 2007;35:S216–23.

    Article  PubMed  Google Scholar 

Download references

Acknowledgements

We would like to thank the support from the Department of Biomedical Engineering of the Air Force Medical University and the Experimental Surgery Department of Xijing Hospital.

Funding

This work was supported by the Key Research and Development Projects of the Science and Technology Committee (2022YFC2404803); the Key Basic Research Projects of the Basic Strengthening Plan of the Science and Technology Committee (2019-JCJQ-ZD-115-00-02); Innovation Foundation for Doctor Dissertation of Northwestern Polytechnical University under grant (CX2023088).

Author information

Author notes

  1. Ming-xu Zhu, Jun-yao Li, and Zhan-xiu Cai contributed equally.

Authors and Affiliations

  1. Department of Biomedical Engineering, Air Force Medical University, Xi’an, China

    Ming-xu Zhu, Jun-yao Li, Wei-ce Wang, Yi-tong Guo, Xue-tao Shi & Wei-chen Li

  2. Shaanxi Provincial Key Laboratory of Bioelectromagnetic Detection and Intelligent Perception, Department of Biomedical Engineering, Air Force Medical University, Xi’an, China

    Ming-xu Zhu, Jun-yao Li, Wei-ce Wang, Yi-tong Guo & Xue-tao Shi

  3. College of Life Science and Technology, Shandong Second Medical University, Weifang, China

    Zhan-xiu Cai

  4. Institute of Medical Research, Northwestern Polytechnical University, Xi’an, China

    Yu Wang

  5. Faculty of Electrical and Control Engineering, Liaoning Technical University, Huludao, China

    Guo-bin Gao

  6. Department of Neurosurgery, Xijing Hospital, Air Force Medical University, Xi’an, China

    Qing-dong Guo

  7. Department of Radiology, Functional and Molecular Imaging Key Lab of Shaanxi Province, Tangdu Hospital, Air Force Medical University, Xi’an, China

    Wei-chen Li

Authors
  1. Ming-xu Zhu
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Jun-yao Li
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Zhan-xiu Cai
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. Yu Wang
    View author publications

    You can also search for this author inPubMed Google Scholar

  5. Wei-ce Wang
    View author publications

    You can also search for this author inPubMed Google Scholar

  6. Yi-tong Guo
    View author publications

    You can also search for this author inPubMed Google Scholar

  7. Guo-bin Gao
    View author publications

    You can also search for this author inPubMed Google Scholar

  8. Qing-dong Guo
    View author publications

    You can also search for this author inPubMed Google Scholar

  9. Xue-tao Shi
    View author publications

    You can also search for this author inPubMed Google Scholar

  10. Wei-chen Li
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

WL, XS and MZ conceived and designed the study; MZ, JL, YW, WW and GG performed the experiment; MZ, JL and ZC analyzed the data, and drafted the manuscript; WL, XS, QG, ZC and YG edited and revised the manuscript; all the authors approved the final version of manuscript. All authors read and approved the final manuscript.

Corresponding authors

Correspondence to Qing-dong Guo, Xue-tao Shi or Wei-chen Li.

Ethics declarations

Ethics approval and consent to participate

All experimental procedures and investigations performed during this study were ethically approved by the Animal Research Ethics Committee of the Air Force Medical University and conducted based on its guidelines on animal experiments (Ethical permission number: IACUC-20241299).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Zhu, Mx., Li, Jy., Cai, Zx. et al. A novel method for detecting intracranial pressure changes by monitoring cerebral perfusion via electrical impedance tomography. Fluids Barriers CNS 22, 10 (2025). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12987-025-00619-y

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12987-025-00619-y

Keywords

Fluids and Barriers of the CNS

ISSN: 2045-8118