Multi-Frequency Electromagnetic Tomography for Acute Stroke Detection Using Frequency-Constrained Sparse Bayesian Learning

Biomechanical simulations of intracerebral hemorrhage expansion show tissue displacement has significant impact on electrical impedance tomography results

OBJECTIVE

Intracerebral hemorrhage (ICH) occupies intracranial space and causes brain tissue displacement and fluid volume shifts. We assess how hematoma expansion (HE) affects electrical impedance tomography (EIT) measurements and reconstructed images of the conductivity change caused by HE.

METHODS

We developed a novel multi-physics model of ICH with mechanical tissue deformation during HE. We simulated EIT measurements with the multi-physics model and a traditional static model using five ICH locations. The effects of tissue deformation on the results of monitoring of ICH with EIT were assessed by comparing the measurement data from the multi-physics and traditional models and by comparing the corresponding reconstructed conductivity change from two image reconstruction algorithms.

RESULTS

The simulated measurement data and the reconstructed images of the conductivity change using the multi-physics and the traditional model are radically different regardless of the image reconstruction algorithm used.

CONCLUSIONS

The effect of tissue displacement caused by HE on EIT monitoring of ICH is significant. Specifically, the displacement of cerebrospinal fluid (CSF) can mask the effects of increased ICH blood volume. However, the effects of displaced CSF could be easier to detect with EIT than the ICH blood volume increase and thus could be used as an indicator of HE in EIT bedside monitoring of ICH and improve the detectability of HE, especially for ICH located deep in the brain.

SIGNIFICANCE

Currently there are virtually no imaging methods for continuous monitoring of stroke. There has been recent resurgence in interest to develop electrical impedance tomography (EIT) devices and algorithms for monitoring progression of stroke. In-silico studies show promising results, but there are very little clinical results. In-silico models are usually used for development and evaluation of algorithms for EIT image reconstruction. In previous studies the stroke has been usually modeled as local change in electrical conductivity and the fluid and tissue displacement caused by the increased blood volume in ICH has not been considered. In this paper we present a novel multi-physics model of ICH, simulated EIT measurement results and reconstructed images with comparisons to the traditionally used ICH modeling methods. Our multi-physics approach to modeling of ICH shows that the effect of tissue and fluid displacement during HE needs consideration when developing clinical applications of EIT.

Multi-Path Fusion in SFCF-Net for Enhanced Multi-Frequency Electrical Impedance Tomography

Multi-frequency electrical impedance tomography (mfEIT) offers a nondestructive imaging technology that reconstructs the distribution of electrical characteristics within a subject based on the impedance spectral differences among biological tissues. However, the technology faces challenges in imaging multi-class lesion targets when the conductivity of background tissues is frequency-dependent. To address these issues, we propose a spatial-frequency cross-fusion network (SFCF-Net) imaging algorithm, built on a multi-path fusion structure. This algorithm uses multi-path structures and hyper-dense connections to capture both spatial and frequency correlations between multi-frequency conductivity images, which achieves differential imaging for lesion targets of multiple categories through cross-fusion of information. According to both simulation and physical experiment results, the proposed SFCF-Net algorithm shows an excellent performance in terms of lesion imaging and category discrimination compared to the weighted frequency-difference, U-Net, and MMV-Net algorithms. The proposed algorithm enhances the ability of mfEIT to simultaneously obtain both structural and spectral information from the tissue being examined and improves the accuracy and reliability of mfEIT, opening new avenues for its application in clinical diagnostics and treatment monitoring.

MMV-Net: A Multiple Measurement Vector Network for Multifrequency Electrical Impedance Tomography

Multifrequency electrical impedance tomography (mfEIT) is an emerging biomedical imaging modality to reveal frequency-dependent conductivity distributions in biomedical applications. Conventional model-based image reconstruction methods suffer from low spatial resolution, unconstrained frequency correlation, and high computational cost. Deep learning has been extensively applied in solving the EIT inverse problem in biomedical and industrial process imaging. However, most existing learning-based approaches deal with the single-frequency setup, which is inefficient and ineffective when extended to the multifrequency setup. This article presents a multiple measurement vector (MMV) model-based learning algorithm named MMV-Net to solve the mfEIT image reconstruction problem. MMV-Net considers the correlations between mfEIT images and unfolds the update steps of the Alternating Direction Method of Multipliers for the MMV problem (MMV-ADMM). The nonlinear shrinkage operator associated with the weighted l2,1 regularization term of MMV-ADMM is generalized in MMV-Net with a cascade of a Spatial Self-Attention module and a Convolutional Long Short-Term Memory (ConvLSTM) module to better capture intrafrequency and interfrequency dependencies. The proposed MMV-Net was validated on our Edinburgh mfEIT Dataset and a series of comprehensive experiments. The results show superior image quality, convergence performance, noise robustness, and computational efficiency against the conventional MMV-ADMM and the state-of-the-art deep learning methods.

Real-time imaging of traumatic brain injury using magnetic induction tomography

Objective. Early diagnosis of traumatic brain injury (TBI) is crucial for its prognosis; however, traditional computed tomography diagnostic methods rely on large medical devices with an associated lag time to receive results. Therefore, an imaging modality is needed that provides real-time monitoring, can easily be carried out to assess the extent of TBI damage, and thus guides treatment.Approach. In the present study, an improved magnetic induction tomography (MIT) data acquisition system was used to monitor TBI in an animal model and distinguish the injury level. A pneumatically controlled cortical impactor was used to strike the parietal lobe of anesthetized rabbits two or three times under the same parameter mode to establish two different rabbit models of TBI. The MIT data acquisition system was used to record data and continuously monitor the brain for one hour without intervention.Main results. A target with increased conductivity was clearly observed in the reconstructed image. The position was relatively fixed and accurate, and the average positioning error of the image was 0.013 72 m. The normalized mean reconstruction value of all images increased with time. The slope of the regression line of the normalized mean reconstruction value differed significantly between the two models (p< 0.0001).Significance. This indicates that in the animal model, the unique features of MIT may facilitate the early monitoring of TBI and distinguish different degrees of injuries, thereby reducing the risk and mortality of associated complications.

Advances of deep learning in electrical impedance tomography image reconstruction

Electrical impedance tomography (EIT) has been widely used in biomedical research because of its advantages of real-time imaging and nature of being non-invasive and radiation-free. Additionally, it can reconstruct the distribution or changes in electrical properties in the sensing area. Recently, with the significant advancements in the use of deep learning in intelligent medical imaging, EIT image reconstruction based on deep learning has received considerable attention. This study introduces the basic principles of EIT and summarizes the application progress of deep learning in EIT image reconstruction with regards to three aspects: a single network reconstruction, deep learning combined with traditional algorithm reconstruction, and multiple network hybrid reconstruction. In future, optimizing the datasets may be the main challenge in applying deep learning for EIT image reconstruction. Adopting a better network structure, focusing on the joint reconstruction of EIT and traditional algorithms, and using multimodal deep learning-based EIT may be the solution to existing problems. In general, deep learning offers a fresh approach for improving the performance of EIT image reconstruction and could be the foundation for building an intelligent integrated EIT diagnostic system in the future.

A portable non-invasive microwave based head imaging system using compact metamaterial loaded 3D unidirectional antenna for stroke detection

A metamaterial (MTM) loaded compact three-dimensional antenna is presented for the portable, low-cost, non-invasive microwave head imaging system. The antenna has two slotted dipole elements with finite arrays of MTM unit cell and a folded parasitic patch that attains directional radiation patterns with 80% of fractional bandwidth. The operating frequency of the antenna is 1.95–4.5 GHz. The optimization of MTM unit cell is performed to increase the operational bandwidth, realized gain, and efficiency of the antenna within the frequency regime. It is also explored to improve radiation efficiency and gain when placed to head proximity. One-dimensional mathematical modelling is analyzed to precisely estimate the power distribution that validates the performance of the proposed antenna. To verify the imaging capability of the proposed system, an array of 9 antennas and a realistic three-dimensional tissue-emulating experimental semi-solid head phantom are fabricated and measured. The backscattered signal is collected from different antenna positions and processed by the updated Iterative Correction of Coherence Factor Delay-Multiply-and-Sum beamforming algorithm to reconstruct the hemorrhage images. The reconstructed images in simulation and experimental environment demonstrate the feasibility of the proposed system as a portable platform to successfully detect and locate the hemorrhages inside the brain.

Adaptively Regularized Bases-Expansion Subspace Optimization Methods for Electrical Impedance Tomography

OBJECTIVE

In this work, to deal with the difficulties in choosing regularization weighting parameters and low spatial resolution problems in difference electrical impedance tomography (EIT), we propose two adaptively regularized bases-expansion subspace optimization methods (AR-BE-SOMs).

METHODS

Firstly, an adaptive L1-norm based total variation (TV) regularization is introduced under the framework of BE-SOM. Secondly, besides the additive regularization, an adaptive weighted L2-norm multiplicative regularization is further proposed. The regularized objective functions are optimized by conjugate gradient method, where the unknowns in both methods are update alternatively between induced contrast current (ICC) and conductivity domain.

CONCLUSION

Both numerical and experimental tests are conducted to validate the proposed methods, where AR-BE-SOMs show better edge-preserving, anti-noise performance, lower relative errors, and higher structure similarity indexes than BE-SOM.

SIGNIFICANCE

Unlike the common regularization techniques in EIT, the proposed regularization factors can be obtained adaptively during the optimization process. More importantly, ARBE-SOMs perform well in reconstructions of some challenging geometry with sharp corners such as the heart and lung phantoms, deformation phantoms, triangles and even rectangles. It is expected that the proposed AR-BE-SOMs will find their applications for high-quality lung health monitoring and other clinical applications.

Adaptive threshold split Bregman algorithm based on magnetic induction tomography for brain injury monitoring imaging

Objective. Traditional magnetic induction tomography (MIT) algorithms have problems in reconstruction, such as large area error (AE), blurred boundaries of reconstructed targets, and considerable image noise (IN). As the size and boundary of a lesion greatly affect the treatment plan, more accurate algorithms are necessary to meet clinical needs.Approach. In this study, adaptive threshold split Bregman (ATSB) is proposed for brain injury monitoring imaging in MIT. We established a 3D brain MIT simulation model with the actual anatomical structure and a phantom model and obtained the reconstructed images of single targets in different positions and multiple targets, using the Tikhonov, eigenvalue threshold regularisation (ETR), split Bregman (SB), and ATSB algorithms.Main results. Compared with the Tikhonov and ETR algorithms, the ATSB algorithm reduced the AE by 95% and the IN by 17% in a simulation and reduced the AE by 87% and IN by 6% in phantom experiments. Compared with the SB algorithm, the ATSB algorithm can reduce the difficulty of adjusting parameters and is easier to use in clinical practice. The simulation and phantom experiments results showed that the ATSB algorithm could reconstruct the target size more accurately and could distinguish multiple targets more effectively than the other three algorithms.Significance. The ATSB algorithm could improve the image quality of MIT and better meet the needs of clinical applications and is expected to promote brain injury monitoring imaging via MIT.

FISTA-Net: Learning a Fast Iterative Shrinkage Thresholding Network for Inverse Problems in Imaging

Inverse problems are essential to imaging applications. In this letter, we propose a model-based deep learning network, named FISTA-Net, by combining the merits of interpretability and generality of the model-based Fast Iterative Shrinkage/Thresholding Algorithm (FISTA) and strong regularization and tuning-free advantages of the data-driven neural network. By unfolding the FISTA into a deep network, the architecture of FISTA-Net consists of multiple gradient descent, proximal mapping, and momentum modules in cascade. Different from FISTA, the gradient matrix in FISTA-Net can be updated during iteration and a proximal operator network is developed for nonlinear thresholding which can be learned through end-to-end training. Key parameters of FISTA-Net including the gradient step size, thresholding value and momentum scalar are tuning-free and learned from training data rather than hand-crafted. We further impose positive and monotonous constraints on these parameters to ensure they converge properly. The experimental results, evaluated both visually and quantitatively, show that the FISTA-Net can optimize parameters for different imaging tasks, i.e. Electromagnetic Tomography (EMT) and X-ray Computational Tomography (X-ray CT). It outperforms the state-of-the-art model-based and deep learning methods and exhibits good generalization ability over other competitive learning-based approaches under different noise levels.