Multi-frequency symmetry difference electrical impedance tomography with machine learning for human stroke diagnosis

Abstract

Multi-Frequency Symmetry Difference Electrical Impedance Tomography (MFSD-EIT) can robustly detect and identify unilateral perturbations in symmetric scenes. Here, an investigation is performed to assess if the algorithm can be successfully applied to identify the aetiology of stroke with the aid of machine learning. Methods: Anatomically realistic fourlayer Finite Element Method models of the head based on stroke patient images are developed and used to generate EIT data over a 5 Hz -100 Hz frequency range with and without bleed and clot lesions present. Reconstruction generates conductivity maps of each head at each frequency. Application of a quantitative metric assessing changes in symmetry across the sagittal plane of the reconstructed image and over the frequency range allows lesion detection and identification. The algorithm is applied to both simulated and human (n=34 subjects) data. A classification algorithm is applied to the metric value in order to differentiate between normal, haemorrhage and clot values. Results: An average accuracy of 85% is achieved when MFSD-EIT with Support Vector Machines (SVM) classification is used to identify and differentiate bleed from clot in human data, with 77% accuracy when differentiating normal from stroke in human data. Conclusion: Applying a classification algorithm to metrics derived from MFSD-EIT images is a novel and promising technique for detection and identification of perturbations in static scenes. Significance: The MFSD-EIT algorithm used with machine learning gives promising results of lesion detection and identification in challenging conditions like stroke. The results imply feasible translation to human patients.

Neural networks for classification of strokes in electrical impedance tomography on a 3D head model

“…We chose to use neural networks (NN) since they clearly outperformed kernel methods for this specific kind of nonlinear dataset in our preliminary numerical tests. Also, the performances of a support vector machine (SVM) for brain stroke classification was already studied in [32,34].…”

“…Also, only a finite set of head shapes is considered and the model lacks the ability to generalize to new sample heads. The more recent work [32] considers a 4-layer model for the heads and uses data from 18 human patients [13] that are classified using SVM. The main difference with our method is that our classification is made directly from raw electrode data, while in [32] a preprocessing step involving a precise knowledge of the anatomy of the patient's head is required.…”

“…The more recent work [32] considers a 4-layer model for the heads and uses data from 18 human patients [13] that are classified using SVM. The main difference with our method is that our classification is made directly from raw electrode data, while in [32] a preprocessing step involving a precise knowledge of the anatomy of the patient's head is required. Moreover, only strokes of size 20 ml or 50 ml are considered in four specific locations, while our datasets include strokes with volume as small as 1.5 ml located anywhere within the brain tissue.…”

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

“…EIT is fast, low-cost, portable, non-intrusive, label-free and radiation-free, making it an up-and-coming candidate in biomedical imaging. Emerging applications include functional lung imaging [2], stroke diagnosis [3], [4], and biological tissue imaging [5], [6]. As the impedance spectra of biological tissues is frequency-dependent, differences in electrical properties between various tissues can be exploited to benefit physiological and pathological diagnostics for tissue differentiation, early cancer detection, and tumor or stroke imaging [6], [7].…”

Target Adaptive Differential Iterative Reconstruction (TADI): A Robust Algorithm for Real-Time Electrical Impedance Tomography

“…Moreover, the regularization parameter selection criteria proposed thus far can be divided into two categories based on (i) prior information such as noise error level, and (ii) a-posteriori parameter selection criterion matching the original data with the inclusion of qualitative or quantitative information to the solution [16]. In context, the unbiased predictive risk estimator (UPRE) is a statistical estimation method based on the quadratic norm of prediction error, assuming that the error is independently and identically distributed with the knowledge of noise variance [17]. The equivalent degrees of freedom consider the linear dependency between the observed and theoretical solutions, assuming that the residual norm is a Chi-square distribution with known noise variance and degrees of freedom [18].…”