Multilinear Singular Value Decomposition Research Articles

Patient specific higher order tensor based approach for the detection and localization of myocardial infarction using 12-lead ECG

Myocardial Infarction (MI) is an emergency condition that requires immediate medical treatment. The rapid and accurate diagnosis of MI using a 12-lead electrocardiogram (ECG) is extremely important in a clinical study to save the patient’s life. The manual interpretation of MI using a 12-lead ECG is tedious and time-consuming. Therefore, a patient-specific software-based computer-aided diagnosis framework is helpful to detect and localize MI disease accurately. This paper proposes a patient-specific higher-order tensor-based approach to detect and localize MI automatically using 12-lead ECG recordings. The 12-lead ECG recordings are segmented into 12-lead ECG beats using the multi-lead fusion-based QRS detection algorithm. The fast and adaptive multivariate empirical mode decomposition (FA-MVEMD) based multiscale analysis method decomposes 12-lead ECG beat into a third-order tensor containing the information from the samples, beat, and intrinsic mode functions (IMFs). Furthermore, a fourth-order tensor is formulated by considering beats, samples, lead, and IMFs information of 12-lead ECG recording. The multilinear singular value decomposition (MLSVD) extracts features from the fourth-order tensors and third-order tensors of 12-lead ECG. The K-nearest neighbor (KNN), support vector machine (SVM), and stacked autoencoder-based deep neural network (SAE-DNN) models are used for the detection and localization of MI using fourth-order and third-order tensor domain features. The proposed approach is evaluated using 73 healthy control (HC) and 100 different types of MI-based 12-lead ECG recordings from a public database. The proposed approach has obtained the classification accuracy values of (98.84%, 98.27%, 98.27%) and (86.64%, 83.17%, and 81.98%) using (KNN, SVM, and SAE-DNN) models for MI detection, and localization, respectively using 30-min duration of 12-lead ECG recordings. For MI detection and localization, the suggested approach has obtained accuracy values of 96.53% and 93.32%, respectively, using the 4-s duration of 12-lead ECG recordings. Our approach outperformed existing MI detection and localization methods using 12-lead ECG recordings regarding classification performance.

Development and application of computational methods for the study of protein dynamics with Pm-HMGR as a model system.

Continuous atomistic descriptions of enzyme mechanisms and atomistic dynamics are crucial to the characterization of protein function and dysfunction—a common cause of disease—but unattainable with current techniques, necessitating novel methods which leverage computational-experimental synergy. Computational studies can provide a wealth of information, but carry little weight without experimental validation. Thus, coupling time-resolved crystallographic studies of enzyme mechanisms with computational mechanistic investigation through molecular dynamics (MD) simulation and predictive model generation provides the opportunity to combine approaches. We demonstrate and refine a novel Markov State-informed Multilinear Singular Value Decomposition (MSiMSVD) method with the Pseudomonas mevalonii HMG CoA Reductase (PmHMGR) pathway as a model system. MSiMSVD couples time-resolved crystallographic studies of enzyme mechanisms with computational mechanistic investigation through MD simulation and predictive model generation. Application of the MSiMSVD method to slow dynamical events-such as the PmHMGR 2nd hydride transfer-is limited by the ability of force fields to reproduce atomistic behavior at a transition state. This need can be met via the generation of a Transition State Force Field (TSFF) with a program such as Quantum-guided Molecular Mechanics (Q2MM). However, a frequently encountered problem in the optimization of TSFFs for biomolecules is the high dimensionality of the optimization surface producing local rather than global minima. To address this, we implement a constrained particle swarm algorithm hybridized with a basic genetic algorithm to automate the optimization of TSFFs with Q2MM. The Q2MM and the MSiMSVD methods will become accessible to non-experts for a wide range of systems, such as large biomolecules. Thus far, initial Markov State Models generated from MD simulation of the PmHMGR 2nd hydride transfer indicate that more sampling is necessary to cover the transition state conformation, i.e. the structure of the activated complex.

Tensor-Based Underdetermined Blind Identification of Instantaneous Mixtures

In order to explore the effective statistical information of observed signals to stack into a tensor, while reduce the computational complexity and speed up the convergence, a new tensor-based underdetermined blind identification method of instantaneous mixture is proposed in this brief. Firstly, the autocovariance matrix of each observed sub-block is calculated to construct the symmetric third-order tensor according to segmentation scheme. Secondly, the original tensor is reduced into a low-rank kernel tensor by truncated multi-linear singular value decomposition (MLSVD). Finally, an enhanced line search (ELS) technique is applied to speed up the convergence of alternating least squares (ALS) to identify the mixing matrix. The simulation results indicate that the presented approach provides a superior estimated performance compared to the existing methods in both real and complex scenarios.

Frequency Estimation Enhancement for Industrial Free Induction Decay Signals Under Low SNR via Hankelization and Modified Covariance

The frequency of a free induction decay (FID) signal from an industrial proton precession magnetometer (PPM) is proportional to the magnetic field strength. To achieve high-precision frequency estimation for an FID signal with a low signal-to-noise ratio (SNR), a long estimation period is always required which limits the application scenarios of the PPMs, such as aeromagnetic detection. To break through the contradiction between frequency estimation precision and time, this paper proposes an enhanced frequency estimation method via Han-kelization and modified covariance, dubbed EFE-HMC. First, the collected one-dimensional FID signal is constructed as a third-order tensor through Hankelization, and then the noise tensors are removed by multi-linear singular value decomposition to improve the SNR; second, the pre-processed third-order tensor is inverted into a one-dimensional signal, and the modified covariance is employed to estimate the corresponding signal frequency; and third, an experimental test platform is constructed to compare the proposed EFE-HMC with three state-of-the-art methods including Carry Chain, equal precision, and Dn-ResUnet. The results demonstrate that when the SNR is lower than -10 dB and the estimation time varies from 50 ms to 200 ms, the frequency estimation precision is improved by 60% on average. Moreover, the frequency estimation time is reduced by about 150 ms when the frequency estimation precision is better than 0.03 Hz.

Near Lossless Compression for 3D Radiological Images Using Optimal Multilinear Singular Value Decomposition (3D-VOI-OMLSVD).

Storage and transmission of high-compression 3D radiological images that create high-quality reconstruction upon decompression are critical necessities for effective and efficient teleradiology. To cater to this need, we propose a near lossless 3D image volume compression method based on optimal multilinear singular value decomposition called “3D-VOI-OMLSVD.” The proposed strategy first eliminates any blank 2D image slices from the 3D image volume and uses the selective bounding volume (SBV) to identify and extract the volume of Interest (VOI). Following this, the VOI is decomposed with an optimal multilinear singular value decomposition (OMLSVD) to obtain the corresponding core tensor, factor matrices, and singular values that are compressed with adaptive binary range coder (ABRC), integrated as an entropy encoder. The compressed file can be transferred or transmitted and then decompressed in order to reconstruct the original image. The resultant decompressed VOI is acquired by reversing the above process and then fusing it with the background, using the bound volume coordinates associated with the compressed 3D image. The proposed method performance was tested on a variety of 3D radiological images with different imaging modalities and dimensions using quantitative evaluation metrics such as the compression rate (CR), bit rate (BR), peak signal to noise ratio (PSNR), and structural similarity index (SSIM). Furthermore, we also investigate the impact of VOI extraction on the model performance, before comparing it with two popular compression methods, namely JPEG and JPEG2000. Our proposed method, 3D-VOI-OMLSVD, displayed a high CR value, with a maximum of 37.31, and a low BR, with the lowest reported to be 0.21. The SSIM score was consistently high, with an average performance of 0.9868, while using < 1 second for decoding the image. We observe that with VOI extraction, the compression rate increases manifold, and bit rate drops significantly, and thus reduces the encoding and decoding time to a great extent. Compared to JPEG and JPEG2000, our method consistently performs better in terms of higher CR and lower BR. The results indicate that the proposed compression methodology performs consistently to create high-quality image compressions, and overall gives a better outcome when compared against two state-of-the-art and widely used methods, JPEG and JPEG2000.

From multilinear SVD to multilinear UTV decomposition

Across a range of applications, low multilinear rank approximation (LMLRA) is used to compress large tensors into a more compact form, while preserving most of their information. A specific instance of LMLRA is the multilinear singular value decomposition (MLSVD), which can be used for multilinear principal component analysis (MLPCA). MLSVDs are obtained by computing SVDs of all tensor unfoldings, but, in practical applications, it is often not necessary to compute full SVDs. In this article, we therefore propose a new decomposition, called the truncated multilinear UTV decomposition (TMLUTVD). This is a tensor decomposition that is also multilinear rank-revealing, yet less expensive to compute than a truncated MLSVD (TMLSVD); it can even be computed in a finite number of steps. We present its properties in an algorithm-independent manner. In particular, we derive bounds on the accuracy in function of the truncation level. Experiments illustrate the good performance in practice.

Block Row Kronecker-Structured Linear Systems With a Low-Rank Tensor Solution

Several problems in compressed sensing and randomized tensor decomposition can be formulated as a structured linear system with a constrained tensor as the solution. In particular, we consider block row Kronecker-structured linear systems with a low multilinear rank multilinear singular value decomposition, a low-rank canonical polyadic decomposition or a low tensor train rank tensor train constrained solution. In this paper, we provide algorithms that serve as tools for finding such solutions for a large, higher-order data tensor, given Kronecker-structured linear combinations of its entries. Consistent with the literature on compressed sensing, the number of linear combinations of entries needed to find a constrained solution is far smaller than the corresponding total number of entries in the original tensor. We derive conditions under which a multilinear singular value decomposition, canonical polyadic decomposition or tensor train solution can be retrieved from this type of structured linear systems and also derive the corresponding generic conditions. Finally, we validate our algorithms by comparing them to related randomized tensor decomposition algorithms and by reconstructing a hyperspectral image from compressed measurements.

Updating the Multilinear UTV Decomposition

Low multilinear rank approximation (LMLRA) is a basic tool for compressing a tensor into a more compact form, while preserving most of its information. An LMLRA of an <inline-formula><tex-math notation="LaTeX">$N$</tex-math></inline-formula>th order tensor consists of <inline-formula><tex-math notation="LaTeX">$N$</tex-math></inline-formula> factor matrices which are combined using a core tensor of smaller dimensions than the original tensor. When the tensor is non-static, for example when over time new tensor slices are being added along a certain mode, one can expect that the LMLRA of the tensor changes, but that an approximation of the new tensor can be derived from the previous one. By updating the factor matrices and the core tensor in an efficient way when a new tensor slice is added, this can indeed be achieved. In this paper, a new method is proposed to track the factor matrices in both the evolving and non-evolving modes, leading to efficient and accurate updates of the LMLRA. We track a truncated multilinear UTV decomposition (TMLUTVD), which is a tensor decomposition that is multilinear rank-revealing, yet less expensive to compute than the popular multilinear singular value decomposition. We derive accuracy bounds for the tracked TMLUTVD and give strategies to speed up its computation. Experiments show that the TMLUTVD tracking method works well in practice, even if the mode-<inline-formula><tex-math notation="LaTeX">$n$</tex-math></inline-formula> principal subspaces change drastically between updates. As an illustration of the possibilities of the technique, we extend the use of LMLRA in a compression step, prior to the computation of a canonical polyadic decomposition (CPD), to the setting of updating and tracking.

Sensitivity schemes for dynamic glucose-enhanced magnetic resonance imaging to detect glucose uptake and clearance in mouse brain at 3T.

We investigated three dynamic glucose-enhanced (DGE) MRI methods for sensitively monitoring glucose uptake and clearance in both brain parenchyma and cerebrospinal fluid (CSF) at clinical field strength (3T). By comparing three sequences, namely, Carr-Purcell-Meiboom-Gill (CPMG), on-resonance variable delay multipulse (onVDMP), and on-resonance spin-lock (onSL), a high-sensitivity DGE MRI scheme with truncated multilinear singular value decomposition (MLSVD) denoising was proposed. The CPMG method showed the highest sensitivity in detecting the parenchymal DGE signal among the three methods, while both onVDMP and onSL were more robust for CSF DGE imaging. Here, onVDMP was applied for CSF imaging, as it displayed the best stability of the DGE results in this study. The truncated MLSVD denoising method was incorporated to further improve the sensitivity. The proposed DGE MRI scheme was examined in mouse brain with 50%/25%/12.5% w/w D-glucose injections. The results showed that this combination could detect DGE signal changes from the brain parenchyma and CSF with as low as a 12.5% w/w D-glucose injection. The proposed DGE MRI schemes could sensitively detect the glucose signal change from brain parenchyma and CSF after D-glucose injection at a clinically relevant concentration, demonstrating high potential for clinical translation.

Whole-brain amide CEST imaging at 3T with a steady-state radial MRI acquisition.

To develop a steady-state saturation with radial readout chemical exchange saturation transfer (starCEST) for acquiring CEST images at 3 Tesla (T). The polynomial Lorentzian line-shape fitting approach was further developed for extracting amideCEST intensities at this field. StarCEST MRI using periodically rotated overlapping parallel lines with enhanced reconstruction-based spatial sampling was implemented to acquire Z-spectra that are robust to brain motion. Multi-linear singular value decomposition postprocessing was applied to enhance the CEST SNR. The egg white phantom studies were performed at 3T to reveal the contributions to the 3.5 ppm CEST signal. Based on the phantom validation, the amideCEST peak was quantified using the polynomial Lorentzian line-shape fitting, which exploits the inverse relationship between Z-spectral intensity and the longitudinal relaxation rate in the rotating frame. The 3D turbo spin echo CEST was also performed to compare with the starCEST method. The amideCEST peak showed a negligible peak B1 dependence between 1.2 µT and 2.4 µT. The amideCEST images acquired with starCEST showed much improved image quality, SNR, and motion robustness compared to the conventional 3D turbo spin echo CEST method with the same scan time. The amideCEST contrast extracted by the polynomial Lorentzian line-shape fitting method trended toward a stronger gray matter signal (1.32% ± 0.30%) than white matter (0.92% ± 0.08%; P = .02, n = 5). When calculating the magnetization transfer contrast and T1 -corrected rotating frame relaxation rate maps, amideCEST again was not significantly different for white matter and gray matter. Rapid multi-slice amideCEST mapping can be achieved by the starCEST method (< 5 min) at 3T by combing with the polynomial Lorentzian line-shape fitting method.

Improving the Signal-to-Noise-Ratio of Free Induction Decay Signals Using a New Multilinear Singular Value Decomposition-Based Filter

The free induction decay (FID) signal output by a proton precession magnetometer (PPM) is usually only of the microvolt level, and its frequency is proportional to the magnetic field strength. Therefore, obtaining a high signal-to-noise ratio (SNR) FID signal is crucial for improving the measurement accuracy of the magnetometer. The current gold standards for noise reduction in FID signals—singular value decomposition (SVD) and principal component analysis (PCA)—still have limited denoising capabilities, especially in cases with strong noise interference. In this study, a new noise-reduction algorithm for FID based on multilinear SVD (MLSVD) is proposed. First, equal delay-based multichannel data sampling is used to obtain multiple correlated signals, and thus, the obtained multiple signals are constructed as a third-order tensor; second, the MLSVD is employed to calculate and remove the noise singular value of the tensor; and third, canonical polyadic decomposition (CPD) is used to fuse the multichannel FID signal processed by MLSVD and eliminate signal noise, further improving the SNR. Subsequently, a PPM experimental test platform was constructed, and extensive simulation and practical comparison tests were conducted. The results show that, when the SNR is −10 dB, the noise-reduction effect of the MLSVD is about 9.12 dB higher than that of the SVD and about 8.15 dB higher than that of the PCA, with an overall increase of 28%. In an environment with strong noise interference—with an SNR of −30 dB—both PCA and SVD are no longer viable, while MLSVD can still effectively suppress noise, with a signal-to-noise improvement ratio (SNIR) being as high as 50.36 dB.

A Tensor-Based Approach Using Multilinear SVD for Hand Gesture Recognition From sEMG Signals

With the increasing number of sensors/ channels, hand gesture recognition from multiple channels has attracted the attention of the researchers in recent years. The analysis of a variety of gestures is still a challenging task in real-time applications such as in wearable devices because of the large number of channels used. This paper proposes a tensor-based approach using multilinear singular value decomposition (MLSVD) for hand gesture recognition where all available channels were used during training whereas only a single channel was used for recognition of new gestures. Tensor decompositions have very limited use in gesture recognition using surface Electromyography (sEMG) despite these signals naturally have multi-way structures. The sEMG data of different subjects were first segmented to reshape it as a fourth-order tensor of the form channels × time × gestures × subjects. The MLSVD was applied to model the tensor and extract features, which were then fed into various classifiers such as support vector machine (SVM), K-nearest neighbors (KNN), TreeBagger (TB), and dictionary learning (DL) classifiers in order to compare their performance. The proposed method was evaluated on three publicly available databases (NinaPro, CapgMyo (DB-a, DB-b and DB-c), CSL-HDEG) by performing intra-session, inter-session and inter-subject evaluations on each database. The experimental results indicated that the proposed method achieved the best accuracy (Acc) using DL, achieving Acc of 77.6% and 89.6% with CapgMyo DB-b and CSL-HDEMG databases, respectively, during inter-session evaluation, and 75.2%, 75.4%, 68.3%, and 67.7% with NinaPro, CapgMyo DB-b, CapgMyo DB-c, and CSL-HDEMG databases, respectively, during inter-subject evaluation. The proposed method performed better during both inter-session and inter-subject evaluations than the state-of-the-art methods.

Efficient update of multilinear singular value decomposition in background subtraction applications

Subtraction techniques are used to distinguish moving objects or foregrounds that are being tracked from a static background. To prevent possible local misclassifications, the subspace spanned by low-rank features computed from a multilinear singular value decomposition (MLSVD), can be used to filter out noises or gradual changes from the background. However, as it is prohibitively expensive to compute a new MLSVD from scratch at every iteration, we propose an adaptive and efficient method for updating an MLSVD by reusing previous decompositions while tracking more accurate decomposition errors. The experimental results reveal that the proposed MLSVD update algorithm exhibits a faster execution speed and better accuracy than other MLSVD update algorithms used in background subtraction applications.

High-sensitivity CEST mapping using a spatiotemporal correlation-enhanced method.

To obtain high-sensitivity CEST maps by exploiting the spatiotemporal correlation between CEST images. A postprocessing method accomplished by multilinear singular value decomposition (MLSVD) was used to enhance the CEST SNR by exploiting the correlation between the Z-spectrum for each voxel and the low-rank property of the overall CEST data. The performance of this method was evaluated using CrCEST in ischemic mouse brain at 11.7 tesla. Then, MLSVD CEST was applied to obtain Cr, amide, and amine CEST maps of the ischemic mouse brain to demonstrate its general applications. Complex-valued Gaussian noise was added to CEST k-space data to mimic a low SNR situation. MLSVD CEST analysis was able to suppress the noise, recover the degraded CEST peak, and provide better CrCEST quality compared to the smoothing and singular value decomposition (SVD)-based denoising methods. High-resolution Cr, amide, and amine CEST maps of an ischemic stroke using MLSVD CEST suggest that CrCEST is also a sensitive pH mapping method, and a wide range of pH changes can be detected by combing CrCEST with amine CEST at high magnetic fields. MLSVD CEST provides a simple and efficient way to improve the SNR of CEST images.

Multilinear Singular Value Decomposition for Millimeter Wave Channel Parameter Estimation

Fifth generation (5G) cellular standards are set to utilize millimeter wave (mmWave) frequencies, which enable data speeds greater than 10 Gbps and sub-centimeter localization accuracy. These capabilities rely on accurate estimates of the channel parameters, which we define as the angle of arrival, angle of departure, and path distance for each path between the transmitter and receiver. Estimating the channel parameters in a computationally efficient manner poses a challenge because it requires estimation of parameters from a high-dimensional measurement – particularly for multi-carrier systems since each subcarrier must be estimated separately. Additionally, channel parameter estimation must be able to handle hybrid beamforming, which uses a combination of digital and analog beamforming to reduce the number of required analog to digital converters. This paper introduces a channel parameter estimation technique based on the multilinear singular value decomposition (MSVD), a Tucker form tensor analog of the singular value decomposition, for massive multiple input multiple output (MIMO) multi-carrier systems with hybrid beamforming. The MSVD tensor estimation approach is more computationally efficient than methods such as the canonical polyadic decomposition (CPD) and the Tucker form of the MSVD enables paths to be extracted based on signal energy. The algorithms performance is compared to the CPD method and shown to closely match the Cramer-Rao bound (CRB) of channel parameter estimates through simulations. Additionally, limitations of channel parameter estimation and communication waveform effects are studied.

Multichannel Time-Frequency Complexity Measures for the Analysis of Age-Related Changes in Neuromagnetic Resting-State Activity.

We propose new multichannel time-frequency complexity measures to evaluate differences on magnetoencephalograpy (MEG) recordings between healthy young and old subjects at rest at different spatial scales. After reviewing the Rényi and singular value decomposition entropies based on time-frequency representations, we introduce multichannel generalizations, using multilinear singular value decomposition for one of them. We test these quantities on synthetic data, illustrating how the introduced complexity measures focus on number of components, nonstationarity, and similarity across channels. Friedman tests are used to confirm the differences between young and old groups, and heterogeneity within groups. Experimental results show a consistent increase in complexity measures for the old group. When analyzing the topographical distribution of complexity values, we found clusters in the frontal sensors. The complexity measures here introduced seem to be a better indicator of the neurophysiologic changes of aging than power envelope connectivity. Here, we applied new multichannel time-frequency complexity measures to resting-state MEG recordings from healthy young and old subjects. We showed that these features are able to reveal regional clusters. The multichannel time-frequency complexities can be used to monitor the aging of subjects. They also allow a mutual information approach, and could be applied to a wider range of problems.

Shape-Adaptive Tensor Factorization Model for Dimensionality Reduction of Hyperspectral Images

Tensor-based dimensionality reduction (DR) of hyperspectral images is a promising research topic. However, patch-based tensorization usually adopts a squared neighborhood with fixed window size, which may be inaccurate in modeling the local spatial information in a hyperspectral image scene. In this work, we propose a novel shape-adaptive tensor factorization (SATF) model for dimensionality reduction and classification of hyperspectral images. Firstly, shape-adaptive patch features are extracted to build fourth-order tensors. Secondly, multilinear singular value decomposition (MLSVD) is adopted for tensor factorization and latent features are extracted via mode-i tensor-matrix product. Finally, classification is conducted by using a sparse multinomial logistic regression (SMLR) model. Experimental results, conducted with two popular hyperspectral data sets collected over the Indian Pines and the University of Pavia, respectively, indicate that the proposed method outperforms the other traditional and tensor-based DR methods.

Local low rank denoising for enhanced atomic resolution imaging

Atomic resolution imaging and spectroscopy suffers from inherently low signal to noise ratios often prohibiting the interpretation of single pixels or spectra. We introduce local low rank (LLR) denoising as tool for efficient noise removal in scanning transmission electron microscopy (STEM) images and electron energy-loss (EEL) spectrum images. LLR denoising utilizes tensor decomposition techniques, in particular the multilinear singular value decomposition (MLSVD), to achieve a denoising in a general setting largely independent of the signal features and data dimension, by assuming that the signal of interest is of low rank in segments of appropriately chosen size. When applied to STEM images of graphene, LLR denoising suppresses statistical noise while retaining fine image features such as scan row-wise distortions, possibly related to rippling of the graphene sheet and consequent motion of atoms. When applied to EEL spectra, LLR denoising reveals fine structures distinguishing different lattice sites in the spinel system CoFe2O4.

Reconstruction by calibration over tensors for multi-coil multi-acquisition balanced SSFP imaging.

To develop a rapid imaging framework for balanced steady-state free precession (bSSFP) that jointly reconstructs undersampled data (by a factor of R) across multiple coils (D) and multiple acquisitions (N). To devise a multi-acquisition coil compression technique for improved computational efficiency. The bSSFP image for a given coil and acquisition is modeled to be modulated by a coil sensitivity and a bSSFP profile. The proposed reconstruction by calibration over tensors (ReCat) recovers missing data by tensor interpolation over the coil and acquisition dimensions. Coil compression is achieved using a new method based on multilinear singular value decomposition (MLCC). ReCat is compared with iterative self-consistent parallel imaging (SPIRiT) and profile encoding (PE-SSFP) reconstructions. Compared to parallel imaging or profile-encoding methods, ReCat attains sensitive depiction of high-spatial-frequency information even at higher R. In the brain, ReCat improves peak SNR (PSNR) by 1.1 ± 1.0 dB over SPIRiT and by 0.9 ± 0.3 dB over PE-SSFP (mean ± SD across subjects; average for N = 2-8, R = 8-16). Furthermore, reconstructions based on MLCC achieve 0.8 ± 0.6 dB higher PSNR compared to those based on geometric coil compression (GCC) (average for N = 2-8, R = 4-16). ReCat is a promising acceleration framework for banding-artifact-free bSSFP imaging with high image quality; and MLCC offers improved computational efficiency for tensor-based reconstructions. Magn Reson Med 79:2542-2554, 2018. © 2017 International Society for Magnetic Resonance in Medicine.

Irregular heartbeat classification using Kronecker Product Equations.

Cardiac arrhythmia or irregular heartbeats are an important feature to assess the risk on sudden cardiac death and other cardiac disorders. Automatic classification of irregular heartbeats is therefore an important part of ECG analysis. We propose a tensor-based method for single- and multi-channel irregular heartbeat classification. The method tensorizes the ECG data matrix by segmenting each signal beat-by-beat and then stacking the result into a third-order tensor with dimensions channel × time × heartbeat. We use the multilinear singular value decomposition to model the obtained tensor. Next, we formulate the classification task as the computation of a Kronecker Product Equation. We apply our method on the INCART dataset, illustrating promising results.