Search

Cross-subject electroencephalogram (EEG) based seizure subtype classification is very important in precise epilepsy diagnostics. Deep learning is a promising solution, due to its ability to automatically extract latent patterns. However, it usually requires a large amount of training data, which may not always be available in clinical practice. This paper proposes Multi-Branch Mutual-Distillation (MBMD) Transformer for cross-subject EEG-based seizure subtype classification, which can be effectively trained from small labeled data. MBMD Transformer replaces all even-numbered encoder blocks of the vanilla Vision Transformer by our designed multi-branch encoder blocks. A mutual-distillation strategy is proposed to transfer knowledge between the raw EEG data and its wavelets of different frequency bands. Experiments on two public EEG datasets demonstrated that our proposed MBMD Transformer outperformed several traditional machine learning and state-of-the-art deep learning approaches. To our knowledge, this is the first work on knowledge distillation for EEG-based seizure subtype classification. [ABSTRACT FROM AUTHOR]

Epilepsy is a pervasive neurological disorder affecting approximately 50 million individuals worldwide. Electroencephalogram (EEG) based seizure subtype classification plays a crucial role in epilepsy diagnosis and treatment. However, automatic seizure subtype classification faces at least two challenges: 1) class imbalance, i.e., certain seizure types are considerably less common than others; and 2) no a priori knowledge integration, so that a large number of labeled EEG samples are needed to train a machine learning model, particularly, deep learning. This paper proposes two novel Mixture of Experts (MoE) models, Seizure-MoE and Mix-MoE, for EEG-based seizure subtype classification. Particularly, Mix-MoE adequately addresses the above two challenges: 1) it introduces a novel imbalanced sampler to address significant class imbalance; and 2) it incorporates a priori knowledge of manual EEG features into the deep neural network to improve the classification performance. Experiments on two public datasets demonstrated that the proposed Seizure-MoE and Mix-MoE outperformed multiple existing approaches in cross-subject EEG-based seizure subtype classification. Our proposed MoE models may also be easily extended to other EEG classification problems with severe class imbalance, e.g., sleep stage classification. [ABSTRACT FROM AUTHOR]

Objective: Monitoring hemodynamic and vascular changes in the acute stages of mouse stroke models is invaluable in studying ischemic stroke pathophysiology. However, there lacks a tool to simultaneously and dynamically investigate these changes. Methods: We integrated laser speckle imaging (LSI) and visible-light optical coherence tomography (Vis-OCT) to reveal dynamic vascular responses in acute stages in the distal middle cerebral artery occlusion (dMCAO) model in rodents. LSI provides full-field, real-time imaging to guide Vis-OCT imaging and monitor the dynamic cerebral blood flow (CBF). Vis-OCT offers depth-resolved angiography and oxygen saturation (sO2) measurements. Results: Our results showed detailed CBF and vasculature changes before, during, and after dMCAO. After dMCAO, we observed insignificant sO2 variation in arteries and arterioles and location–dependent sO2 drop in veins and venules. We observed that higher branch-order veins had larger drops in sO2 at the reperfusion stage after dMCAO. Conclusion: This work suggests that integrated LSI and Vis-OCT is a promising tool for investigating ischemic stroke in mouse models. Significance: For the first time, LSI and Vis-OCT are integrated to investigate ischemic strokes in rodent models. [ABSTRACT FROM AUTHOR]

In this paper, we investigate the influence of the size distribution of magnetic nanoparticles (MNPs) on the magnetic signal and mathematical model of ac magnetic nanothermometry (MNTM). Based on the Taylor expansion of the Langevin function, two novel and real-time temperature estimation models that consider the size distribution of MNPs are proposed. These models use the summation and ratio of odd-harmonic amplitudes of signal induced by MNPs in the differential pick-up coil, respectively. The simulation results show that there is a large difference between the harmonic amplitudes of monosized and multisized MNPs. The experiments are performed under low-frequency (117 Hz) and weak (30 Oe) sinusoidal ac magnetic field. Therefore, the influence of magnetic relaxations and truncation error on MNTM can be ignored. Experimental results show that the maximum temperature estimation error of model I is about 0.63 K with a standard deviation of 0.21 K, whereas that of model II is about 0.60 K with a standard deviation of 0.32 K in the measurement time of 1 s. In addition, the problem of relative temperature sensitivity and noise for these two models are discussed. [ABSTRACT FROM AUTHOR]

In this study, we present a novel noninvasive and real-time magnetic nanothermometry (MNTM) using both odd and even harmonics of magnetic nanoparticles (MNPs) under low frequency (f = 117 Hz) ac and dc magnetic fields. Based on the first-order Langevin function, the mathematical models of weighted amplitude summation of odd and even harmonics of MNPs are proposed. Antinoise ability and sensitivity of MNTM determined by saturation magnetization, Ms, and particle diameter, D, are discussed in detail to provide guidance for the synthesis of temperature-sensing MNPs applied in different temperature ranges. We have found that the method presented in this study has better antinoise ability, and MNPs (Ms = 477 kA/m) with diameters of approximately 25 nm are more suitable for temperature probing in physiological temperature range. Experimental result shows that the maximum temperature measurement error of this novel method is below 0.28 K with a standard deviation of 0.09 K in 1 s measurement. [ABSTRACT FROM PUBLISHER]

This paper reports on a new system of magnetic nanoparticle spectrometer to measure the magnetization spectrum of a magnetic nanoparticle based sample induced in alternating magnetic field with low frequency ( f = 192 Hz) for noninvasive magnetic nanothermometry. The achieved magnetic nanothermometer employs the measured fundamental and third harmonics in ac magnetic field (fundamental and second harmonics in ac plus dc magnetic fields), as well as the Fourier series of the first-order Langevin function, to calculate temperature. Our experimental result shows that the maximum error of temperature probing is about 0.67 K with a standard deviation of 0.29 K in ac magnetic field whereas it is about 0.48 K with a standard deviation of 0.19 K in ac plus dc magnetic fields with response time of 1 s. Furthermore, experimental results indicate that an increase in response time from 1 to 8 s improves the accuracy of temperature probing from 0.29 to 0.08 K in ac magnetic field (from 0.19 to 0.06 K in ac plus dc magnetic fields). In addition, the dependence of temperature probing accuracy on measurement signal-to-noise ratio of magnetic nanoparticle magnetization spectrum is discussed by comparing the measured second and third harmonics, which are induced in ac + dc and ac magnetic fields, respectively. [ABSTRACT FROM AUTHOR]

Optical coherence tomography (OCT) oximetry explores the possibility to measure retinal hemoglobin oxygen saturation level (sO2). We investigated the accuracy of OCT retinal oximetry using Monte Carlo simulation in a commonly used four-layer retinal model. After we determined the appropriate number of simulated photon packets, we studied the effects of blood vessel diameter, signal sampling position, physiological sO2 level, and the blood packing factor on the accuracy of sO2 estimation in OCT retinal oximetry. The simulation results showed that a packing factor between 0.2 and 0.4 yields a reasonably accurate estimation of sO2 within a 5% error tolerance, which is independent of vessel diameter and sampling position, when visible-light illumination is used in OCT. We further explored the optimal optical spectral range for OCT retinal oximetry. The simulation results suggest that visible spectral range around 560 nm is better suited than near-infrared spectral range around 800 nm for OCT oximetry to warrant accurate measurements. [ABSTRACT FROM AUTHOR]

In this paper, we study the computational modeling of electromagnetically induced heating in magnetic fluid hyperthermia. Owing to the Brownian rotation and Neel relaxation of induced magnetic moments, ferrofluids can generate heat when exposed to an alternating current magnetic field. To destroy all tumors cells while preventing deleterious physiological responses, input parameters such as the frequency and intensity of magnetic fields and the complex susceptibility of ferrofluids should be determined precisely. In this paper, a solution to Maxwell’s equation for a model of a tumor and its neighboring tissues are coupled as input to Penne’s bioheat equation. Both sets of equations are solved using the finite element analysis method with perfectly matched layers for isothermal boundary conditions in COMSOL. We use a bilayered spherical model with blood perfusion and metabolism to simulate the temperature distribution in tumor regions during hyperthermia therapy. Power density due to electromagnetic field simulation serves as input to the bioheat transfer equation and determines the heat generated by the ferrofluids. The obtained results indicate that tumor regions are heated without adversely affecting healthy regions. [ABSTRACT FROM AUTHOR]

The method to determine the size distribution of nanosized ferrofluid by optimizing the discretization strategy used in calculations is demostrated in this paper. The discretization strategy is achieved with a modified optimal quantization theory to eventually reduce the ill-posed property of the magnetization numerical model. The optimization then enables the suppression of artificial oscillation found in previous estimation. The suppression of oscillations caused by the ill-posed feature can accurately resolve dimers in ferrofluid. The suppression of oscillations caused by the ill-posed feature can accurately resolve dimers in ferrofluid. [ABSTRACT FROM PUBLISHER]