Increase In Signal-to-noise Ratio Research Articles

Compartment-based reconstruction of 3D acquisition-weighted 31 P cardiac magnetic resonance spectroscopic imaging at 7 T: A reproducibility study.

Even at 7 T, cardiac 31 P magnetic resonance spectroscopic imaging (MRSI) is fundamentally limited by low signal-to-noise ratio (SNR), leading to long scan times and poor temporal and spatial resolutions. Compartment-based reconstruction algorithms such as magnetic resonance spectroscopy with linear algebraic modeling (SLAM) and spectral localization by imaging (SLIM) may improve SNR or reduce scan time without changes to acquisition. Here, we compare the repeatability and SNR performance of these compartment-based methods, applied to three different acquisition schemes at 7 T. Twelve healthy volunteers were scanned twice. Each scan session consisted of a 6.5-min 3D acquisition-weighted (AW) cardiac 31 P phase encode-based MRSI acquisition and two 6.5-min truncated k-space acquisitions with increased averaging (4 × 4 × 4 central k-space phase encodes and fractional SLAM [fSLAM] optimized k-space phase encodes). Spectra were reconstructed using (i) AW Fourier reconstruction; (ii) AW SLAM; (iii) AW SLIM; (iv) 4 × 4 × 4 SLAM; (v) 4 × 4 × 4 SLIM; and (vi) fSLAM acquisition-reconstruction combinations. The phosphocreatine-to-adenosine triphosphate (PCr/ATP) ratio, the PCr SNR, and spatial response functions were computed, in addition to coefficients of reproducibility and variability. Using the compartment-based reconstruction algorithms with the AW 31 P acquisition resulted in a significant increase in SNR compared with previously published Fourier-based MRSI reconstruction methods while maintaining the measured PCr/ATP ratio and improving interscan reproducibility. The alternative acquisition strategies with truncated k-space performed no better than the common AW approach. Compartment-based spectroscopy approaches provide an attractive reconstruction method for cardiac 31 P spectroscopy at 7 T, improving reproducibility and SNR without the need for a dedicated k-space sampling strategy.

Usefulness Evaluation of Water as an Oral Contrast Medium and Virtual Monoenergetic Images in Liver CT

The usefulness of intake water as an oral contrast medium and MonoE(Virtual Monoenergetic Images) 50 keV of DECT(Dual Energy Computed Tomography) was evaluated in Liver CT. In order to compare water intake image and non-water intake image and MonoE 50 keV image of non-water intake data were obtained from 50 patients who underwent liver follow-up CT at our hospital from January 1, 2021 to June 31, 2022 at our hospital, using IQon Spectral CT (Philips Healthcare, Netherlands). In the same patient, compared to non-water intake image, both Water intake image and MonoE 50 keV image of non-water intake showed an increase in the average values of HU of portal vein, liver parenchyma, SNR, CNR. As a result of quantitative analysis using the paired-sample t-test of SPSS (Statistical Package For Social Science, IBM, US), showed a statistically significant increase in portal vein, liver parenchyma, SNR, CNR in all phases except for the liver parenchyma of Artery Phase. In particular, in the case of MonoE 50 keV, it was confirmed that SNR and CNR increased significantly. Both water intake as an oral contrast and MonoE 50 keV of DECT showed significant increases in SNR and CNR compared to when non-water intake. If MonoE 50 keV using DECT is applied to patients who are restricted water intake and difficult to intake due to poor general conditions, it is believed that the safety and convenience of patients will be increased, and the satisfaction of CT will be improved.

MaReSPS for energy efficient spectral precoding technique in large scale MIMO-OFDM

Spectral pre-coding is a capable method to restrain Out-Of-Band Emission (OOBE) and act in accordance with leaking parameters over neighboring frequency channels while masking unnecessary emissions. Nevertheless, spectral pre-coding might deform the real data vector that is articulated as the Error Vector Magnitude (EVM), which shows a harmful effect on the performance of Multiple-Input Multiple-Output-Orthogonal Frequency Division Multiplexing (MIMO-OFDM)-oriented schemes. In this research, a new Mapper Reducer for spectral pre-coded signal (MaReSPS) for energy-constrained signal receiver is proposed for energy efficient spectral precoding in the MIMO-OFDM system. This model involves Mapper Reducer (MR) framework for detecting the received signal, which renders an error rate, and graceful degradation is observed in the throughput under channel uncertainty. The proposed scheme alleviates the resultant Transmit EVM (TxEVM) observed at the receiver by capitalizing on the massive MIMO system, and as a result the throughput is improved. The comparison is done with respect to Block Error Rate (BLER), throughput, and Power Spectral Density (PSD) for proving the betterment of the proposed precoding model for MIMO-OFDM. In particular, the normalized throughput for conventional No OOBE Reduction (OOBER), Mask Compliant Spectral Pre-coder (MSP), Notching Spectral Pre-coder + Zero Forcing (NSP + ZF), P1/P2: CVX and P1/P2: Top- Alternating Direction Method of Multipliers (ADMM) models, as well as proposed MaReSPS model, is lower at a Signal to Noise Ratio (SNR) from 0 dB to 15 dB. With an increase in SNR, the normalized throughput increases and when SNR =40 dB, the normalized throughput values reach their peak values. However, compared to existing models, the proposed MaReSPS model showed high normalized throughput.

The distortions of the free water model for diffusion MRI data when assuming single compartment relaxometry and proton density

Objective. To document the bias of the simplified free water model of diffusion MRI (dMRI) signal vis-à-vis a specific model which, in addition to diffusion, incorporates compartment-specific proton density (PD), T1 recovery during repetition time (TR), and T2 decay during echo time (TE). Approach. Both models assume that volume fraction f of the total signal in any voxel arises from the free water compartment (fw) such as cerebrospinal fluid or edema, and the remainder (1-f) from hindered water (hw) which is constrained by cellular structures such as white matter (WM). The specific and simplified models are compared on a synthetic dataset, using a range of PD, T1 and T2 values. We then fit the models to an in vivo healthy brain dMRI dataset. For both synthetic and in vivo data we use experimentally feasible TR, TE, signal-to-noise ratio (SNR) and physiologically plausible diffusion profiles. Main results. From the simulations we see that the difference between the estimated simplified f and specific f is largest for mid-range ground-truth f, and it increases as SNR increases. The estimation of volume fraction f is sensitive to the choice of model, simplified or specific, but the estimated diffusion parameters are robust to small perturbations in the simulation. Specific f is more accurate and precise than simplified f. In the white matter (WM) regions of the in vivo images, specific f is lower than simplified f. Significance. In dMRI models for free water, accounting for compartment specific PD, T1 and T2, in addition to diffusion, improves the estimation of model parameters. This extra model specification attenuates the estimation bias of compartmental volume fraction without affecting the estimation of other diffusion parameters.

Channel capacity and power allocation of MIMO visible light communication system

In this paper, the channel capacity of the multiple-input multiple-output (MIMO) visible light communication (VLC) system is investigated under the peak, average optical and electrical power constraints. Finding the channel capacity of MIMO VLC is shown to be a mixed integer programming problem. To address this open problem, we propose an inexact gradient projection method to find the channel capacity-achieving discrete input distribution and the channel capacity of MIMO VLC. Also we derive both upper and lower bounds of the capacity of MIMO VLC with the closed-form expressions. Furthermore, by considering practical discrete constellation inputs, we develop the optimal power allocation scheme to maximize transmission rate of MIMO VLC system. Simulation results show that more discrete points are needed to achieve the channel capacity as SNR increases. Both the upper and lower bounds of channel capacity are tight at low SNR region. In addition, comparing the equal power allocation, the proposed power allocation scheme can significantly increase the rate for the low-order modulation inputs.

Acquisition Time and Probabilities of Detection and False Alarm in Direct Sequence Code Division Multiple Access Systems

Multiple access interference (MAI) is one of the most important issues to consider in the design, implementation and operation of 5G systems using code division multiple access (CDMA) technique based on spread spectrum that offers a more flexible solution, compared to other techniques. Especially, it permits a much higher user bit rate, due to the bandwidth allocated to the emitted signal; thus, allowing very attractive multimedia services. In this paper, we develop the expressions of the detection probability, false alarm probability and the average acquisition time - in a direct sequence CDMA (DS-CDMA) transmission system - using a serial search and taking into consideration MAI. The evolution of these parameters according to the signal- to-noise ratio (SNR) and the decision threshold leads to study performances of the system in terms of detection and average acquisition time. The obtained results reveal that the acquisition time decreases when the SNR increases. Additionally, it is found that the probability of detection increases when the threshold decreases, and increasing SNR it reaches a constant value for an SNR/chip around 0 dB for all thresholds.

High signal-to-noise-ratio ultrasonic imaging of crack defects in particles filled composite explosives

<sec>Polymer bonded explosive (PBX) is a kind of composite material with highly filled molding explosive particles (normally more than 95%) and a small quantity of binders (less than 5%). The effective detection of internal cracks in PBX is of great significance in evaluating structural integrity and safety reliability.</sec><sec>Ultrasonic phased array detection and imaging methods show great advantages and potential in detecting crack defects. But acoustic test results indicate that the PBX has unique characteristics with low longitudinal wave velocity (～3000 m·s<sup>–1</sup>) and strong attenuation (attenuation coefficient ～400 dB·m<sup>–1</sup> for 2.5 MHz ultrasound). When the defect is imaged by traditional ultrasonic total focusing method (TFM), the structural noises at the boundaries between particles lead to low signal-to-noise ratio (SNR) in the FMC signals and strong background noise in reconstructed image, which will disturb the detection of cracks.</sec><sec>To realize the high SNR imaging of crack defects in PBX, an ultrasonic imaging algorithm based on baseband nonlinear synthetic focusing (BB-NSF) is proposed. By utilizing the spatial coherence of the received signals in full matrix capture (FMC) data, the pixel intensity at defect position can be enhanced while the background noise can be drastically weakened. The delay rule of the algorithm is modified according to the characteristics of PBX surface configuration. In this way, the high SNR imaging of crack defects with different orientations of PBX surface configuration is realized, and the quality of the reconstructed images is compared and evaluated quantitatively. Meanwhile, the base band transformation in calculation process optimization could significantly reduce calculation burden and increase imaging efficiency.</sec><sec>Experimental results show that the proposed algorithm can effectively suppress background noise and significantly improve the ability to detect the PBX cracks. The effective suppression to background noise makes the defect more highlighting and distinguished easily. For the BB-NSF algorithm, spatial coherence coefficient <i>p</i> is a crucial parameter used for dynamically regulating the SNR of reconstructed image. When <i>p</i> value is more than 2.0, the SNR of the ultrasonic reconstructed image of PBX crack defect is improved by more than 10 dB compared with that of the traditional linear synthetic focusing imaging. With the increase of <i>p</i> value, the SNR is further improved, while the calculation efficiency for a single image is almost kept stable. Moreover, the increase of SNR to some extent will improve the far-field detect capability.</sec><sec>Besides, with the BB-NSF algorithm, flexible transducer inhibits different imaging characteristics of for cracks with different orientations and depths in curved PBX specimens. For defects with large orientation angle and buried depth, the tip, root and shape of cracks can be completely present. For defects with small orientation angle and buried depth, part of shape and contour features will be lost.</sec><sec>In conclusion, the BB-NSF algorithm shows the advantage of high SNR and calculation efficiency in imaging PBX cracks, and exhibits great application prospect in imaging internal defects of other strongly attenuated composites.</sec>

Effect of DNN Approximation for Channel Estimation and Signal Detection on OFDM Applications

This paper offers a deep learning approximation to realize channel estimation and signal detection that creates the main communication structure skeleton for the orthogonal frequency-division multiplexing (OFDM) system known as an efficient modulation type on 5G. This letter offers an application of deep learning to handle the wireless OFDM channels' end-to-end conduct. First, channel state information (CSI) is predicted explicitly that differs from existing OFDM receivers, then detected the transmitted symbols utilizing the predicted CSI. In the end, CSI is predicted by the suggested deep learning approximation indirectly and transmitted symbols are directly recovered. The structure of the designed receiver occurs of a layer of DNN and soft decisions, which resolves the issues channel estimation error, time delay, and limitation of decoding between users in classic detection techniques. In the simulation results, it is observed that the receiver has powerful stability on the power distribution of user, not only convenient for the linear channel, but also for nonlinear channel when enhancement the number of users, also detection can be well on the receiver. Generally, the efficiency of the modulation system decreases with the features of the multipath channel utilized for transmission. Channel estimation and detection of symbols utilize to reduce the impacts of the channel, which needs high computation and bandwidth conventionally. This paper is used deep neural networks (DNN) for detecting the signal, in this way much effort in detecting the channel is prevented. The proposed method saves priceless bandwidth via used CP in OFDM with a big increase in SNR.

Further Research on Frequency Allocation (19-Cell System) Method based on IFR and FFR

Many of the most advanced technology has a basic need for efficient telecommunication. In order to achieve this goal, the team should allocate those frequency channels in an intelligent way so that the interference can be reduced to the maximum extent. Here the team introduced a new method based on IFR3&IFR1 (Integer frequency reuse) in a 19-cell network, but also apply the FFR (fractional frequency reuse) method to reduce the interference ulteriorly. In the pure IFR situation, the integer frequency reuse factor and the cellular network specification collectively decide the network capacity. While in FFR only a fraction of the whole bandwidth is used by the users in the edge. By contrast, the users near the BS can make use of the total bandwidth. The cell is divided into an outer and an inner part in this effective way. Fractional Frequency Reuse (FFR) and its derivative method are accepted widely in the downlink (DL) inter-cell interference coordination(ICIC) schemes. The team also optimized the FFR method by measuring the distance in a more precise way. The team utilized Matlab to make a simulation of the system applying the new method and come to a couple of reasonable conclusions by comparing the FFR+IFR3 with FFR+IFR3. FFR+IFR3 has a better performance against FFR+IFR1 in a system whose distance between each cell is with more accuracy. While the average number of users per cell and transmission power for cell edge have little impact on average capacity. When the SNR increases, the network capacity starts steep but later flattens. Our method is more intelligent and can minish interference efficiently.

Impact of deep learning image reconstructions (DLIR) on coronary artery calcium quantification

BackgroundDeep learning image reconstructions (DLIR) have been recently introduced as an alternative to filtered back projection (FBP) and iterative reconstruction (IR) algorithms for computed tomography (CT) image reconstruction. The aim of this study was to evaluate the effect of DLIR on image quality and quantification of coronary artery calcium (CAC) in comparison to FBP.MethodsOne hundred patients were consecutively enrolled. Image quality–associated variables (noise, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR)) as well as CAC-derived parameters (Agatston score, mass, and volume) were calculated from images reconstructed by using FBP and three different strengths of DLIR (low (DLIR_L), medium (DLIR_M), and high (DLIR_H)). Patients were stratified into 4 risk categories according to the Coronary Artery Calcium - Data and Reporting System (CAC-DRS) classification: 0 Agatston score (very low risk), 1–99 Agatston score (mildly increased risk), Agatston 100–299 (moderately increased risk), and ≥ 300 Agatston score (moderately-to-severely increased risk).ResultsIn comparison to standard FBP, increasing strength of DLIR was associated with a significant and progressive decrease of image noise (p < 0.001) alongside a significant and progressive increase of both SNR and CNR (p < 0.001). The use of incremental levels of DLIR was associated with a significant decrease of Agatston CAC score and CAC volume (p < 0.001), while mass score remained unchanged when compared to FBP (p = 0.232). The underestimation of Agatston CAC led to a CAC-DRS misclassification rate of 8%.ConclusionDLIR systematically underestimates Agatston CAC score. Therefore, DLIR should be used cautiously for cardiovascular risk assessment.Key Points• In coronary artery calcium imaging, the implementation of deep learning image reconstructions improves image quality, by decreasing the level of image noise.• Deep learning image reconstructions systematically underestimate Agatston coronary artery calcium score.• Deep learning image reconstructions should be used cautiously in clinical routine to measure Agatston coronary artery calcium score for cardiovascular risk assessment.

Tissue optimization strategies for high-quality ex vivo diffusion imaging.

Ex vivo diffusion imaging can be used to study healthy and pathological tissue microstructure in the rodent brain with high resolution, providing a link between in vivo MRI and ex vivo microscopy techniques. Major challenges for the successful acquisition of ex vivo diffusion imaging data however are changes in the relaxivity and diffusivity of brain tissue following perfusion fixation. In this study we address this question by examining the combined effects of tissue preparation factors that influence signal-to-noise ratio (SNR) and consequently image quality, including fixative concentration, contrast agent concentration and tissue rehydration time. We present an optimization strategy combining these factors to manipulate the and of fixed tissue and maximize SNR efficiency. We apply this strategy in the rat brain, for a diffusion-weighted spin echo protocol with TE = 27 ms on a 9.4 T scanner with a 39 mm volume coil and 660 mT/m 114 mm gradient insert. We used a reduced fixative concentration of 2% paraformaldehyde (PFA), rehydration time more than 20 days, 15 mM Gd-DTPA in perfusate and TR 250 ms. This resulted in a doubling of SNR and an increase in SNR per unit time of 135% in cortical grey matter and 88% in white matter compared with 4% PFA and no contrast agent. This improved SNR efficiency enabled the acquisition of excellent-quality high-resolution (78 m isotropic voxel size) diffusion data with b = 4000 s/mm , 30 diffusion directions and a field of view of mm3 in less than 4 days. It was also possible to achieve comparable data quality for a standard resolution (150 m) diffusion dataset in h. In conclusion, the tissue optimization strategy presented here may be used to improve SNR, increase spatial resolution and/or allow faster acquisitions in preclinical ex vivo diffusion MRI experiments.

Seeing is believing: ultra-high field magnetic resonance imaging in vascular and neurodegeneration research

Ultra-high field (UHF) magnetic resonance (MR) refers to a main operating field strength of 7 teslas (T) or higher. UHF MR offers unique opportunities for revealing new insights into the microstructures and functions of the brain that are not available on conventional field strength MR. Recently, the U.S. Food and Drug Administration (FDA) has cleared the 7 T magnetic resonance imaging (MRI) device for clinical use, which has doubled the 7 T systems installed in the US. Central to this powerful evolution is the concomitant increase of signal-to-noise ratio (SNR) and susceptibility contrast at UHF MRI. Compared to 3 T, 7 T offers a 2 to 3-fold increase in image SNR, which will allow us to image high resolution with voxel volume of less than 0.1 mm3, while the largely increased susceptibility or T2* contrast at ultra-high field MRI will be the key benefit for brain imaging. Sequences sensitive to magnetic susceptibility include susceptibility-weighted imaging (SWI) and quantitative susceptibility mapping (QSM). Thanks to the susceptibility-induced blooming effect, small veins at a micro level (e.g. &lt;50 µm) can be seen on 220 x 220 µm matrix size SWI.1 With the strengths being provided on UHF MR, several key clinical benefits are discussed. The perivenous lesion development or central vein sign in multiple sclerosis (MS) has been revealed in vivo for the first time,2 which has proved useful for a differential diagnosis of MS.3 Seven teslas MRI showed improved detection of cortical MS lesions that are consistent with histopathological patterns of cortical demyelinating lesions regarding their cortical layer involvement that are not seen on 3 T MRI.4 The visibility of small venous architecture, including small cortical and medullary veins, has an important implication for tissue oxygen utilization and neuronal degeneration.5 By introducing an ultrasmall-superparamagnetic-iron-oxide (USPIO) contrast agent (e.g., ferumoxytol), the small arterioles can also be seen on 7 T SWI, which allows the detection of small arterial tortuosity in the elderly, likely the major cause of small vessel disease. Seven teslas markedly increased the sensitivity in detecting iron deposition and microbleeds associated with brain structures and lesions, and this provides critical insights on brain pathology, in particular in age-related neurodegenerative diseases. Seven teslas MRI improves the characterization of age-related choroid plexus degeneration, including both volume and signal changes, which may have an impact on waste clearance and cerebrospinal fluid (CSF) production. Figure 1 demonstrates a few representative examples of the potential of 7 T. In summary, seeing the unseen on 7 T provides opportunities not only for an early diagnosis, but also for clinical research of disease mechanisms and treatment strategy development.

High SNR rapid T1 -weighted MPRAGE using spiral imaging with long readouts and improved deblurring.

The goal of this work is to present the implementation of 3D spiral high-resolution MPRAGE and to demonstrate that SNR and scan efficiency increase with the increment of readout time. Simplified signal equations for MPRAGE indicate that the T1 contrast can be kept approximately the same by a simple relationship between the flip angle and the TR. Furthermore, if T1 contrast remains the same, image SNR depends on the square root of the product of the total scan time and the readout time. MPRAGE spiral sequences were implemented with distributed spirals and spiral staircase on 3 Tesla scanners. Brain images of threevolunteers were acquired with different readout times. Spiral images were processed with a joint water-fat separation and deblurring algorithm and compared to Cartesian images. Pure noise data sets were also acquired for SNR evaluation. Consistent T1 weighting can be achieved with various spiral readout lengths, and between spiral MPRAGE imaging and the traditional Cartesian MPRAGE imaging. Noise performance analysis demonstrates higher SNR efficiency of spiral MPRAGE imaging with matched T1 contrast compared to the Cartesian reference imaging. Fast, high SNR MPRAGE imaging is feasible with long readout spiral trajectories.

Challenges regarding MR compatibility of an MRgFUS robotic system

Numerous challenges are faced when employing Magnetic Resonance guided Focused Ultrasound (MRgFUS) hardware in the Magnetic Resonance Imaging (MRI) setting. The current study aimed to provide insights on this topic through a series of experiments performed in the framework of evaluating the MRI compatibility of an MRgFUS robotic device. All experiments were performed in a 1.5 T MRI scanner. The main metric for MRI compatibility assessment was the signal to noise ratio (SNR). Measurements were carried out in a tissue mimicking phantom and freshly excised pork tissue under various activation states of the system. In the effort to minimize magnetic interference and image distortion, various set-up parameters were examined. Significant SNR degradation and image distortion occurred when the FUS transducer was activated mainly owing to FUS-induced target and coil vibrations and was getting worse as the output power was increased. Proper design and stable positioning of the imaged phantom play a critical role in reducing these vibrations. Moreover, isolation of the phantom from the imaging coil was proven essential for avoiding FUS-induced vibrations from being transferred to the coil during sonication and resulted in a more than 3-fold increase in SNR. The use of a multi-channel coil increased the SNR by up to 50 % compared to a single-channel coil. Placement of the electronics outside the coil detection area increased the SNR by about 65 %. A similar SNR improvement was observed when the encoders' counting pulses were deactivated. Overall, this study raises awareness about major challenges regarding operation of an MRgFUS system in the MRI environment and proposes simple measures that could mitigate the impact of noise sources so that the monitoring value of MR imaging in FUS applications is not compromised.

Estimation of the Number of Signal Sources Based on Hard Thresholding

Aiming at the problem of large error in so far as low SNR and limited number of snapshots in the actual techniques for evaluation the amount of signal sources, a method for evaluation the amount of signal sources ground on the hard threshold method is proposed. The method works by comparing the eigenvalues of the received signal covariance matrix with a hard threshold. And use the distance betwixt the compared eigenvalues and the eigenvalues of the ideal ideal data as the criterion, and the amount of sources identified is ground on the minimum distance criterion. The simulation consequences display that the approach can effectively guess at the amount of sources under the situation of low SNR and small snapshot number, and the estimation is consistent with the increase of SNR and snapshot number.

The effect of parameters on the signal compensation performance in the modified non-negative Tikhonov regularization method within a Bayesian frame

Recently, a modified non-negative Tikhonov regularization method within the Bayesian frame has been successfully proposed, and demonstrated to have excellent compensation performance towards the signal distortion induced by long-distance transmission in coaxial cable. However, how the parameters of this modified method affect compensation performance has not yet been investigated and analyzed, which may in turn bring some obstacles in the practical utilization of this modified method. Herein, we carried out a systematic study on the effect of parameters on the signal compensation performance of this modified method. The bias constant D has almost no influence on the compensation performance if the unknown input signal satisfies the non-negative constraint. When the signal-to-noise ratio (SNR) is less than 40 dB, the compensation performance is continuously improved as the SNR increases. Once the SNR attains ∼40 dB, the compensated signal is nearly identical to the input signal. Further increment of SNR does not give rise to an improved compensation performance and the relative error between the compensated signal and input signal remains nearly unchanged. The sampling frequency interval of S21 should satisfy the restriction of , with and representing the sampling frequency of the digital oscilloscope and number of points of the output signal, respectively. The compensation performance is excellent initially, whereas it rapidly deteriorates once exceeds a certain value that is highly related to the length of the impulse response function. This study provides a beneficial reference for application of the modified signal compensation method.

Self-supervised neural network improves tri-exponential intravoxel incoherent motion model fitting compared to least-squares fitting in non-alcoholic fatty liver disease.

Recent literature suggests that tri-exponential models may provide additional information and fit liver intravoxel incoherent motion (IVIM) data more accurately than conventional bi-exponential models. However, voxel-wise fitting of IVIM results in noisy and unreliable parameter maps. For bi-exponential IVIM, neural networks (NN) were able to produce superior parameter maps than conventional least-squares (LSQ) generated images. Hence, to improve parameter map quality of tri-exponential IVIM, we developed an unsupervised physics-informed deep neural network (IVIM3-NET). We assessed its performance in simulations and in patients with non-alcoholic fatty liver disease (NAFLD) and compared outcomes with bi-exponential LSQ and NN fits and tri-exponential LSQ fits. Scanning was performed using a 3.0T free-breathing multi-slice diffusion-weighted single-shot echo-planar imaging sequence with 18 b-values. Images were analysed for visual quality, comparing the bi- and tri-exponential IVIM models for LSQ fits and NN fits using parameter-map signal-to-noise ratios (SNR) and adjusted R 2. IVIM parameters were compared to histological fibrosis, disease activity and steatosis grades. Parameter map quality improved with bi- and tri-exponential NN approaches, with a significant increase in average parameter-map SNR from 3.38 to 5.59 and 2.45 to 4.01 for bi- and tri-exponential LSQ and NN models respectively. In 33 out of 36 patients, the tri-exponential model exhibited higher adjusted R 2 values than the bi-exponential model. Correlating IVIM data to liver histology showed that the bi- and tri-exponential NN outperformed both LSQ models for the majority of IVIM parameters (10 out of 15 significant correlations). Overall, our results support the use of a tri-exponential IVIM model in NAFLD. We show that the IVIM3-NET can be used to improve image quality compared to a tri-exponential LSQ fit and provides promising correlations with histopathology similar to the bi-exponential neural network fit, while generating potentially complementary additional parameters.

Estimation and Compensation of IQ Imbalance for OTSM Systems

Recently, orthogonal time sequency multiplexing (OTSM) waveform was proposed using Walsh Hadamard transform (WHT) for doubly-dispersive channels at a lower modulation/demodulation complexity than orthogonal time frequency space (OTFS) waveform. At high frequencies, hardware imperfections such as in-phase and quadrature (IQ) imbalance is prominent at the transceiver. In this letter, the input-output (I/O) relation in the delay-sequency domain is derived under these impairments at the receiver. The impairment parameters are estimated in the delay-time domain and are compensated in the time domain using proposed dual pilot approach. The normalized mean square error (NMSE) of the estimated imbalance parameters is evaluated and it is observed that NMSE reduces with increase in SNR and with reduction in mobility. The results are evaluated with OTFS system and it is observed that the NMSE performance of OTSM is similar to OTFS under high pilot power and high SNR at no excess pilot power. The bit error rate performance of the compensated system is evaluated and its performance is close to the ideal system.

Convolutional neural network transformer (CNNT) for free-breathing real-time cine imaging

Abstract Funding Acknowledgements Type of funding sources: None. Background Real-time cine imaging does not require breath-holding and is a robust cine imaging technique in the presence of irregular heartbeats. It is a good alternative to the conventional breath-hold retro-gated cine for simplified acquisition and improved patient comfort. Real-time acquisition is achieved with the single-shot BSSFP readout without retro-gating. To maintain good temporal and spatial resolution, higher acceleration (e.g. &amp;gt;4x parallel imaging) is required. As a result, the real-time cine images experience reduced signal-to-noise ratio (SNR), which limits its clinical acceptance. Purpose We developed a novel deep learning model architecture, the Convolutional Neural Network Transformer (CNNT), to improve the quality of real-time cine, under 4x, 5x and 6x acceleration. Method Convolutional Neural Networks (CNN) are widely used in CMR research to process cardiac images. Cardiac images are often acquired as a time series with strong inter-phase correlation. We combined the CNN with the more recent transformer model to develop a novel CNNT architecture. It takes in the entire 2D+T time series as input and has advantages of CNN for efficient computation and spatial invariance. It further inherits the advantages of attention layer in the transformer and is able to efficiently utilize the temporal correlation within a time series. A CNNT model is developed to improve the SNR of real-time cine imaging. N=10 patients were scanned at a heart center, with 4x, 5x and 6x acceleration. Typical imaging parameters are: FOV 360×270mm2, flip angle 50°, acquired matrix size 160×90 for R=4 acceleration, 192×108 for R=5 and 6, temporal resolution 40ms for R=4, 42ms for R=5 and 35ms for R=6. The real-time images went through a TGRAPPA reconstruction [1] and the CNNT model. The SNR of TGRAPPA was measured with SNR units [2]. The Monte-Carlo pseudo-replica test was used to measure SNR for the CNNT model. For every cine series, two phases were picked for the end-systole and end-diastole. For every image picked, two region-of-interests were drawn in the myocardium and in the LV blood pool. The CNNT model was deployed inline on the MR scanner using the Gadgetron InlineAI [3]. Results Figure 1 gives real-time cine images for three accelerations, reconstructed with TGRAPPA and CNNT. The parallel imaging TGRAPPA reconstruction suffers significant SNR loss from elevated g-factor and less acquired data. The deep learning CNNT model recovered SNR even at the very high 6x acceleration, without observed loss of boundary sharpness. Table 1 lists the SNR measurement results. The TGRAPPA SNR decreased ∼4x from R=4 to R=6 for both the blood and myocardium. For the blood, the CNNT increased the SNR by 170%, 335%, 371% at R=4, 5 and 6. For the myocardium, the SNR increases were 335%, 634% and 828%. Conclusion We developed a convolutional neural network transformer model to recover the SNR for real-time cine imaging at higher acceleration.

The importance of ocular artifact removal in single-trial ERP analysis: The case of the N250 in face learning

ObjectiveSingle-trial event-related potentials (ERPs) offer fine-grained information about the trajectories of the neurocognitive processes but are highly sensitive to any artifacts in the EEG signal. The primary aim of this study was to assess the impact of ocular artifact removal on the single-trial N250 ERP analysis of face learning in individual participants. MethodsWe present a detailed description of our research-grade EEG hardware setup and a highly reproducible code (https://osf.io/aqhmn/) for generating time series of single-trial N250 ERP amplitudes and precise identification of a changepoint between face memory trace acquisition and maintenance. Ocular artifacts were removed using a new semi-automatic approach with only one hyperparameter based on the correlation between EEG components from independent component analysis (ICA) and the EOG signal. ResultsResults from the simulation study showed that our ocular artifact filtration decreased the average RMSE by half and achieved the highest increase of SNR among all the compared methods. It decreased standard deviations and improved the fit of the broken-line regression models for all participants by 25% ± 17% (min. 2%, max. 63%). Conclusions and significanceOcular artifact filtration had a substantial positive impact on the regression modeling of single-trial ERP amplitudes. Lack of ocular artifact removal can drastically distort the conclusions about the face learning process from single-trial N250 ERP experiments for individual participants. The changepoint locations changed for 13 out of 15 participants. This is the first published analysis of time series of single-trial N250 ERP amplitudes in face learning.