Transform-domain least mean squares (TDLMS) adaptive filters encompass the class of learning algorithms where the input data are subjected to a data-independent unitary transform followed by a power normalization stage as preprocessing steps. Because conventional transformations are not data-dependent, this preconditioning procedure was shown theoretically to improve the convergence of the least mean squares (LMS) filter only for certain classes of input data. So, one can tailor the transformation to the class of data. However, in reality, if the class of input data is not known beforehand, it is difficult to decide which transformation to use. Thus, there is a need to devise a learning framework to obtain such a preconditioning transformation using input data prior to applying on the input data. It is hypothesized that the underlying topology of the data affects the selection of the transformation. With the input modeled as a weighted finite graph, our method, called preconditioning using graph (PrecoG), adaptively learns the desired transform by recursive estimation of the graph Laplacian matrix. We show the efficacy of the transform as a generalized split preconditioner on a linear system of equations and in Hebbian-LMS learning models. In terms of the improvement of the condition number after applying the transformation, PrecoG performs significantly better than the existing state-of-the-art techniques that involve unitary and nonunitary transforms.

PurposeTo develop a B1‐corrrected single flip‐angle continuous acquisition strategy with free‐breathing and cardiac self‐gating for spiral T1 mapping, and compare it to a previous dual flip‐angle technique.MethodsData were continuously acquired using a spiral‐out trajectory, rotated by the golden angle in time. During the first 2 s, off‐resonance Fermi RF pulses were applied to generate a Bloch‐Siegert shift B1 map, and the subsequent data were acquired with an inversion RF pulse applied every 4 s to create a T1* map. The final T1 map was generated from the B1 and the T1* maps by using a look‐up table that accounted for slice profile effects, yielding more accurate T1 values. T1 values were compared to those from inversion recovery (IR) spin echo (phantom only), MOLLI, SAturation‐recovery single‐SHot Acquisition (SASHA), and previously proposed dual flip‐angle results. This strategy was evaluated in a phantom and 25 human subjects.ResultsThe proposed technique showed good agreement with IR spin‐echo results in the phantom experiment. For in‐vivo studies, the proposed technique and the previously proposed dual flip‐angle method were more similar to SASHA results than to MOLLI results.ConclusionsB1‐corrected single flip‐angle T1 mapping successfully acquired B1 and T1 maps in a free‐breathing, continuous‐IR spiral acquisition, providing a method with improved accuracy to measure T1 using a continuous Look‐Locker acquisition, as compared to the previously proposed dual excitation flip‐angle technique.

Purpose The synergistic use of k‐t undersampling and multiband (MB) imaging has the potential to provide extended slice coverage and high spatial resolution for first‐pass perfusion MRI. The low‐rank plus sparse (L + S) model has shown excellent performance for accelerating single‐band (SB) perfusion MRI. Methods A MB data consistency method employing ESPIRiT maps and through‐plane coil information was developed. This data consistency method was combined with the temporal L + S constraint to form the slice‐L + S method. Slice‐L + S was compared to SB L + S and the sequential operations of split slice‐GRAPPA and SB L + S (seq‐SG‐L + S) using synthetic data formed from multislice SB images. Prospectively k‐t undersampled MB data were also acquired and reconstructed using seq‐SG‐L + S and slice‐L + S. Results Using synthetic data with total acceleration rates of 6–12, slice‐L + S outperformed SB L + S and seq‐SG‐L + S (N = 7 subjects) with respect to normalized RMSE and the structural similarity index ( P < 0.05 for both). For the specific case with MB factor = 3 and rate 3 undersampling, or for SB imaging with rate 9 undersampling (N = 7 subjects), the normalized RMSE values were 0.037 ± 0.007, 0.042 ± 0.005, and 0.031 ± 0.004; and the structural similarity index values were 0.88 ± 0.03, 0.85 ± 0.03, and 0.89 ± 0.02 for SB L + S, seq‐SG‐L + S, and slice‐L + S, respectively ( P < 0.05 for both). For prospectively undersampled MB data, slice‐L + S provided better image quality than seq‐SG‐L + S for rate 6 (N = 7) and rate 9 acceleration (N = 7) as scored by blinded experts. Conclusion Slice‐L + S outperformed SB‐L + S and seq‐SG‐L + S and provides 9 slice coverage of the left ventricle with a spatial resolution of 1.5 mm × 1.5 mm with good image quality.

Seizures cause retrograde amnesia, but underlying mechanisms are poorly understood. We tested whether seizure activated neuronal circuits overlap with spatial memory engram and whether seizures saturate LTP in engram cells. A seizure caused retrograde amnesia for spatial memory task. Spatial learning and a seizure caused cFos expression and synaptic plasticity overlapping set of neurons in the CA1 of the hippocampus. Recordings from learning-labeled CA1 pyramidal neurons showed potentiated synapses. Seizure-tagged neurons were also more excitable with larger rectifying excitatory postsynaptic currents than surrounding unlabeled neurons. These neurons had enlarged dendritic spines and saturated LTP. A seizure immediately after learning, reset the memory engram. Seizures cause retrograde amnesia through shared ensembles and mechanisms.

Purpose To develop and evaluate a high spatial resolution (1.25 × 1.25 mm 2 ) spiral first‐pass myocardial perfusion imaging technique with whole‐heart coverage at 3T, to better assess transmural differences in perfusion between the endocardium and epicardium, to quantify the myocardial ischemic burden, and to improve the detection of obstructive coronary artery disease. Methods Whole‐heart high‐resolution spiral perfusion pulse sequences and corresponding motion‐compensated reconstruction techniques for both interleaved single‐slice (SS) and simultaneous multi‐slice (SMS) acquisition with or without outer‐volume suppression (OVS) were developed. The proposed techniques were evaluated in 34 healthy volunteers and 8 patients (55 data sets). SS and SMS images were reconstructed using motion‐compensated L1‐SPIRiT and SMS‐Slice‐L1‐SPIRiT, respectively. Images were blindly graded by 2 experienced cardiologists on a 5‐point scale (5, excellent; 1, poor). Results High‐quality perfusion imaging was achieved for both SS and SMS acquisitions with or without OVS. The SS technique without OVS had the highest scores (4.5 [4, 5]), which were greater than scores for SS with OVS (3.5 [3.25, 3.75], P < .05), MB = 2 without OVS (3.75 [3.25, 4], P < .05), and MB = 2 with OVS (3.75 [2.75, 4], P < .05), but significantly higher than those for MB = 3 without OVS (4 [4, 4], P = .95). SMS image quality was improved using SMS‐Slice‐L1‐SPIRiT as compared to SMS‐L1‐SPIRiT ( P < .05 for both reviewers). Conclusion We demonstrated the successful implementation of whole‐heart spiral perfusion imaging with high resolution at 3T. Good image quality was achieved, and the SS without OVS showed the best image quality. Evaluation in patients with expected ischemic heart disease is warranted.

In the published version of the article by Zhou et al we found the following errors: In section 4.3, ‘Figure 5A’ on the third row should be ‘Figure 6A’. We regret these errors in our manuscript.

With different states of two intrinsic fluorophores, nicotinamide adenine dinucleotide (phosphate) (NAD(P)H) and flavin adenine dinucleotide (FAD), we developed a single-layer autoencoder (AE) for feature extraction, which outputs condensed features representing the full metabolic FLIM information with lower dimensionality. We also described distributions of AE features and fluorescence lifetime redox ratio (FLIRR) from single cells by Gaussian mixture models (GMM), and predicted the values of FLIRR based on feature data from each time point for the HeLa cell lines and Caucasian-American (LNCaP) prostate cancer cell lines by the polynomial regression model and the random forest regression model.

To develop a free‐breathing cardiac self‐gated technique that provides cine images and B1+ slice profile–corrected T1 maps from a single acquisition.

The objective of the current study was to develop and evaluate a DEep learning‐based rapid Spiral Image REconstruction (DESIRE) technique for high‐resolution spiral first‐pass myocardial perfusion imaging with whole‐heart coverage, to provide fast and accurate image reconstruction for both single‐slice (SS) and simultaneous multislice (SMS) acquisitions. Three‐dimensional U‐Net–based image enhancement architectures were evaluated for high‐resolution spiral perfusion imaging at 3 T. The SS and SMS MB = 2 networks were trained on SS perfusion images from 156 slices from 20 subjects. Structural similarity index (SSIM), peak signal‐to‐noise ratio (PSNR), and normalized root mean square error (NRMSE) were assessed, and prospective images were blindly graded by two experienced cardiologists (5: excellent; 1: poor). Excellent performance was demonstrated for the proposed technique. For SS, SSIM, PSNR, and NRMSE were 0.977 [0.972, 0.982], 42.113 [40.174, 43.493] dB, and 0.102 [0.080, 0.125], respectively, for the best network. For SMS MB = 2 retrospective data, SSIM, PSNR, and NRMSE were 0.961 [0.950, 0.969], 40.834 [39.619, 42.004] dB, and 0.107 [0.086, 0.133], respectively, for the best network. The image quality scores were 4.5 [4.1, 4.8], 4.5 [4.3, 4.6], 3.5 [3.3, 4], and 3.5 [3.3, 3.8] for SS DESIRE, SS L1‐SPIRiT, MB = 2 DESIRE, and MB = 2 SMS‐slice‐L1‐SPIRiT, respectively, showing no statistically significant difference (p = 1 and p = 1 for SS and SMS, respectively) between L1‐SPIRiT and the proposed DESIRE technique. The network inference time was ~100 ms per dynamic perfusion series with DESIRE, while the reconstruction time of L1‐SPIRiT with GPU acceleration was ~ 30 min. It was concluded that DESIRE enabled fast and high‐quality image reconstruction for both SS and SMS MB = 2 whole‐heart high‐resolution spiral perfusion imaging.

Microglia play significant roles in normal and pathologic brains. Microscopy images or videos of the mouse brain are the primary source for studying these cells in vivo. Due to low image contrast, enhancement is necessary before higher-level analysis. Existing techniques are not suitable for enhancing the complex structure of microglia. This paper proposes a novel technique (VBET) to enhance the blob and vessel-like structures of these cells. Experiments on 3D and 2D microglial images show that VBET preserves more structures than existing enhancement methods, and subsequent analyses, such as segmentation and microglial ramification, have higher accuracy than before (25% improvement in 3D Dice index, and 50% improvement in ramification index).

Noise and aliasing artifacts arise in various accelerated cardiac magnetic resonance (CMR) imaging applications. In accelerated myocardial T1-mapping, the traditional three-parameter based nonlinear regression may not provide accurate estimates due to sensitivity to noise. A deep neural network-based framework is proposed to address this issue. The DeepT1 framework consists of recurrent and U-net convolution networks to produce a single output map from the noisy and incomplete measurements. The results show that DeepT1 provides noise-robust estimates compared to the traditional pixel-wise three parameter fitting.

This paper proposes an automatic parameter selection framework for optimizing the performance of parameter-dependent regularized reconstruction algorithms. The proposed approach exploits a convolutional neural network for direct estimation of the regularization parameters from the acquired imaging data. This method can provide very reliable parameter estimates in a computationally efficient way. The effectiveness of the proposed approach is verified on transform-learning-based magnetic resonance image reconstructions of two different publicly available datasets. This experiment qualitatively and quantitatively measures improvement in image reconstruction quality using the proposed parameter selection strategy versus both existing parameter selection solutions and a fully deep-learning reconstruction with limited training data. Based on the experimental results, the proposed method improves average reconstructed image peak signal-to-noise ratio by a dB or more versus all competing methods in both brain and knee datasets, over a range of subsampling factors and input noise levels.

Automatic 3D reconstruction of glia morphology is a powerful tool necessary for investigating the role of microglia in neurological disorders in the central nervous system. Current glia skeleton reconstruction techniques fail to capture an accurate tracing of the processes over time, useful for the study of the microglia motility and morphology in the brain during healthy and diseased states. We propose Hieroglyph, a fully automatic temporal 3D skeleton reconstruction algorithm for glia imaged via 3D multiphoton microscopy. Hieroglyph yielded a 21% performance increase compared to state of the art automatic skeleton reconstruction methods and outperforms the state of the art in different measures of consistency on datasets of 3D images of microglia. The results from this method provide a 3D graph and digital reconstruction of glia useful for a myriad of morphological analyses that could impact studies in brain immunology and disease.

Cell imaging and analysis are fundamental to biomedical research because cells are the basic functional units of life. Among different cell-related analysis, cell counting and detection are widely used. In this paper, we focus on one common step of learning-based cell counting approaches: coding the raw dot labels into more suitable maps for learning. Two criteria of coding raw dot labels are discussed, and a new coding scheme is proposed in this paper. The two criteria measure how easy it is to train the model with a coding scheme, and how robust the recovered raw dot labels are when predicting. The most compelling advantage of the proposed coding scheme is the ability to distinguish neighboring cells in crowded regions. Cell counting and detection experiments are conducted for five coding schemes on four types of cells and two network architectures. The proposed coding scheme improves the counting accuracy versus the widely-used Gaussian and rectangle kernels up to 12%, and also improves the detection accuracy versus the common proximity coding up to 14%.

In this paper, we describe a novel enhancement method for images containing filamentous structures. Our method combines a gradient sparsity constraint with a filamentous structure constraint for the effective removal of clutter and noise from the background. The method is applied and evaluated on three types of data: 1) confocal microscopy images of neurons; 2) calcium imaging data; and 3) images of road pavement. We found that the images enhanced by our method preserve both the structure and the intensity details of the original object. In the case of neuron microscopy, we find that the neurons enhanced by our method are better correlated with the original structure intensities than the neurons enhanced by well-known vessel enhancement methods. Experiments on simulated calcium imaging data indicate that both the number of detected neurons and the accuracy of the derived calcium activity are improved. Applying our method to real calcium data, more regions exhibiting calcium activity in the full field of view were found. In road pavement crack detection, smaller or milder cracks were detected after using our enhancement method.

The ADVANCE program strives to change institutional culture to promote gender equity among faculty in science, engineering, and social and behavioral sciences. This is a challenging goal, as most ADVANCE teams have experienced. We found that one powerful way to approach changing individuals’ perspectives and institutions’ cultures is through art. Our ADVANCE team at the University of Virginia built three exhibits—two physical exhibits on campus and one online exhibit—to portray and celebrate our women faculty through photographic portraits and oral history narratives of their experiences. Our aim was to make women visible in spaces in which they were historically barred or overlooked, such as in particular disciplines as well as at UVA itself, which only admitted women students in the 1970s. This paper documents how and why we used art to challenge assumptions about women scholars and how we evaluated the exhibits’ impact on the UVA community.

PurposeTo develop a continuous‐acquisition cardiac self‐gated spiral pulse sequence and a respiratory motion‐compensated reconstruction strategy for free‐breathing cine imaging.MethodsCine data were acquired continuously on a 3T scanner for 8 seconds per slice without ECG gating or breath‐holding, using a golden‐angle gradient echo spiral pulse sequence. Cardiac motion information was extracted by applying principal component analysis on the gridded 8 × 8 k‐space center data. Respiratory motion was corrected by rigid registration on each heartbeat. Images were reconstructed using a low‐rank and sparse (L+S) technique. This strategy was evaluated in 37 healthy subjects and 8 subjects undergoing clinical cardiac MR studies. Image quality was scored (1–5 scale) in a blinded fashion by 2 experienced cardiologists. In 13 subjects with whole‐heart coverage, left ventricular ejection fraction (LVEF) from SPiral Acquisition with Respiratory correction and Cardiac Self‐gating (SPARCS) was compared to that from a standard ECG‐gated breath‐hold balanced steady‐state free precession (bSSFP) cine sequence.ResultsThe self‐gated signal was successfully extracted in all cases and demonstrated close agreement with the acquired ECG signal (mean bias, –0.22 ms). The mean image score across all subjects was 4.0 for reconstruction using the L+S model. There was good agreement between the LVEF derived from SPARCS and the gold‐standard bSSFP technique.ConclusionSPARCS successfully images cardiac function without the need for ECG gating or breath‐holding. With an 8‐second data acquisition per slice, whole‐heart cine images with clinically acceptable spatial and temporal resolution and image quality can be acquired in <90 seconds of free‐breathing acquisition.

PurposeSpiral MRI has advantages for cardiac imaging, and multiband (MB) spiral MRI of the heart shows promise. However, current reconstruction methods for MB spiral imaging have limitations. We sought to develop improved reconstruction methods for MB spiral cardiac MRI.MethodsTwo reconstruction methods were developed. The first is non‐Cartesian slice‐GRAPPA (NCSG), which uses phase demodulation and gridding operations before application of a Cartesian slice‐separating kernel. The second method, slice‐SPIRiT, formulates the reconstruction as a minimization problem that enforces in‐plane coil consistency and consistency with the acquired MB data, and uses through‐plane coil sensitivity information in the iterative solution. These methods were compared with conjugate‐gradient SENSE in phantoms and volunteers. Temporal alternation of CAIPIRINHA (controlled aliasing in parallel imaging results in higher acceleration) phase and the use of a temporal filter were also investigated.ResultsPhantom experiments with 3 simultaneous slices (MB = 3) showed that mean artifact power was highest for conjugate‐gradient SENSE, lower for NCSG, and lowest for slice‐SPIRiT. For volunteer cine imaging (MB = 3, N = 5), the artifact power was 0.182 ± 0.037, 0.148 ± 0.036, and 0.139 ± 0.034 for conjugate‐gradient SENSE, NCSG, and slice‐SPIRiT, respectively (P < .05, analysis of variance). Temporal alternation of CAIPIRINHA reduced artifacts for both NCSG and slice‐SPIRiT.ConclusionThe NCSG and slice‐SPIRiT methods provide more accurate reconstructions for MB spiral cine imaging compared with conjugate‐gradient SENSE. These methods hold promise for non‐Cartesian MB imaging.

Estimating a time-varying signal, such as head motion from magnetic resonance imaging data, becomes particularly challenging in the face of other temporal dynamics such as functional activation. This paper describes a new Kalman filter-like framework that includes a sparse residual term in the measurement model. This additional term allows the extended Kalman filter to generate real-time motion estimates suitable for prospective motion correction when such dynamics occur. An iterative augmented Lagrangian algorithm similar to the alterating direction method of multipliers implements the update step for this Kalman filter. This paper evaluates the accuracy and convergence rate of this iterative method for small and large motion in terms of its sensitivity to parameter selection. The included experiment on a simulated functional magnetic resonance imaging acquisition demonstrates that the resulting method improves the maximum Youden's J index of the time series analysis by 2-3% versus retrospective motion correction, while the sensitivity index increases from 4.3 to 5.4 when combining prospective and retrospective correction.