Nonlinear Gradient Field Estimation in Diffusion MRI Tensor Simulation.

Gradient nonlinearities not only induce spatial distortion in magnetic resonance imaging (MRI), but also introduce discrepancies between intended and acquired diffusion sensitization in diffusion weighted (DW) MRI. Advances in scanner performance have increased the importance of correcting gradient nonlinearities. The most common approaches for gradient nonlinear field estimations rely on phantom calibration field maps which are not always feasible, especially on retrospective data. Here, we derive a quadratic minimization problem for the complete gradient nonlinear field (L(r)). This approach starts with corrupt diffusion signal and estimates the L(r) in two scenarios: (1) the true diffusion tensor known and (2) the true diffusion tensor unknown (i.e., diffusion tensor is estimated). We show the validity of this mathematical approach, both theoretically and through tensor simulation. The estimated field is assessed through diffusion tensor metrics: mean diffusivity (MD), fractional anisotropy (FA), and principal eigenvector (V1). In simulation with 300 diffusion tensors, the study shows the mathematical model is not ill-posed and remains stable. We find when the true diffusion tensor is known (1) the change in determinant of the estimated L(r) field and the true field is near zero and (2) the median difference in estimated L(r) corrected diffusion metrics to true values is near zero. We find the results of L(r) estimation are dependent on the level of L(r) corruption. This work provides an approach to estimate gradient field without the need for additional calibration scans. To the best of our knowledge, the mathematical derivation presented here is novel.

Functional correlation tensors in brain white matter and the effects of normal aging.

Resting state correlations between blood oxygenation level dependent (BOLD) MRI signals from voxels in white matter (WM) are demonstrably anisotropic, so that functional correlation tensors (FCT) may be used to quantify the underlying microstructure of BOLD effects in WM tracts. However, the overall spatial distribution of FCTs and their metrics in specific populations has not yet been established, and the factors that affect their precise arrangements remain unclear. Changes in WM occur with normal aging, and these may be expected to affect FCTs. We hypothesized that FCTs exhibit a characteristic spatial pattern and may show systematic changes with aging or other factors. Here we report our analyses of the FCT characteristics of fMRI images of a large cohort of 461 cognitively normal subjects (190 females, 271 males) sourced from the Open Access Series of Imaging Studies (OASIS), with age distributions of 42 y/o - 95 y/o. Group averages and statistics of FCT indices, including axial functional correlations, radial functional correlations, mean functional correlations and fractional anisotropy, were quantified in WM bundles defined by the JHU ICBM-DTI-81 WM atlas. In addition, their variations with normal aging were examined. The results revealed a dimorphic distribution of changes in FCT metrics with age, with decreases of the functional correlations in some regions and increases in others. Supplementary analysis revealed that females exhibited significant age effects on a greater number of WM areas, but the interaction between age and sex was not significant. The findings demonstrate the reproducibility of the spatial distribution of FCT metrics and reveal subtle regional changes with age.

Brain structure and activity predicting cognitive maturation in adolescence.

Cognitive abilities of primates, including humans, continue to improve through adolescence 1,2. While a range of changes in brain structure and connectivity have been documented 3,4, how they affect neuronal activity that ultimately determines performance of cognitive functions remains unknown. Here, we conducted a multilevel longitudinal study of monkey adolescent neurocognitive development. The developmental trajectory of neural activity in the prefrontal cortex accounted remarkably well for working memory improvements. While complex aspects of activity changed progressively during adolescence, such as the rotation of stimulus representation in multidimensional neuronal space, which has been implicated in cognitive flexibility, even simpler attributes, such as the baseline firing rate in the period preceding a stimulus appearance had predictive power over behavior. Unexpectedly, decreases in brain volume and thickness, which are widely thought to underlie cognitive changes in humans 5 did not predict well the trajectory of neural activity or cognitive performance changes. Whole brain cortical volume in particular, exhibited an increase and reached a local maximum in late adolescence, at a time of rapid behavioral improvement. Maturation of long-distance white matter tracts linking the frontal lobe with areas of the association cortex and subcortical regions best predicted changes in neuronal activity and behavior. Our results provide evidence that optimization of neural activity depending on widely distributed circuitry effects cognitive development in adolescence.

Spatiospectral image processing workflow considerations for advanced MR spectroscopy of the brain.

Magnetic resonance spectroscopy (MRS) is one of the few non-invasive imaging modalities capable of making neurochemical and metabolic measurements in vivo. Traditionally, the clinical utility of MRS has been narrow. The most common use has been the "single-voxel spectroscopy" variant to discern the presence of a lactate peak in the spectra in one location in the brain, typically to evaluate for ischemia in neonates. Thus, the reduction of rich spectral data to a binary variable has not classically necessitated much signal processing. However, scanners have become more powerful and MRS sequences more advanced, increasing data complexity and adding 2 to 3 spatial dimensions in addition to the spectral one. The result is a spatially- and spectrally-variant MRS image ripe for image processing innovation. Despite this potential, the logistics for robustly accessing and manipulating MRS data across different scanners, data formats, and software standards remain unclear. Thus, as research into MRS advances, there is a clear need to better characterize its image processing considerations to facilitate innovation from scientists and engineers. Building on established neuroimaging standards, we describe a framework for manipulating these images that generalizes to the voxel, spectral, and metabolite level across space and multiple imaging sites while integrating with LCModel, a widely used quantitative MRS peak-fitting platform. In doing so, we provide examples to demonstrate the advantages of such a workflow in relation to recent publications and with new data. Overall, we hope our characterizations will lower the barrier of entry to MRS processing for neuroimaging researchers.

DeepN4: Learning N4ITK Bias Field Correction for T1-weighted Images.

T1-weighted (T1w) MRI has low frequency intensity artifacts due to magnetic field inhomogeneities. Removal of these biases in T1w MRI images is a critical preprocessing step to ensure spatially consistent image interpretation. N4ITK bias field correction, the current state-of-the-art, is implemented in such a way that makes it difficult to port between different pipelines and workflows, thus making it hard to reimplement and reproduce results across local, cloud, and edge platforms. Moreover, N4ITK is opaque to optimization before and after its application, meaning that methodological development must work around the inhomogeneity correction step. Given the importance of bias fields correction in structural preprocessing and flexible implementation, we pursue a deep learning approximation / reinterpretation of the N4ITK bias fields correction to create a method which is portable, flexible, and fully differentiable. In this paper, we trained a deep learning network "DeepN4" on eight independent cohorts from 72 different scanners and age ranges with N4ITK-corrected T1w MRI and bias field for supervision in log space. We found that we can closely approximate N4ITK bias fields correction with naïve networks. We evaluate the peak signal to noise ratio (PSNR) in test dataset against the N4ITK corrected images. The median PSNR of corrected images between N4ITK and DeepN4 was 47.96 dB. In addition, we assess the DeepN4 model on eight additional external datasets and show the generalizability of the approach. This study establishes that incompatible N4ITK preprocessing steps can be closely approximated by naïve deep neural networks, facilitating more flexibility. All code and models are released at https://github.com/MASILab/DeepN4 .

Full-dose challenge of moderate, severe, and unknown beta-lactam allergies in the emergency department.

OBJECTIVE: This study aims to assess the outcome of challenging documented moderate, severe, or unknown beta-lactam allergies with full dose administration of a beta-lactam antibiotic in emergency department (ED) patients admitted for acute bacterial infection. METHODS: A single-center, retrospective, descriptive study of adult patients challenged with a full dose of beta-lactam in the ED from January 2021 to December 2022 was conducted. Included patients had at least one documented moderate, severe, or unknown beta-lactam allergy in the electronic medical record (EMR) without documentation of prior tolerance. Patient demographics, prior beta-lactam antibiotic reaction, beta-lactam administered in the ED, inpatient beta-lactam continuation, adverse drug reactions, and updates to allergy profiles were collected. Descriptive statistics for data analysis were performed using SPSS Version 22. RESULTS: Of the 184 ED encounters with full-dose beta-lactam challenges, five (2.7%) patients with documented moderate, severe, or unknown beta-lactam allergies experienced an allergic reaction after the challenge; one (0.5%) patient had an allergic reaction in the ED, and the remaining four (2.2%) occurred after admission. No anaphylactic reactions occurred. All allergic reactions were limited to mild rash or itching. Most patients (98.9%) were challenged with a cephalosporin. A beta-lactam was continued in 86.4% of cases, and the allergy profile was updated for future utilization in 73.4% of patients. CONCLUSIONS: This study suggests that full-dose challenge of moderate, severe, or unknown beta-lactam allergies can be safely accomplished in the ED. This approach avoids unnecessary penicillin allergy skin testing and reduces utilization of suboptimal alternative antibiotic regimens.

Quantification of mediation effects of white matter functional characteristics on cognitive decline in aging.

Cognitive decline with aging involves multifactorial processes, including changes in brain structure and function. This study focuses on the role of white matter functional characteristics, as reflected in blood oxygenation level-dependent signals, in age-related cognitive deterioration. Building on previous research confirming the reproducibility and age-dependence of blood oxygenation level-dependent signals acquired via functional magnetic resonance imaging, we here employ mediation analysis to test if aging affects cognition through white matter blood oxygenation level-dependent signal changes, impacting various cognitive domains and specific white matter regions. We used independent component analysis of resting-state blood oxygenation level-dependent signals to segment white matter into coherent hubs, offering a data-driven view of white matter's functional architecture. Through correlation analysis, we constructed a graph network and derived metrics to quantitatively assess regional functional properties based on resting-state blood oxygenation level-dependent fluctuations. Our analysis identified significant mediators in the age-cognition relationship, indicating that aging differentially influences cognitive functions by altering the functional characteristics of distinct white matter regions. These findings enhance our understanding of the neurobiological basis of cognitive aging, highlighting the critical role of white matter in maintaining cognitive integrity and proposing new approaches to assess interventions targeting cognitive decline in older populations.

Individualized spectral filters alleviate persistent photophobia, headaches and migraines in active duty military and Veterans following brain trauma.

PURPOSE: Consistent with association between photophobia and headache, growing evidence suggests an underlying causal relationship between light sensitivity and central pain. We investigated whether an intervention to regulate light sensitivity by filtering only wavelengths causing difficulties for the specific individual could alleviate headaches/migraines resulting from traumatic brain injury (TBI). METHODS: Secondary data analysis of a clinical database including N = 392 military personnel (97% men, 3% women), ranging in age from 20 to 51 years, diagnosed with TBI, persistent headaches/migraines, and light sensitivity. The average elapsed time from TBI diagnosis to intervention was 3 years. Headache/migraine severity, frequency, medication use, and difficulties related to daily functioning were assessed pre and 4-12 weeks post-intervention with individualized spectral filters. RESULTS: Monthly migraine frequency decreased significantly from an average of 14.8 to 1.9, with 74% reporting no migraines post-intervention. Prescription and over-the-counter medication use decreased by more than 70%. Individuals also reported significant improvement in light sensitivity, headaches/migraine severity, and physical and perceptual symptoms. CONCLUSIONS: Wearing individualized spectral filters was associated with symptom relief, increased subjective quality of reported health and well-being, and decreased objective medication use for TBI-related persistent headaches/migraines. These results support a suggested relationship between dysregulated light sensitivity and central regulation of pain.

Nonlinear changes in pupillary attentional orienting responses across the lifespan.

The cognitive aging process is not necessarily linear. Central task-evoked pupillary responses, representing a brainstem-pupil relationship, may vary across the lifespan. Thus we examined, in 75 adults ranging in age from 19 to 86, whether task-evoked pupillary responses to an attention task may serve in as an index of cognitive aging. This is because the locus coeruleus (LC), located in the brainstem, is not only among the earliest sites of degeneration in pathological aging, but also supports both attentional and pupillary behaviors. We assessed brief, task-evoked phasic attentional orienting to behaviorally relevant and irrelevant auditory tones, stimuli known specifically to recruit the LC in the brainstem and evoke pupillary responses. Due to potential nonlinear changes across the lifespan, we used a novel data-driven analysis on 6 dynamic pupillary behaviors on 10% of the data to reveal cut off points that best characterized the three age bands: young (19-41 years old), middle aged (42-68 years old), and older adults (69 + years old). Follow-up analyses on independent data, the remaining 90%, revealed age-related changes such as monotonic decreases in tonic pupillary diameter and dynamic range, along with curvilinear phasic pupillary responses to the behaviorally relevant target events, increasing in the middle-aged group and then decreasing in the older group. Additionally, the older group showed decreased differentiation of pupillary responses between target and distractor events. This pattern is consistent with potential compensatory LC activity in midlife that is diminished in old age, resulting in decreased adaptive gain. Beyond regulating responses to light, pupillary dynamics reveal a nonlinear capacity for neurally mediated gain across the lifespan, thus providing evidence in support of the LC adaptive gain hypothesis.

Comprehensive behavioral and physiologic assessment of peripheral and central auditory function in individuals with mild traumatic brain injury.

Auditory complaints are frequently reported by individuals with mild traumatic brain injury (mTBI) yet remain difficult to detect in the absence of clinically significant hearing loss. This highlights a growing need to identify sensitive indices of auditory-related mTBI pathophysiology beyond pure-tone thresholds for improved hearing healthcare diagnosis and treatment. Given the heterogeneity of mTBI etiology and the diverse peripheral and central processes required for normal auditory function, the present study sought to determine the audiologic assessments sensitive to mTBI pathophysiology at the group level using a well-rounded test battery of both peripheral and central auditory system function. This test battery included pure-tone detection thresholds, word understanding in quiet, sentence understanding in noise, distortion product otoacoustic emissions (DPOAEs), middle-ear muscle reflexes (MEMRs), and auditory evoked potentials (AEPs), including auditory brainstem responses (ABRs), middle latency responses (MLRs), and late latency responses (LLRs). Each participant also received magnetic resonance imaging (MRI). Compared to the control group, we found that individuals with mTBI had reduced DPOAE amplitudes that revealed a compound effect of age, elevated MEMR thresholds for an ipsilateral broadband noise elicitor, longer ABR Wave I latencies for click and 4 kHz tone burst elicitors, longer ABR Wave III latencies for 4 kHz tone bursts, larger MLR Na and Nb amplitudes, smaller MLR Pb amplitudes, longer MLR Pa latencies, and smaller LLR N1 amplitudes for older individuals with mTBI. Further, mTBI individuals with combined hearing difficulty and noise sensitivity had a greater number of deficits on thalamic and cortical AEP measures compared to those with only one/no self-reported auditory symptoms. This finding was corroborated with MRI, which revealed significant structural differences in the auditory cortical areas of mTBI participants who reported combined hearing difficulty and noise sensitivity, including an enlargement of left transverse temporal gyrus (TTG) and bilateral planum polare (PP). These findings highlight the need for continued investigations toward identifying individualized audiologic assessments and treatments that are sensitive to mTBI pathophysiology.

Optimization of the flip angles of narrow-band editing pulses in J-difference edited MRS of lactate at 3T.

PURPOSE: Application of highly selective editing RF pulses provides a means of minimizing co-editing of contaminants in J-difference MRS (MEGA), but it causes reduction in editing yield. We examined the flip angles (FAs) of narrow-band editing pulses to maximize the lactate edited signal with minimal co-editing of threonine. METHODS: The effect of editing-pulse FA on the editing performance was examined, with numerical and phantom analyses, for bandwidths of 17.6-300 Hz in MEGA-PRESS editing of lactate at 3T. The FA and envelope of 46 ms Gaussian editing pulses were tailored to maximize the lactate edited signal at 1.3 ppm and minimize co-editing of threonine. The optimized editing-pulse FA MEGA scheme was tested in brain tumor patients. RESULTS: Simulation and phantom data indicated that the optimum FA of MEGA editing pulses is progressively larger than 180° as the editing-pulse bandwidth decreases. For 46 ms long 17.6 Hz bandwidth Gaussian pulses and other given sequence parameters, the lactate edited signal was maximum at the first and second editing-pulse FAs of 241° and 249°, respectively. The edit-on and difference-edited lactate peak areas of the optimized FA MEGA were greater by 43% and 25% compared to the 180°-FA MEGA, respectively. In-vivo data confirmed the simulation and phantom results. The lesions of the brain tumor patients showed elevated lactate and physiological levels of threonine. CONCLUSION: The lactate MEGA editing yield is significantly increased with editing-pulse FA much larger than 180° when the editing-pulse bandwidth is comparable to the lactate quartet frequency width.

Effect of Meditation on Brain Activity during an Attention Task: A Comparison Study of ASL and BOLD Task fMRI.

Focused attention meditation (FAM) training has been shown to improve attention, but the neural basis of FAM on attention has not been thoroughly understood. Here, we aim to investigate the neural effect of a 2-month FAM training on novice meditators in a visual oddball task (a frequently adopted task to evaluate attention), evaluated with both ASL and BOLD fMRI. Using ASL, activation was increased in the middle cingulate (part of the salience network, SN) and temporoparietal (part of the frontoparietal network, FPN) regions; the FAM practice time was negatively associated with the longitudinal changes in activation in the medial prefrontal (part of the default mode network, DMN) and middle frontal (part of the FPN) regions. Using BOLD, the FAM practice time was positively associated with the longitudinal changes of activation in the inferior parietal (part of the dorsal attention network, DAN), dorsolateral prefrontal (part of the FPN), and precentral (part of the DAN) regions. The effect sizes for the activation changes and their association with practice time using ASL are significantly larger than those using BOLD. Our study suggests that FAM training may improve attention via modulation of the DMN, DAN, SN, and FPN, and ASL may be a sensitive tool to study the FAM effect on attention.

DeepN4: Learning N4ITK Bias Field Correction for T1-weighted Images.

T1-weighted (T1w) MRI has low frequency intensity artifacts due to magnetic field inhomogeneities. Removal of these biases in T1w MRI images is a critical preprocessing step to ensure spatially consistent image interpretation. N4ITK bias field correction, the current state-of-the-art, is implemented in such a way that makes it difficult to port between different pipelines and workflows, thus making it hard to reimplement and reproduce results across local, cloud, and edge platforms. Moreover, N4ITK is opaque to optimization before and after its application, meaning that methodological development must work around the inhomogeneity correction step. Given the importance of bias fields correction in structural preprocessing and flexible implementation, we pursue a deep learning approximation / reinterpretation of the N4ITK bias fields correction to create a method which is portable, flexible, and fully differentiable. In this paper, we trained a deep learning network "DeepN4" on eight independent cohorts from 72 different scanners and age ranges with N4ITK-corrected T1w MRI and bias field for supervision in log space. We found that we can closely approximate N4ITK bias fields correction with naïve networks. We evaluate the peak signal to noise ratio (PSNR) in test dataset against the N4ITK corrected images. The median PSNR of corrected images between N4ITK and DeepN4 was 47.96 dB. In addition, we assess the DeepN4 model on eight additional external datasets and show the generalizability of the approach. This study establishes that incompatible N4ITK preprocessing steps can be closely approximated by naïve deep neural networks, facilitating more flexibility. All code and models are released at https://github.com/MASILab/DeepN4.

Attention-dependent coupling with forebrain and brainstem neuromodulatory nuclei changes across the lifespan.

Attentional states continuously reflect the predictability and uncertainty in one's environment having important consequences for learning and memory. Beyond well known cortical contributions, rapid shifts in attention are hypothesized to also originate from deep nuclei, such as the basal forebrain (BF) and locus coeruleus (LC) neuromodulatory systems. These systems are also the first to change with aging. Here we characterized the interplay between these systems and their regulation of afferent targets - the hippocampus (HPC) and posterior cingulate cortex (PCC) - across the lifespan. To examine the role of attentional salience on task-dependent functional connectivity, we used a target-distractor go/no go task presented during functional MRI. In younger adults, BF coupling with the HPC, and LC coupling with the PCC, increased with behavioral relevance (targets vs distractors). Although the strength and presence of significant regional coupling changed in middle age, the most striking change in network connectivity was in old age, such that in older adults BF and LC coupling with their cortical afferents was largely absent and replaced by stronger interconnectivity between LC-BF nuclei. Overall rapid changes in attention related to behavioral relevance revealed distinct roles of subcortical neuromodulatory systems. The pronounced changes in functional network architecture across the lifespan suggest a decrease in these distinct roles, with deafferentation of cholinergic and noradrenergic systems associated with a shift towards mutual support during attention guided to external stimuli.

Locus coeruleus neuromelanin accumulation and dissipation across the lifespan.

The pigment neuromelanin, produced in the locus coeruleus (LC) as a byproduct of catecholamine synthesis, gives the "blue spot" its name, and both identifies LC neurons and is thought to play an important yet complex role in normal and pathological aging. Using neuromelanin-sensitive T1-weighted turbo spin echo MRI scans we characterized volume and neuromelanin signal intensity in the LC of 96 participants between the ages of 19 and 86. Although LC volume did not change significantly throughout the lifespan, LC neuromelanin signal intensity increased from early adulthood, peaked around age 60 and precipitously declined thereafter. Neuromelanin intensity was greater in the caudal relative to rostral extent and in women relative to men. With regard to function, rostral LC neuromelanin intensity was associated with fluid cognition in older adults (60+) only in those above the 50th percentile of cognitive ability for age. The gradual accumulation of LC neuromelanin across the lifespan, its sudden dissipation in later life, and relation to preserved cognitive function, is consistent with its complex role in normal and pathological aging.

Whole-brain, gray, and white matter time-locked functional signal changes with simple tasks and model-free analysis.

Recent studies have revealed the production of time-locked blood oxygenation level-dependent (BOLD) functional MRI (fMRI) signals throughout the entire brain in response to tasks, challenging the existence of sparse and localized brain functions and highlighting the pervasiveness of potential false negative fMRI findings. "Whole-brain" actually refers to gray matter, the only tissue traditionally studied with fMRI. However, several reports have demonstrated reliable detection of BOLD signals in white matter, which have previously been largely ignored. Using simple tasks and analyses, we demonstrate BOLD signal changes across the whole brain, in both white and gray matters, in similar manner to previous reports of whole brain studies. We investigated whether white matter displays time-locked BOLD signals across multiple structural pathways in response to a stimulus in a similar manner to the cortex. We find that both white and gray matter show time-locked activations across the whole brain, with a majority of both tissue types showing statistically significant signal changes for all task stimuli investigated. We observed a wide range of signal responses to tasks, with different regions showing different BOLD signal changes to the same task. Moreover, we find that each region may display different BOLD responses to different stimuli. Overall, we present compelling evidence that, just like all gray matter, essentially all white matter in the brain shows time-locked BOLD signal changes in response to multiple stimuli, challenging the idea of sparse functional localization and the prevailing wisdom of treating white matter BOLD signals as artifacts to be removed.

Seasonal variations of functional connectivity of human brains.

Seasonal variations have long been observed in various aspects of human life. While there is an abundance of research that has characterized seasonality effects in, for example, cognition, mood, and behavior, including studies of underlying biophysical mechanisms, direct measurements of seasonal variations of brain functional activities have not gained wide attention. We have quantified seasonal effects on functional connectivity as derived from MRI scans. A cohort of healthy human subjects was divided into four groups based on the seasons of their scanning dates as documented in the image database of the Human Connectome Project. Sinusoidal functions were used as regressors to determine whether there were significant seasonal variations in measures of brain activities. We began with the analysis of seasonal variations of the fractional amplitudes of low frequency fluctuations of regional functional signals, followed by the seasonal variations of functional connectivity in both global- and network-level. Furthermore, relevant environmental factors, including average temperature and daylength, were found to be significantly associated with brain functional activities, which may explain how the observed seasonal fluctuations arise. Finally, topological properties of the brain functional network also showed significant variations across seasons. All the observations accumulated revealed seasonality effects of human brain activities in a resting-state, which may have important practical implications for neuroimaging research.

Spatiospectral image processing workflow considerations for advanced MR spectroscopy of the brain.

Magnetic resonance spectroscopy (MRS) is one of the few non-invasive imaging modalities capable of making neurochemical and metabolic measurements in vivo. Traditionally, the clinical utility of MRS has been narrow. The most common use has been the "single-voxel spectroscopy" variant to discern the presence of a lactate peak in the spectra in one location in the brain, typically to evaluate for ischemia in neonates. Thus, the reduction of rich spectral data to a binary variable has not classically necessitated much signal processing. However, scanners have become more powerful and MRS sequences more advanced, increasing data complexity and adding 2 to 3 spatial dimensions in addition to the spectral one. The result is a spatially- and spectrally-variant MRS image ripe for image processing innovation. Despite this potential, the logistics for robustly accessing and manipulating MRS data across different scanners, data formats, and software standards remain unclear. Thus, as research into MRS advances, there is a clear need to better characterize its image processing considerations to facilitate innovation from scientists and engineers. Building on established neuroimaging standards, we describe a framework for manipulating these images that generalizes to the voxel, spectral, and metabolite level across space and multiple imaging sites while integrating with LCModel, a widely used quantitative MRS peak-fitting platform. In doing so, we provide examples to demonstrate the advantages of such a workflow in relation to recent publications and with new data. Overall, we hope our characterizations will lower the barrier of entry to MRS processing for neuroimaging researchers.

Affine image registration of arterial spin labeling MRI using deep learning networks.

Convolutional neural networks (CNN) have demonstrated good accuracy and speed in spatially registering high signal-to-noise ratio (SNR) structural magnetic resonance imaging (sMRI) images. However, some functional magnetic resonance imaging (fMRI) images, e.g., those acquired from arterial spin labeling (ASL) perfusion fMRI, are of intrinsically low SNR and therefore the quality of registering ASL images using CNN is not clear. In this work, we aimed to explore the feasibility of a CNN-based affine registration network (ARN) for registration of low-SNR three-dimensional ASL perfusion image time series and compare its performance with that from the state-of-the-art statistical parametric mapping (SPM) algorithm. The six affine parameters were learned from the ARN using both simulated motion and real acquisitions from ASL perfusion fMRI data and the registered images were generated by applying the transformation derived from the affine parameters. The speed and registration accuracy were compared between ARN and SPM. Several independent datasets, including meditation study (10 subjects × 2), bipolar disorder study (26 controls, 19 bipolar disorder subjects), and aging study (27 young subjects, 33 older subjects), were used to validate the generality of the trained ARN model. The ARN method achieves superior image affine registration accuracy (total translation/total rotation errors of ARN vs. SPM: 1.17 mm/1.23° vs. 6.09 mm/12.90° for simulated images and reduced MSE/L1/DSSIM/Total errors of 18.07% / 19.02% / 0.04% / 29.59% for real ASL test images) and 4.4 times (ARN vs. SPM: 0.50 s vs. 2.21 s) faster speed compared to SPM. The trained ARN can be generalized to align ASL perfusion image time series acquired with different scanners, and from different image resolutions, and from healthy or diseased populations. The results demonstrated that our ARN markedly outperforms the iteration-based SPM both for simulated motion and real acquisitions in terms of registration accuracy, speed, and generalization.