Search

IntroductionCoronavirus disease 2019 (COVID-19) has caused >3.5 million deaths worldwide and affected >160 million people. At least twice as many have been infected but remained asymptomatic or minimally symptomatic. COVID-19 includes central nervous system manifestations mediated by inflammation and cerebrovascular, anoxic, and/or viral neurotoxicity mechanisms. More than one third of patients with COVID-19 develop neurologic problems during the acute phase of the illness, including loss of sense of smell or taste, seizures, and stroke. Damage or functional changes to the brain may result in chronic sequelae. The risk of incident cognitive and neuropsychiatric complications appears independent from the severity of the original pulmonary illness. It behooves the scientific and medical community to attempt to understand the molecular and/or systemic factors linking COVID-19 to neurologic illness, both short and long term.MethodsThis article describes what is known so far in terms of links among COVID-19, the brain, neurological symptoms, and Alzheimer's disease (AD) and related dementias. We focus on risk factors and possible molecular, inflammatory, and viral mechanisms underlying neurological injury. We also provide a comprehensive description of the Alzheimer's Association Consortium on Chronic Neuropsychiatric Sequelae of SARS-CoV-2 infection (CNS SC2) harmonized methodology to address these questions using a worldwide network of researchers and institutions.ResultsSuccessful harmonization of designs and methods was achieved through a consensus process initially fragmented by specific interest groups (epidemiology, clinical assessments, cognitive evaluation, biomarkers, and neuroimaging). Conclusions from subcommittees were presented to the whole group and discussed extensively. Presently data collection is ongoing at 19 sites in 12 countries representing Asia, Africa, the Americas, and Europe.DiscussionThe Alzheimer's Association Global Consortium harmonized methodology is proposed as a model to study long-term neurocognitive sequelae of SARS-CoV-2 infection.Key pointsThe following review describes what is known so far in terms of molecular and epidemiological links among COVID-19, the brain, neurological symptoms, and AD and related dementias (ADRD)The primary objective of this large-scale collaboration is to clarify the pathogenesis of ADRD and to advance our understanding of the impact of a neurotropic virus on the long-term risk of cognitive decline and other CNS sequelae. No available evidence supports the notion that cognitive impairment after SARS-CoV-2 infection is a form of dementia (ADRD or otherwise). The longitudinal methodologies espoused by the consortium are intended to provide data to answer this question as clearly as possible controlling for possible confounders. Our specific hypothesis is that SARS-CoV-2 triggers ADRD-like pathology following the extended olfactory cortical network (EOCN) in older individuals with specific genetic susceptibility.The proposed harmonization strategies and flexible study designs offer the possibility to include large samples of under-represented racial and ethnic groups, creating a rich set of harmonized cohorts for future studies of the pathophysiology, determinants, long-term consequences, and trends in cognitive aging, ADRD, and vascular disease.We provide a framework for current and future studies to be carried out within the Consortium. and offers a "green paper" to the research community with a very broad, global base of support, on tools suitable for low- and middle-income countries aimed to compare and combine future longitudinal data on the topic.The Consortium proposes a combination of design and statistical methods as a means of approaching causal inference of the COVID-19 neuropsychiatric sequelae. We expect that deep phenotyping of neuropsychiatric sequelae may provide a series of candidate syndromes with phenomenological and biological characterization that can be further explored. By generating high-quality harmonized data across sites we aim to capture both descriptive and, where possible, causal associations.

PurposeTo improve hyperpolarized 13 C (HP-13 C) MRI by image denoising with a new approach, patch-based higher-order singular value decomposition (HOSVD).MethodsThe benefit of using a patch-based HOSVD method to denoise dynamic HP-13 C MR imaging data was investigated. Image quality and the accuracy of quantitative analyses following denoising were evaluated first using simulated data of [1-13 C]pyruvate and its metabolic product, [1-13 C]lactate, and compared the results to a global HOSVD method. The patch-based HOSVD method was then applied to healthy volunteer HP [1-13 C]pyruvate EPI studies. Voxel-wise kinetic modeling was performed on both non-denoised and denoised data to compare the number of voxels quantifiable based on SNR criteria and fitting error.ResultsSimulation results demonstrated an 8-fold increase in the calculated SNR of [1-13 C]pyruvate and [1-13 C]lactate with the patch-based HOSVD denoising. The voxel-wise quantification of kPL (pyruvate-to-lactate conversion rate) showed a 9-fold decrease in standard errors for the fitted kPL after denoising. The patch-based denoising performed superior to the global denoising in recovering kPL information. In volunteer data sets, [1-13 C]lactate and [13 C]bicarbonate signals became distinguishable from noise across captured time points with over a 5-fold apparent SNR gain. This resulted in >3-fold increase in the number of voxels quantifiable for mapping kPB (pyruvate-to-bicarbonate conversion rate) and whole brain coverage for mapping kPL .ConclusionsSensitivity enhancement provided by this denoising significantly improved quantification of metabolite dynamics and could benefit future studies by improving image quality, enabling higher spatial resolution, and facilitating the extraction of metabolic information for clinical research.

PurposeWith the initiation of human hyperpolarized 13 C (HP-13 C) trials at multiple sites and the development of improved acquisition methods, there is an imminent need to maximally extract diagnostic information to facilitate clinical interpretation. This study aims to improve human HP-13 C MR spectroscopic imaging through means of Tensor Rank truncation-Image enhancement (TRI) and optimal receiver combination (ORC).MethodsA data-driven processing framework for dynamic HP 13 C MR spectroscopic imaging (MRSI) was developed. Using patient data sets acquired with both multichannel arrays and single-element receivers from the brain, abdomen, and pelvis, we examined the theory and application of TRI, as well as 2 ORC techniques: whitened singular value decomposition (WSVD) and first-point phasing. Optimal conditions for TRI were derived based on bias-variance trade-off.ResultsTRI and ORC techniques together provided a 63-fold mean apparent signal-to-noise ratio (aSNR) gain for receiver arrays and a 31-fold gain for single-element configurations, which particularly improved quantification of the lower-SNR-[13 C]bicarbonate and [1-13 C]alanine signals that were otherwise not detectable in many cases. Substantial SNR enhancements were observed for data sets that were acquired even with suboptimal experimental conditions, including delayed (114 s) injection (8× aSNR gain solely by TRI), or from challenging anatomy or geometry, as in the case of a pediatric patient with brainstem tumor (597× using combined TRI and WSVD). Improved correlation between elevated pyruvate-to-lactate conversion, biopsy-confirmed cancer, and mp-MRI lesions demonstrated that TRI recovered quantitative diagnostic information.ConclusionOverall, this combined approach was effective across imaging targets and receiver configurations and could greatly benefit ongoing and future HP 13 C MRI research through major aSNR improvements.

Background: Training convolutional neural networks using pathology whole slide images (WSIs) is traditionally prefaced by the extraction of a training dataset of image patches. While effective, for large datasets of WSIs, this dataset preparation is inefficient. Methods: We created a custom pipeline (histo-fetch) to efficiently extract random patches and labels from pathology WSIs for input to a neural network on-the-fly. We prefetch these patches as needed during network training, avoiding the need for WSI preparation such as chopping/tiling. Results & conclusions: We demonstrate the utility of this pipeline to perform artificial stain transfer and image generation using the popular networks CycleGAN and ProGAN, respectively. For a large WSI dataset, histo-fetch is 98.6% faster to start training and used 7535x less disk space.