Towards computer-aided severity assessment via deep neural networks for geographic and opacity extent scoring of SARS-CoV-2 chest X-rays.

A critical step in effective care and treatment planning for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the cause for the coronavirus disease 2019 (COVID-19) pandemic, is the assessment of the severity of disease progression. Chest x-rays (CXRs) are often used to assess SARS-CoV-2 severity, with two important assessment metrics being extent of lung involvement and degree of opacity. In this proof-of-concept study, we assess the feasibility of computer-aided scoring of CXRs of SARS-CoV-2 lung disease severity using a deep learning system. Data consisted of 396 CXRs from SARS-CoV-2 positive patient cases. Geographic extent and opacity extent were scored by two board-certified expert chest radiologists (with 20+ years of experience) and a 2nd-year radiology resident. The deep neural networks used in this study, which we name COVID-Net S, are based on a COVID-Net network architecture. 100 versions of the network were independently learned (50 to perform geographic extent scoring and 50 to perform opacity extent scoring) using random subsets of CXRs from the study, and we evaluated the networks using stratified Monte Carlo cross-validation experiments. The COVID-Net S deep neural networks yielded R[Formula: see text] of [Formula: see text] and [Formula: see text] between predicted scores and radiologist scores for geographic extent and opacity extent, respectively, in stratified Monte Carlo cross-validation experiments. The best performing COVID-Net S networks achieved R[Formula: see text] of 0.739 and 0.741 between predicted scores and radiologist scores for geographic extent and opacity extent, respectively. The results are promising and suggest that the use of deep neural networks on CXRs could be an effective tool for computer-aided assessment of SARS-CoV-2 lung disease severity, although additional studies are needed before adoption for routine clinical use.

Initial chest radiograph scores inform COVID-19 status, intensive care unit admission and need for mechanical ventilation.

AIM: To evaluate whether portable chest radiography (CXR) scores are associated with coronavirus disease 2019 (COVID-19) status and various clinical outcomes. MATERIALS AND METHODS: This retrospective study included 500 initial CXR from COVID-19-suspected patients. Each CXR was scored based on geographic extent and degree of opacity as indicators of disease severity. COVID-19 status and clinical outcomes including intensive care unit (ICU) admission, mechanical ventilation, mortality, length of hospitalisation, and duration on ventilator were collected. Multivariable logistic regression analysis was performed to evaluate the relationship between CXR scores and COVID-19 status, CXR scores and clinical outcomes, adjusted for code status, age, gender and co-morbidities. RESULTS: The interclass correlation coefficients amongst raters were 0.94 and 0.90 for the extent score and opacity score, respectively. CXR scores were significantly (p < 0.01) associated with COVID-19 positivity (odd ratio [OR] = 1.49; 95% confidence interval [CI]: 1.27 - 1.75 for extent score and OR = 1.75; 95% CI: 1.42 - 2.15 for opacity score), ICU admission (OR = 1.19; 95% CI: 1.09 - 1.31 for extent score and OR = 1.26; 95% CI: 1.10 - 1.44 for opacity score), and invasive mechanical ventilation (OR = 1.22; 95% CI: 1.11 - 1.35 for geographic score and OR = 1.21; 95% CI: 1.05 - 1.38 for opacity score). CXR scores were not significantly different between survivors and non-survivors after adjusting for code status (p>0.05). CXR scores were not associated with length of hospitalisation or duration on ventilation (p>0.05). CONCLUSIONS: Initial CXR scores have prognostic value and are associated with COVID-19 positivity, ICU admission, and mechanical ventilation.

Emission Noise in an Interacting Quantum Dot: Role of Inelastic Scattering and Asymmetric Coupling to the Reservoirs.

A theory is developed for the emission noise at frequency ν in a quantum dot in the presence of Coulomb interactions and asymmetric couplings to the reservoirs. We give an analytical expression for the noise in terms of the various transmission amplitudes. Including the inelastic scattering contribution, it can be seen as the analog of the Meir-Wingreen formula for the current. A physical interpretation is given on the basis of the transmission of one electron-hole pair to the concerned reservoir where it emits an energy after recombination. We then treat the interactions by solving the self-consistent equations of motion for the Green functions. The results for the noise derivative versus eV show a zero value until eV=hν, followed by a Kondo peak in the Kondo regime, in good agreement with recent measurements in carbon nanotube quantum dots.

Multimodal MRI characterization of experimental subarachnoid hemorrhage.

Subarachnoid hemorrhage (SAH) is associated with significant morbidity and mortality. We implemented an in-scanner rat model of mild SAH in which blood or vehicle was injected into the cistern magna, and applied multimodal MRI to study the brain prior to, immediately after (5min to 4h), and upto 7days after SAH. Vehicle injection did not change arterial lumen diameter, apparent diffusion coefficient (ADC), T2, venous signal, vascular reactivity to hypercapnia, or foot-fault scores, but mildly reduce cerebral blood flow (CBF) up to 4h, and open-field activity up to 7days post injection. By contrast, blood injection caused: (i) vasospasm 30min after SAH but not thereafter, (ii) venous abnormalities at 3h and 2days, delayed relative to vasospasm, (iii) reduced basal CBF and to hypercapnia 1-4h but not thereafter, (iv) reduced ADC immediately after SAH but no ADC and T2 changes on days 2 and 7, and (v) reduced open-field activities in both SAH and vehicle animals, but no significant differences in open-field activities and foot-fault tests between groups. Mild SAH exhibited transient and mild hemodynamic disturbances and diffusion changes, but did not show apparent ischemic brain injury nor functional deficits.

Changes in MRI signal intensity during hypercapnic challenge under conscious and anesthetized conditions.

Most functional magnetic resonance imaging (fMRI) studies in animals are conducted under anesthesia to minimize motion artifacts. However, methods and techniques have been developed recently for imaging fully conscious rats. Functional MRI studies on conscious animals report enhanced BOLD signal changes as compared to the anesthetized condition. In this study, rats were exposed to different concentrations of carbon dioxide (CO(2)) while conscious and anesthetized to test whether cerebrovascular reactivity may be contributing to these enhanced BOLD signal changes. Hypercapnia produced significantly greater increases in MRI signal intensity in fully conscious animals (6.7-13.3% changes) as when anesthetized with 1% isoflurane (3.2-4.9% changes). In addition, the response to hypercapnia was more immediate in the conscious condition (< 30s) with signal risetimes twice as fast as in the anesthetized state (60s). Both cortical and subcortical brain regions showed a robust, dose- dependent increase in MRI signal intensity with hypercapnic challenge while the animals were conscious but little or no change when anesthetized. Baseline variations in MRI signal were higher while animals were conscious but this was off set by greater signal intensity changes leading to a greater contrast-to-noise ratio, 13.1 in conscious animals, as compared to 8.0 in the anesthetized condition. In summary, cerebral vasculature appears to be more sensitive to hypercapnic challenge in the conscious condition resulting in enhanced T2* MRI signal intensity and the potential for better BOLD signal changes during functional imaging.

Localized cerebral blood flow response at submillimeter columnar resolution.

Functional magnetic resonance imaging (fMRI) has been widely used for imaging brain functions. However, the extent of the fMRI hemodynamic response around the active sites, at submillimeter resolution, remains poorly understood and controversial. With the use of perfusion-based fMRI, we evaluated the hemodynamic response in the cat visual cortex after orientation-specific stimuli. Activation maps obtained by using cerebral blood flow fMRI measurements were predominantly devoid of large draining vein contamination and reproducible at columnar resolution. Stimulus-specific cerebral blood flow responses were spatially localized to individual cortical columns, and columnar layouts were resolved. The periodic spacing of orientation columnar structures was estimated to be 1.1 +/- 0.2 mm (n = 14 orientations, five animals), consistent with previous findings. The estimated cerebral blood flow response at full width at half-maximum was 470 microm under single-stimulus conditions without differential subtraction. These results suggest that hemodynamic-based fMRI can indeed be used to map individual functional columns if large-vessel contributions can be minimized or eliminated.

Functional imaging of brain activity in conscious monkeys responding to sexually arousing cues.

Olfactory cues can elicit intense emotional responses. This study used fMRI in male common marmoset monkeys to identify brain areas associated with sexual arousal in response to odors of ovulating female monkeys. Under light anesthesia, monkeys were secured in a specially designed restrainer and positioned in a 9.4 T magnetic resonance spectrometer. When fully conscious, they were presented with the scents of both ovariectomized and ovulating monkeys. The sexually arousing odors of the ovulating monkeys enhanced signal intensity in the preoptic area and anterior hypothalamus compared to the odors of ovariectomized monkeys. These data corroborate previous findings in monkeys based on invasive electrical lesion and stimulation techniques and demonstrate the feasibility of using non-invasive functional imaging on fully conscious common marmosets to study cue-elicited emotional responses.

Relative changes of cerebral arterial and venous blood volumes during increased cerebral blood flow: implications for BOLD fMRI.

Measurement of cerebral arterial and venous blood volumes during increased cerebral blood flow can provide important information regarding hemodynamic regulation under normal, pathological, and neuronally active conditions. In particular, the change in venous blood volume induced by neural activity is one critical component of the blood oxygenation level-dependent (BOLD) signal because BOLD contrast is dependent only on venous blood, not arterial blood. Thus, relative venous and arterial blood volume (rCBV) and cerebral blood flow (rCBF) in alpha-chlorolase-anesthetized rats under hypercapnia were measured by novel diffusion-weighted (19)F NMR following an i.v. administration of intravascular tracer, perfluorocarbons, and continuous arterial spin labeling methods, respectively. The relationship between rCBF and total rCBV during hypercapnia was rCBV(total) = rCBF(0.40), which is consistent with previous PET measurement in monkeys. This relationship can be linearized in a CBF range of 50-130 ml/100 g/min as DeltarCBV(total)/ DeltarCBF = 0.31 where DeltarCBV and DeltarCBF represent rCBV and rCBF changes. The average arterial volume fraction was 0.25 at a basal condition with CBF of approximately 60 ml/100 g/min and increased up to 0.4 during hypercapnia. The change in venous rCBV was 2-fold smaller than that of total rCBV (DeltarCBV(vein)/DeltarCBF = 0.15), while the arterial rCBV change was 2.5 times larger than that of total rCBV (DeltarCBV(artery)/DeltarCBF = 0.79). These NMR results were confirmed by vessel diameter measurements with in vivo videomicroscopy. The absolute venous blood volume change contributes up to 36% of the total blood volume change during hypercapnia. Our findings provide a quantitative physiological model of BOLD contrast.

Extracellular apparent diffusion in rat brain.

The apparent diffusion coefficients (ADCs) of a series of markers concentrated in the extracellular space of normal rat brain were measured to evaluate, by inference, the ADC of water in the extracellular space. The markers (mannitol, phenylphosphonate, and polyethylene glycols) are defined as "compartment selective" because tissue culture experiments demonstrate some leakage into the intracellular space, making them less "compartment specific" than commonly believed. These primarily extracellular markers have ADCs similar to those of intracellular metabolites of comparable hydrodynamic radius, suggesting that water ADC values in the intra- and extracellular spaces are similar. If this is the case, then it is unlikely that a net shift of water from the extra- to the intracellular space contributes significantly to the reduction in water ADC detected following brain injury. Rather, this reduction is more likely due primarily to a reduction of the ADC of intracellular water associated with injury.

Effect of hyperoxia, hypercapnia, and hypoxia on cerebral interstitial oxygen tension and cerebral blood flow.

The assessment of cerebral interstitial oxygen tension (piO(2)) can provide valuable information regarding cerebrovascular physiology and brain function. Compartment-specific cerebral piO(2) was measured by (19)F NMR following the infusion of an oxygen-sensitive perfluorocarbon directly into the interstitial and ventricular space of the in vivo rat brain. (19)F T(1) measurements were made and cerebral piO(2) were obtained through in vitro calibrations. The effects of graded hyperoxia, hypercapnia, and hypoxia on piO(2) and cerebral blood flow (CBF) were investigated. Under normoxia (arterial pO(2) approximately 120 mm Hg), piO(2) was approximately 30 mm Hg and jugular venous pO(2) was approximately 50 mm Hg. During hyperoxia (arterial pO(2) = 90-300 mm Hg), piO(2) increased linearly with the arterial pO(2). Following hypercapnia (arterial pCO(2) = 20-60 mm Hg), the piO(2) increased sigmoidally with increasing CBF. With hypoxia (arterial pO(2) = 30-40 mm Hg), CBF increased approximately 56% and piO(2) decreased to approximately 15 mm Hg. The hypoxia-induced CBF increase was effective to some extent in compensating for the reduced piO(2). This methodology may prove useful for investigating cerebral piO(2) under pathologically or functionally altered conditions. Magn Reson Med 45:61-70, 2001.

Spatiotemporal dynamics of the BOLD fMRI signals: toward mapping submillimeter cortical columns using the early negative response.

The existence of the early-negative blood-oxygenation-level-dependent (BOLD) response is controversial and its practical utility for mapping brain functions with columnar spatial specificity remains questionable. To address these issues, gradient-echo BOLD fMRI studies were performed at 4.7 T and 9.4 T using the well-established orientation column model in the cat visual cortex. A robust transient early-negative BOLD response was consistently observed in anesthetized cat (-0.35 +/- 0.09%, mean +/- SD, n = 8 at 2.9 +/- 0.5 sec poststimulus onset for 4.7 T, TE = 31 ms; -0.29 +/- 0.10%, n = 4 at 3.0 +/- 0.8 sec poststimulus onset for 9.4 T, TE = 12 ms). In addition to its temporal evolution, the BOLD response also evolved dynamically in the spatial domain. The initially spatially localized early-negative signal appeared to dynamically drain from the active sites toward large vessels, followed by a wave of the delayed positive signal, which exhibited similar spatiotemporal dynamics. Only the early-negative BOLD response within 2 sec of the stimulus onset (not the entire dip) yielded columnar layouts without differential subtraction. The functional maps of two orthogonal orientations using the first 2-sec dip were indeed complementary. On the other hand, the delayed positive BOLD response appeared diffused and extended beyond the active sites. It was thus less suitable to resolve columnar layouts. These results have implications for the design and interpretation of the BOLD fMRI at columnar resolution. Magn Reson Med 44:231-242, 2000.

In vivo MR measurements of regional arterial and venous blood volume fractions in intact rat brain.

In vivo measurement of cerebral arterial and venous volume fractions is important to the understanding of brain physiology and function. By using an intravascular perfluorocarbon and 19F NMR at 4.7 T, regional arterial and venous volume fractions from an intact rat brain were resolved based on the pseudodiffusion coefficients, which were (33 +/- 7) x 10(-3) and (0.45 +/- 0.13) x 10(-3) mm(2)/sec (mean +/- SD, n = 7) for the fast- and slow-moving component, respectively. By exploiting the linear dependence of the perfluorocarbon 19F 1/T1 on the dissolved paramagnetic oxygen concentration, combined inversion-recovery and diffusion measurements were made to correlate the short T1 (high-oxygenation) component with the fast-moving component and the long T1 (low-oxygenation) component with the slow-moving component. The arterial blood volume fraction was 29 +/- 7% of the total cerebral blood volume. Finally, experiments were performed in which different oxygen concentrations were inhaled to validate this technique.

Functional MRI of calcium-dependent synaptic activity: cross correlation with CBF and BOLD measurements.

Spatial specificities of the calcium-dependent synaptic activity, hemodynamic-based blood oxygenation level-dependent (BOLD) and cerebral blood flow (CBF) fMRI were quantitatively compared in the same animals. Calcium-dependent synaptic activity was imaged by exploiting the manganese ion (Mn++) as a calcium analog and an MRI contrast agent at 9.4 T. Following forepaw stimulation in alpha-chloralose anesthetized rat, water T1 of the contralateral forepaw somatosensory cortex (SI) was focally and markedly reduced from 1.99 +/- 0.03 sec to 1.30 +/- 0.18 sec (mean +/- SD, N = 7), resulting from the preferential intracellular Mn++ accumulation. Based on an in vitro calibration, the estimated contralateral somatosensory cortex [Mn++] was approximately 100M, which was 2-5-fold higher than the neighboring tissue and the ipsilateral SI. Regions with the highest calcium activities were localized around cortical layer IV. Stimulus-induced BOLD and CBF changes were 3.4 +/- 1.6% and 98 +/- 33%, respectively. The T1 synaptic activity maps extended along the cortex, whereas the hemodynamic-based activation maps extended radially along the vessels. Spatial overlaps among the synaptic activity, BOLD, and CBF activation maps showed excellent co-registrations. The center-of-mass offsets between any two activation maps were less than 200 microm, suggesting that hemodynamic-based fMRI techniques (at least at high field) can be used to accurately map the spatial loci of synaptic activity.

High-resolution mapping of iso-orientation columns by fMRI.

Blood-oxygenation level-dependent (BOLD) functional magnetic resonance imaging (fMRI) is an important tool for localizing brain functions in vivo. However, the ability of BOLD fMRI to map cortical columnar structures is highly controversial, as the ultimate functional specificity of BOLD remains unknown. Here we report a biphasic BOLD response to visual stimulation in the primary visual cortex of cats. In functional imaging, the initial BOLD signal decrease accurately labeled individual iso-orientation columns. In contrast, the delayed positive BOLD changes indicated the pattern of overall activation in the visual cortex, but were less suited to discriminate active from inactive columns.

Evaluation of extra- and intracellular apparent diffusion in normal and globally ischemic rat brain via 19F NMR.

The biophysical mechanism(s) underlying diffusion-weighted MRI contrast following brain injury remains to be elucidated. Although it is generally accepted that water apparent diffusion coefficient (ADC) decreases after brain injury, it is unknown whether this is associated with a decrease in intracellular or extracellular water displacement, or both. To address this question, 2-[19F]luoro-2-deoxyglucose-6-phosphate (2FDG-6P) was employed as a compartment-specific marker in normal and globally ischemic rat brain. Through judicious choice of routes of administration, 2FDG-6P was confined to the intra- or extracellular space. There was no statistical difference between intra- and extracellular 2FDG-6P ADCs in normal or in globally ischemic brain (P > 0.16), suggesting that water ADCs in both compartments are similar. However, ischemia did result in a 40% ADC decrease in both compartments (P < 0.001). Assuming that 2FDG-6P reflects water motion, this study shows that water ADC decreases in both spaces after ischemia, with the reduction of intracellular water motion being the primary source of diffusion-weighted contrast.

Evaluation of intracellular diffusion in normal and globally-ischemic rat brain via 133Cs NMR.

The question of whether the apparent diffusion coefficient (ADC) of intracellular water changes after brain injury was addressed by using 133Cs as an indicator to report on the state of the intracellular environment. Cesium is an NMR-detectable potassium analog that accumulates in the intracellular space and is detectable in rat brain after being added to the animal's diet. The ADC of cesium was measured before and after the death of the rat. The cesium ADC fell from 0.91 +/- 0.05 x 10(-3) mm2/s (mean +/- SEM, n=5) in the alive rat to 0.71 +/- 0.05 x 10(-3) mm2/s within 20 min (the best time resolution of the experiment) of the death of the animal and stayed at this value for at least 3 h (p < 0.001). Assuming that the ADC of cesium reflects motion in the intracellular environment, these results support the idea that there are changes associated with cell injury that would cause a reduction in the ADC of intracellular water. Hence, one factor contributing to the decrease in water ADC after brain injury is a change in the ADC of intracellular water.