A novel method for tracking nitrogen kinetics in vivo under hyperbaric conditions using radioactive nitrogen-13 gas and positron emission tomography.

Decompression sickness (DCS) is caused by gaseous nitrogen dissolved in tissues forming bubbles during decompression. To date, no method exists to identify nitrogen within tissues, but with advances in positron-emission tomography (PET) technology, it may be possible to track gaseous radionuclides into tissues. We aimed to develop a method to track nitrogen movement in vivo and under hyperbaric pressure that could then be used to further our understanding of DCS using nitrogen-13 (13N2). A single anesthetized female Sprague-Dawley rat was exposed to 625 kPa, composed of air, isoflurane, and 13N2 for 10 min. The PET scanner recorded 13N2 during the hyperbaric exposure with energy windows of 250-750 keV. The PET showed an increase in 13N2 concentration in the lung, heart, and abdominal regions, which all reached a plateau after ∼4 min. This showed that it is possible to gain noninvasive in vivo measurements of nitrogen kinetics through the body while at hyperbaric pressures. Tissue samples showed radioactivity above background levels in the blood, brain, liver, femur, and thigh muscle when assessed using a γ counter. The method can be used to evaluate an array of challenges to our understanding of decompression physiology by quantifying nitrogen load through γ counts of 13N2, and signal intensity of the PET. Further development of the method will improve the specificity of the measured outcomes, and enable it to be used with larger mammals, including humans.NEW & NOTEWORTHY This article describes a method for the in vivo quantification and tracking of nitrogen through the mammalian body whilst exposed to hyperbaric pressure. The method has the potential to further our understanding of decompression sickness, and quantitatively evaluate the effectiveness of both the treatment and prevention of decompression sickness.

Longitudinal assessment of demographic representativeness in the Medical Imaging and Data Resource Center open data commons.

Purpose: The Medical Imaging and Data Resource Center (MIDRC) open data commons was launched to accelerate the development of artificial intelligence (AI) algorithms to help address the COVID-19 pandemic. The purpose of this study was to quantify longitudinal representativeness of the demographic characteristics of the primary MIDRC dataset compared to the United States general population (US Census) and COVID-19 positive case counts from the Centers for Disease Control and Prevention (CDC). Approach: The Jensen-Shannon distance (JSD), a measure of similarity of two distributions, was used to longitudinally measure the representativeness of the distribution of (1) all unique patients in the MIDRC data to the 2020 US Census and (2) all unique COVID-19 positive patients in the MIDRC data to the case counts reported by the CDC. The distributions were evaluated in the demographic categories of age at index, sex, race, ethnicity, and the combination of race and ethnicity. Results: Representativeness of the MIDRC data by ethnicity and the combination of race and ethnicity was impacted by the percentage of CDC case counts for which this was not reported. The distributions by sex and race have retained their level of representativeness over time. Conclusion: The representativeness of the open medical imaging datasets in the curated public data commons at MIDRC has evolved over time as the number of contributing institutions and overall number of subjects have grown. The use of metrics, such as the JSD support measurement of representativeness, is one step needed for fair and generalizable AI algorithm development.

Machine learning with multimodal data for COVID-19.

In response to the unprecedented global healthcare crisis of the COVID-19 pandemic, the scientific community has joined forces to tackle the challenges and prepare for future pandemics. Multiple modalities of data have been investigated to understand the nature of COVID-19. In this paper, MIDRC investigators present an overview of the state-of-the-art development of multimodal machine learning for COVID-19 and model assessment considerations for future studies. We begin with a discussion of the lessons learned from radiogenomic studies for cancer diagnosis. We then summarize the multi-modality COVID-19 data investigated in the literature including symptoms and other clinical data, laboratory tests, imaging, pathology, physiology, and other omics data. Publicly available multimodal COVID-19 data provided by MIDRC and other sources are summarized. After an overview of machine learning developments using multimodal data for COVID-19, we present our perspectives on the future development of multimodal machine learning models for COVID-19.

Toward fairness in artificial intelligence for medical image analysis: identification and mitigation of potential biases in the roadmap from data collection to model deployment.

Purpose: To recognize and address various sources of bias essential for algorithmic fairness and trustworthiness and to contribute to a just and equitable deployment of AI in medical imaging, there is an increasing interest in developing medical imaging-based machine learning methods, also known as medical imaging artificial intelligence (AI), for the detection, diagnosis, prognosis, and risk assessment of disease with the goal of clinical implementation. These tools are intended to help improve traditional human decision-making in medical imaging. However, biases introduced in the steps toward clinical deployment may impede their intended function, potentially exacerbating inequities. Specifically, medical imaging AI can propagate or amplify biases introduced in the many steps from model inception to deployment, resulting in a systematic difference in the treatment of different groups. Approach: Our multi-institutional team included medical physicists, medical imaging artificial intelligence/machine learning (AI/ML) researchers, experts in AI/ML bias, statisticians, physicians, and scientists from regulatory bodies. We identified sources of bias in AI/ML, mitigation strategies for these biases, and developed recommendations for best practices in medical imaging AI/ML development. Results: Five main steps along the roadmap of medical imaging AI/ML were identified: (1) data collection, (2) data preparation and annotation, (3) model development, (4) model evaluation, and (5) model deployment. Within these steps, or bias categories, we identified 29 sources of potential bias, many of which can impact multiple steps, as well as mitigation strategies. Conclusions: Our findings provide a valuable resource to researchers, clinicians, and the public at large.

Special Section Guest Editorial: Global Health, Bias, and Diversity in AI in Medical Imaging.

The editorial comments on the JMI Special Section on Global Health, Bias, and Diversity in AI in Medical Imaging.

Measuring short-term changes in specific ventilation using dynamic specific ventilation imaging.

Specific ventilation imaging (SVI) measures the spatial distribution of specific ventilation (SV) in the lung with MRI by using inhaled oxygen as a contrast agent. Because of the inherently low signal-to-noise ratio in the technique, multiple switches between inspiring air and O2 are utilized, and the high spatial resolution SV distribution is determined as an average over the entire imaging period (∼20 min). We hypothesized that a trade-off between spatial and temporal resolution could allow imaging at a higher temporal resolution, at the cost of a coarser, yet acceptable, spatial resolution. The appropriate window length and spatial resolution compromise were determined by generating synthetic data with signal- and contrast-to-noise characteristics reflective of that in previously published experimental data, with a known and unchanging distribution of SV, and showed that acceptable results could be obtained in an imaging period of ∼7 min (80 breaths), with a spatial resolution of ∼1 cm3. Previously published data were then reanalyzed. The average heterogeneity of the temporally resolved maps of SV was not different from the previous overall analysis, however, the temporally resolved maps were less effective at detecting the amount of bronchoconstriction resulting from methacholine administration. The results further indicated that the initial response to inhaled methacholine and subsequent inhalation of albuterol were largely complete within ∼22 min and ∼9 min, respectively, although there was a tendency for an ongoing developing effect in both cases. These results suggest that it is feasible to use a shortened SVI protocol, with a modest sacrifice in spatial resolution, to measure temporally dynamic processes.NEW & NOTEWORTHY Dynamic imaging providing maps of specific ventilation with a temporal resolution of ∼7 min with a spatial resolution of ∼1 cm3 using MRI was shown to be practical. The technique provides an ionizing radiation free means of temporally following the spatial pattern of specific ventilation. Reanalysis of previously published data showed that the effects of inhaled methacholine and albuterol were largely complete at ∼22 min and ∼9 min, respectively after administration.

Prone positioning redistributes gravitational stress in the lung in normal conditions and in simulations of oedema.

NEW FINDINGS: What is the central question of this study? How does the interaction between posture and gravity affect the stresses on the lung, particularly in highly inflated gravitationally non-dependent regions, which are potentially vulnerable to increased mechanical stress and injury? What is the main finding and its importance? Changes in stress attributable to gravity are not well characterized between postures. Using a new metric of gravitational stress, we show that regions of the lung near maximal inflation have the greatest gravitational stresses while supine, but not while prone. In simulations of increased lung weight consistent with severe pulmonary oedema, the prone lung has lower gravitational stress in vulnerable, non-dependent regions, potentially protecting them from overinflation and injury. ABSTRACT: Prone posture changes the gravitational vector, and potentially the stress induced by tissue deformation, because a larger lung volume is gravitationally dependent when supine, but non-dependent when prone. To evaluate this, 10 normal subjects (six male and four female; age, means ± SD = 27 ± 6 years; height, 171 ± 9 cm; weight, 69 ± 13 kg; forced expiratory volume in the first second/forced expiratory volume as a percentage of predicted, 93 ± 6%) were imaged at functional residual capacity, supine and prone, using magnetic resonance imaging, to quantify regional lung density. We defined regional gravitational stress as the cumulative weight, per unit area, of the column of lung tissue below each point. Gravitational stress was compared between regions of differing inflation to evaluate differences between highly stretched, and thus potentially vulnerable, regions and less stretched lung. Using reference density values for normal lungs at total lung capacity (0.10 ± 0.03 g/ml), regions were classified as highly inflated (density < 0.13 g/ml, i.e., close to total lung capacity), intermediate (0.13 ≤ density < 0.16 g/ml) or normally inflated (density ≥ 0.16 g/ml). Gravitational stress differed between inflation categories while supine (-1.6 ± 0.3 cmH2 O highly inflated; -1.4 ± 0.3 cmH2 O intermediate; -1.1 ± 0.1 cmH2 O normally inflated; P = 0.05) but not while prone (-1.4 ± 0.2 cmH2 O highly inflated; -1.3 ± 0.2 cmH2 O intermediate; -1.3 ± 0.1 cmH2 O normally inflated; P = 0.39), and increased more with height from dependent lung while supine (-0.24 ± 0.02 cmH2 O/cm supine; -0.18 ± 0.04 cmH2 O/cm prone; P = 0.05). In simulated severe pulmonary oedema, the gradient in gravitational stress increased in both postures (all P < 0.0001), was greater in the supine posture than when prone (-0.57 ± 0.21 cmH2 O/cm supine; -0.34 ± 0.16 cmH2 O/cm prone; P = 0.0004) and was similar to the gradient calculated from supine computed tomography images in a patient with acute respiratory distress syndrome (-0.51 cmH2 O/cm). The non-dependent lung has greater gravitational stress while supine and might be protected while prone, particularly in the presence of oedema.

Abnormal pulmonary perfusion heterogeneity in patients with Fontan circulation and pulmonary arterial hypertension.

KEY POINTS: The distribution of pulmonary perfusion is affected by gravity, vascular branching structure and active regulatory mechanisms, which may be disrupted by cardiopulmonary disease, but this is not well studied, particularly in rare conditions. We evaluated pulmonary perfusion in patients who had undergone Fontan procedure, patients with pulmonary arterial hypertension (PAH) and two groups of controls using a proton magnetic resonance imaging technique, arterial spin labelling to measure perfusion. Heterogeneity was assessed by the relative dispersion (SD/mean) and gravitational gradients. Gravitational gradients were similar between all groups, but heterogeneity was significantly increased in both patient groups compared to controls and persisted after removing contributions from large blood vessels and gravitational gradients. Patients with Fontan physiology and patients with PAH have increased pulmonary perfusion heterogeneity that is not explainable by differences in mean perfusion, gravitational gradients, or large vessel anatomy. This probably reflects vascular remodelling in PAH and possibly in Fontan physiology. ABSTRACT: Many factors affect the distribution of pulmonary perfusion, which may be disrupted by cardiopulmonary disease, but this is not well studied, particularly in rare conditions. An example is following the Fontan procedure, where pulmonary perfusion is passive, and heterogeneity may be increased because of the underlying pathophysiology leading to Fontan palliation, remodelling, or increased gravitational gradients from low flow. Another is pulmonary arterial hypertension (PAH), where gravitational gradients may be reduced secondary to high pressures, but remodelling may increase perfusion heterogeneity. We evaluated regional pulmonary perfusion in Fontan patients (n = 5), healthy young controls (Fontan control, n = 5), patients with PAH (n = 6) and healthy older controls (PAH control) using proton magnetic resonance imaging. Regional perfusion was measured using arterial spin labelling. Heterogeneity was assessed by the relative dispersion (SD/mean) and gravitational gradients. Mean perfusion was similar (Fontan = 2.50 ± 1.02 ml min-1  ml-1 ; Fontan control = 3.09 ± 0.58, PAH = 3.63 ± 1.95; PAH control = 3.98 ± 0.91, P = 0.26), and the slopes of gravitational gradients were not different (Fontan = -0.23 ± 0.09 ml min-1  ml-1  cm-1 ; Fontan control = -0.29 ± 0.23, PAH = -0.27 ± 0.09, PAH control = -0.25 ± 0.18, P = 0.91) between groups. Perfusion relative dispersion was greater in both Fontan and PAH than controls (Fontan = 1.46 ± 0.18; Fontan control = 0.99 ± 0.21, P = 0.005; PAH = 1.22 ± 0.27, PAH control = 0.91 ± 0.12, P = 0.02) but similar between patient groups (P = 0.13). These findings persisted after removing contributions from large blood vessels and gravitational gradients (all P < 0.05). We conclude that patients with Fontan physiology and PAH have increased pulmonary perfusion heterogeneity that is not explained by differences in mean perfusion, gravitational gradients, or large vessel anatomy. This probably reflects the effects of remodelling in PAH and possibly in Fontan physiology.

Vaping disrupts ventilation-perfusion matching in asymptomatic users.

Inhalation of e-cigarette's aerosols (vaping) has the potential to disrupt pulmonary gas exchange, but the effects in asymptomatic users are unknown. We assessed ventilation-perfusion (V̇A/Q̇) mismatch in asymptomatic e-cigarette users, using magnetic resonance imaging (MRI). We hypothesized that vaping induces V̇A/Q̇ mismatch through alterations in both ventilation and perfusion distributions. Nine young, asymptomatic "Vapers" with >1-yr vaping history, and no history of cardiopulmonary disease, were imaged supine using proton MRI, to assess the right lung at baseline and immediately after vaping. Seven young "Controls" were imaged at baseline only. Relative dispersion (SD/means) was used to quantify the heterogeneity of the individual ventilation and perfusion distributions. V̇A/Q̇ mismatch was quantified using the second moments of the ventilation and perfusion versus V̇A/Q̇ ratio distributions, log scale, LogSDV̇, and LogSDQ̇, respectively, analogous to the multiple inert gas elimination technique. Spirometry was normal in both groups. Ventilation heterogeneity was similar between groups at baseline (Vapers, 0.43 ± 0.13; Controls, 0.51 ± 0.11; P = 0.13) but increased after vaping (to 0.57 ± 0.17; P = 0.03). Perfusion heterogeneity was greater (P = 0.04) in Vapers at baseline (0.53 ± 0.06) compared with Controls (0.44 ± 0.10) but decreased after vaping (to 0.42 ± 0.07; P = 0.005). Vapers had greater (P = 0.01) V̇A/Q̇ mismatch at baseline compared with Controls (LogSDQ̇ = 0.61 ± 0.12 vs. 0.43 ± 0.12), which was increased after vaping (LogSDQ̇ = 0.73 ± 0.16; P = 0.03). V̇A/Q̇ mismatch is greater in Vapers and worsens after vaping. This suggests subclinical alterations in lung function not detected by spirometry.NEW & NOTEWORTHY This research provides evidence of vaping-induced disruptions in ventilation-perfusion matching in young, healthy, asymptomatic adults with normal spirometry who habitually vape. The changes in ventilation and perfusion distributions, both at baseline and acutely after vaping, and the potential implications on hypoxic vasoconstriction are particularly relevant in understanding the pathogenesis of vaping-induced dysfunction. Our imaging-based approach provides evidence of potential subclinical alterations in lung function below thresholds of detection using spirometry.

Ventilatory heterogeneity in the normal human lung is unchanged by controlled breathing.

Measurement of ventilation heterogeneity with the multiple-breath nitrogen washout (MBW) is usually performed using controlled breathing with a fixed tidal volume and breathing frequency. However, it is unclear whether controlled breathing alters the underlying ventilatory heterogeneity. We hypothesized that the width of the specific ventilation distribution (a measure of heterogeneity) would be greater in tests performed during free breathing compared with those performed using controlled breathing. Eight normal subjects (age range = 23-50 yr, 5 female/3 male) twice underwent MRI-based specific ventilation imaging consisting of five repeated cycles with the inspired gas switching between 21% and 100% O2 every ~2 min (total imaging time = ~20 min). In each session, tests were performed with free breathing (FB, no constraints) and controlled breathing (CB) at a respiratory rate of 12 breaths/min and no tidal volume control. The specific ventilation (SV) distribution in a mid-sagittal slice of the right lung was calculated, and the heterogeneity was calculated as the full width at half max of a Gaussian distribution fitted on a log scale (SV width). Free breathing resulted in a range of breathing frequencies from 8.7 to 15.9 breaths/min (mean = 11.5 ± 2.2, P = 0.62, compared with CB). Heterogeneity (SV width) was unchanged by controlled breathing (FB: 0.38 ± 0.12; CB: 0.34 ± 0.09, P = 0.18, repeated-measures ANOVA). The imposition of a controlled breathing frequency did not significantly affect the heterogeneity of ventilation in the normal lung, suggesting that MBW and specific ventilation imaging as typically performed provide an unperturbed measure of ventilatory heterogeneity.NEW & NOTEWORTHY By using MRI-based specific ventilation imaging (SVI), we showed that the heterogeneity of specific ventilation was not different comparing free breathing and breathing with the imposition of a fixed breathing frequency of 12 breaths/min. Thus, multiple-breath washout and SVI as typically performed provide an unperturbed measure of ventilatory heterogeneity.