Uncovering atherosclerotic cardiovascular disease by PET imaging

Assessing atherosclerosis severity is essential for precise patient stratification. Specifically, there is a need to identify patients with residual inflammation because these patients remain at high risk of cardiovascular events despite optimal management of cardiovascular risk factors. Molecular imaging techniques, such as PET, can have an essential role in this context. PET imaging can indicate tissue-based disease status, detect early molecular changes and provide whole-body information. Advances in molecular biology and bioinformatics continue to help to decipher the complex pathogenesis of atherosclerosis and inform the development of imaging tracers. Concomitant advances in tracer synthesis methods and PET imaging technology provide future possibilities for atherosclerosis imaging. In this Review, we summarize the latest developments in PET imaging techniques and technologies for assessment of atherosclerotic cardiovascular disease and discuss the relationship between imaging readouts and transcriptomics-based plaque phenotyping.

Hybrid Magnetic Resonance Positron Emission Tomography Is Associated With Cardiac-Related Outcomes in Cardiac Sarcoidosis

BACKGROUND

Imaging with late gadolinium enhancement (LGE) magnetic resonance (MR) and 18F-fluorodeoxyglucose (18F-FDG) PET allows complementary assessment of myocardial injury and disease activity and has shown promise for improved characterization of active cardiac sarcoidosis (CS) based on the combined positive imaging outcome, MR(+)PET(+).

OBJECTIVES

This study aims to evaluate qualitative and quantitative assessments of hybrid MR/PET imaging in CS and to evaluate its association with cardiac-related outcomes.

METHODS

A total of 148 patients with suspected CS underwent hybrid MR/PET imaging. Patients were classified based on the presence/absence of LGE (MR+/MR-), presence/absence of 18F-FDG (PET+/PET-), and pattern of 18F-FDG uptake (focal/diffuse) into the following categories: MR(+)PET(+)FOCAL, MR(+)PET(+)DIFFUSE, MR(+)PET(-), MR(-)PET(+)FOCAL, MR(-)PET(+)DIFFUSE, MR(-)PET(-). Further analysis classified MR positivity based on %LGE exceeding 5.7% as MR(+/-)5.7%. Quantitative values of standard uptake value, target-to-background ratio, target-to-normal-myocardium ratio (TNMRmax), and T2 were measured. The primary clinical endpoint was met by the occurrence of cardiac arrest, ventricular tachycardia, or secondary prevention implantable cardioverter-defibrillator (ICD) before the end of the study. The secondary endpoint was met by any of the primary endpoint criteria plus heart failure or heart block. MR/PET imaging results were compared between those meeting or not meeting the clinical endpoints.

RESULTS

Patients designated MR(+)5.7%PET(+)FOCAL had increased odds of meeting the primary clinical endpoint compared to those with all other imaging classifications (unadjusted OR: 9.2 [95% CI: 3.0-28.7]; P = 0.0001), which was higher than the odds based on MR or PET alone. TNMRmax achieved an area under the receiver-operating characteristic curve of 0.90 for separating MR(+)PET(+)FOCAL from non-MR(+)PET(+)FOCAL, and 0.77 for separating those reaching the clinical endpoint from those not reaching the clinical endpoint.

CONCLUSIONS

Hybrid MR/PET image-based classification of CS was statistically associated with clinical outcomes in CS. TNMRmax had modest sensitivity and specificity for quantifying the imaging-based classification MR(+)PET(+)FOCAL and was associated with outcomes. Use of combined MR and PET image-based classification may have use in prognostication and treatment management in CS.

18F-FDG PET/CT in left ventricular assist device infections: In-depth characterization and clinical implications

BACKGROUND

Previous retrospective studies suggest a good diagnostic performance of 18F-fluorodeoxyglucose positron emission tomography (18F-FDG-PET)/computed tomography (CT) in left ventricular assist device (LVAD) infections. Our aim was to prospectively evaluate the role of PET/CT in the characterization and impact on clinical management of LVAD infections.

METHODS

A total of 40 patients (aged 58 [53-62] years) with suspected LVAD infection and 5 controls (aged 69 [64-71] years) underwent 18F-FDG-PET/CT. Four LVAD components were evaluated: exit site and subcutaneous driveline (peripheral), pump pocket, and outflow graft. The location with maximal uptake was considered the presumed site of infection. Infection was confirmed by positive culture (exit site or blood) and/or surgical findings.

RESULTS

Visual uptake was present in 40 patients (100%) in the infection group vs 4 (80%) control subjects. For each individual component, the presence of uptake was more frequent in the infection than in the control group. The location of maximal uptake was most frequently the pump pocket (48%) in the infection group and the peripheral components (75%) in the control group. Maximum standard uptake values (SUVmax) were higher in the infection than in the control group: SUVmax (average all components): 6.9 (5.1-8.5) vs 3.8 (3.7-4.3), p = 0.002; SUVmax (location of maximal uptake): 10.6 ± 4.0 vs 5.4 ± 1.9, p = 0.01. Pump pocket infections were more frequent in patients with bacteremia than without bacteremia (79% vs 31%, p = 0.011). Pseudomonas (32%) and methicillin-susceptible Staphylococcus aureus (29%) were the most frequent pathogens and were associated with pump pocket infections, while Staphylococcus epidermis (11%) was associated with peripheral infections. PET/CT affected the clinical management of 83% of patients with infection, resulting in surgical debridement (8%), pump exchange (13%), and upgrade in the transplant listing status (10%), leading to 8% of urgent transplants.

CONCLUSIONS

18F-FDG-PET/CT enables the diagnosis and characterization of the extent of LVAD infections, which can significantly affect the clinical management of these patients.

A review of self-supervised, generative, and few-shot deep learning methods for data-limited magnetic resonance imaging segmentation

Magnetic resonance imaging (MRI) is a ubiquitous medical imaging technology with applications in disease diagnostics, intervention, and treatment planning. Accurate MRI segmentation is critical for diagnosing abnormalities, monitoring diseases, and deciding on a course of treatment. With the advent of advanced deep learning frameworks, fully automated and accurate MRI segmentation is advancing. Traditional supervised deep learning techniques have advanced tremendously, reaching clinical-level accuracy in the field of segmentation. However, these algorithms still require a large amount of annotated data, which is oftentimes unavailable or impractical. One way to circumvent this issue is to utilize algorithms that exploit a limited amount of labeled data. This paper aims to review such state-of-the-art algorithms that use a limited number of annotated samples. We explain the fundamental principles of self-supervised learning, generative models, few-shot learning, and semi-supervised learning and summarize their applications in cardiac, abdomen, and brain MRI segmentation. Throughout this review, we highlight algorithms that can be employed based on the quantity of annotated data available. We also present a comprehensive list of notable publicly available MRI segmentation datasets. To conclude, we discuss possible future directions of the field-including emerging algorithms, such as contrastive language-image pretraining, and potential combinations across the methods discussed-that can further increase the efficacy of image segmentation with limited labels.

PTSD increases risk for major adverse cardiovascular events through neural and cardio-inflammatory pathways

While posttraumatic stress disorder (PTSD) is known to associate with an elevated risk for major adverse cardiovascular events (MACE), few studies have examined mechanisms underlying this link. Recent studies have demonstrated that neuro-immune mechanisms, (manifested by heightened stress-associated neural activity (SNA), autonomic nervous system activity, and inflammation), link common stress syndromes to MACE. However, it is unknown if neuro-immune mechanisms similarly link PTSD to MACE. The current study aimed to test the hypothesis that upregulated neuro-immune mechanisms increase MACE risk among individuals with PTSD. This study included N = 118,827 participants from a large hospital-based biobank. Demographic, diagnostic, and medical history data collected from the biobank. SNA (n = 1,520), heart rate variability (HRV; [n = 11,463]), and high sensitivity C-reactive protein (hs-CRP; [n = 15,164]) were obtained for a subset of participants. PTSD predicted MACE after adjusting for traditional MACE risk factors (hazard ratio (HR) [95 % confidence interval (CI)] = 1.317 [1.098, 1.580], β = 0.276, p = 0.003). The PTSD-to-MACE association was mediated by SNA (CI = 0.005, 0.133, p < 0.05), HRV (CI = 0.024, 0.056, p < 0.05), and hs-CRP (CI = 0.010, 0.040, p < 0.05). This study provides evidence that neuro-immune pathways may play important roles in the mechanisms linking PTSD to MACE. Future studies are needed to determine if these markers are relevant targets for PTSD treatment and if improvements in SNA, HRV, and hs-CRP associate with reduced MACE risk in this patient population.

Acid ceramidase regulates innate immune memory

Innate immune memory, also called "trained immunity," is a functional state of myeloid cells enabling enhanced immune responses. This phenomenon is important for host defense, but also plays a role in various immune-mediated conditions. We show that exogenously administered sphingolipids and inhibition of sphingolipid metabolizing enzymes modulate trained immunity. In particular, we reveal that acid ceramidase, an enzyme that converts ceramide to sphingosine, is a potent regulator of trained immunity. We show that acid ceramidase regulates the transcription of histone-modifying enzymes, resulting in profound changes in histone 3 lysine 27 acetylation and histone 3 lysine 4 trimethylation. We confirm our findings by identifying single-nucleotide polymorphisms in the region of ASAH1, the gene encoding acid ceramidase, that are associated with the trained immunity cytokine response. Our findings reveal an immunomodulatory effect of sphingolipids and identify acid ceramidase as a relevant therapeutic target to modulate trained immunity responses in innate immune-driven disorders.

Discovery Viewer (DV): Web-Based Medical AI Model Development Platform and Deployment Hub

The rapid rise of artificial intelligence (AI) in medicine in the last few years highlights the importance of developing bigger and better systems for data and model sharing. However, the presence of Protected Health Information (PHI) in medical data poses a challenge when it comes to sharing. One potential solution to mitigate the risk of PHI breaches is to exclusively share pre-trained models developed using private datasets. Despite the availability of these pre-trained networks, there remains a need for an adaptable environment to test and fine-tune specific models tailored for clinical tasks. This environment should be open for peer testing, feedback, and continuous model refinement, allowing dynamic model updates that are especially important in the medical field, where diseases and scanning techniques evolve rapidly. In this context, the Discovery Viewer (DV) platform was developed in-house at the Biomedical Engineering and Imaging Institute at Mount Sinai (BMEII) to facilitate the creation and distribution of cutting-edge medical AI models that remain accessible after their development. The all-in-one platform offers a unique environment for non-AI experts to learn, develop, and share their own deep learning (DL) concepts. This paper presents various use cases of the platform, with its primary goal being to demonstrate how DV holds the potential to empower individuals without expertise in AI to create high-performing DL models. We tasked three non-AI experts to develop different musculoskeletal AI projects that encompassed segmentation, regression, and classification tasks. In each project, 80% of the samples were provided with a subset of these samples annotated to aid the volunteers in understanding the expected annotation task. Subsequently, they were responsible for annotating the remaining samples and training their models through the platform's "Training Module". The resulting models were then tested on the separate 20% hold-off dataset to assess their performance. The classification model achieved an accuracy of 0.94, a sensitivity of 0.92, and a specificity of 1. The regression model yielded a mean absolute error of 14.27 pixels. And the segmentation model attained a Dice Score of 0.93, with a sensitivity of 0.9 and a specificity of 0.99. This initiative seeks to broaden the community of medical AI model developers and democratize the access of this technology to all stakeholders. The ultimate goal is to facilitate the transition of medical AI models from research to clinical settings.

Enhancing the brain MRI at ultra-high field systems using a meta-array structure

BACKGROUND

The main advantage of ultra-high field (UHF) magnetic resonance neuroimaging is theincreased signal-to-noise ratio (SNR) compared with lower field strength imaging. However, the wavelength effect associated with UHF MRI results in radiofrequency (RF) inhomogeneity, compromising whole brain coverage for many commercial coils. Approaches to resolving this issue of transmit field inhomogeneity include the design of parallel transmit systems (PTx), RF pulse design, and applying passive RF shimming such as high dielectric materials. However, these methods have some drawbacks such as unstable material parameters of dielectric pads, high-cost, and complexity of PTx systems. Metasurfaces are artificial structures with a unique platform that can control the propagation of the electromagnetic (EM) waves, and they are very promising for engineering EM device. Implementation of meta-arrays enhancing MRI has been explored previously in several studies.

PURPOSE

The aim of this study was to assess the effect of new meta-array technology on enhancing the brain MRI at 7T. A meta-array based on a hybrid structure consisting of an array of broadside-coupled split-ring resonators and high-permittivity materials was designed to work at the Larmor frequency of a 7 Tesla (7T) MRI scanner. When placed behind the head and neck, this construct improves the SNR in the region of the cerebellum,brainstem and the inferior aspect of the temporal lobes.

METHODS

Numerical electromagnetic simulations were performed to optimize the meta-array design parameters and determine the RF circuit configuration. The resultant transmit-efficiency and signal sensitivity improvements were experimentally analyzed in phantoms followed by healthy volunteers using a 7T whole-body MRI scanner equipped with a standard one-channel transmit, 32-channel receive head coil. Efficacy was evaluated through acquisition with and without the meta-array using two basic sequences: gradient-recalled-echo (GRE) and turbo-spin-echo (TSE).

RESULTS

Experimental phantom analysis confirmed two-fold improvement in the transmit efficiency and 1.4-fold improvement in the signal sensitivity in the target region. In vivo GRE and TSE images with the meta-array in place showed enhanced visualization in inferior regions of the brain, especially of the cerebellum, brainstem, and cervical spinal cord.

CONCLUSION

Addition of the meta-array to commonly used MRI coils can enhance SNR to extend the anatomical coverage of the coil and improve overall MRI coil performance. This enhancement in SNR can be leveraged to obtain a higher resolution image over the same time slot or faster acquisition can be achieved with same resolution. Using this technique could improve the performance of existing commercial coils at 7T for whole brain and other applications.

Development of the ehive Digital Health App: Protocol for a Centralized Research Platform

BACKGROUND

The increasing use of smartphones, wearables, and connected devices has enabled the increasing application of digital technologies for research. Remote digital study platforms comprise a patient-interfacing digital application that enables multimodal data collection from a mobile app and connected sources. They offer an opportunity to recruit at scale, acquire data longitudinally at a high frequency, and engage study participants at any time of the day in any place. Few published descriptions of centralized digital research platforms provide a framework for their development.

OBJECTIVE

This study aims to serve as a road map for those seeking to develop a centralized digital research platform. We describe the technical and functional aspects of the ehive app, the centralized digital research platform of the Hasso Plattner Institute for Digital Health at Mount Sinai Hospital, New York, New York. We then provide information about ongoing studies hosted on ehive, including usership statistics and data infrastructure. Finally, we discuss our experience with ehive in the broader context of the current landscape of digital health research platforms.

METHODS

The ehive app is a multifaceted and patient-facing central digital research platform that permits the collection of e-consent for digital health studies. An overview of its development, its e-consent process, and the tools it uses for participant recruitment and retention are provided. Data integration with the platform and the infrastructure supporting its operations are discussed; furthermore, a description of its participant- and researcher-facing dashboard interfaces and the e-consent architecture is provided.

RESULTS

The ehive platform was launched in 2020 and has successfully hosted 8 studies, namely 6 observational studies and 2 clinical trials. Approximately 1484 participants downloaded the app across 36 states in the United States. The use of recruitment methods such as bulk messaging through the EPIC electronic health records and standard email portals enables broad recruitment. Light-touch engagement methods, used in an automated fashion through the platform, maintain high degrees of engagement and retention. The ehive platform demonstrates the successful deployment of a central digital research platform that can be modified across study designs.

CONCLUSIONS

Centralized digital research platforms such as ehive provide a novel tool that allows investigators to expand their research beyond their institution, engage in large-scale longitudinal studies, and combine multimodal data streams. The ehive platform serves as a model for groups seeking to develop similar digital health research programs.

INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID)

DERR1-10.2196/49204.

A Hybrid Decision Tree and Deep Learning Approach Combining Medical Imaging and Electronic Medical Records to Predict Intubation Among Hospitalized Patients With COVID-19: Algorithm Development and Validation

BACKGROUND

Early prediction of the need for invasive mechanical ventilation (IMV) in patients hospitalized with COVID-19 symptoms can help in the allocation of resources appropriately and improve patient outcomes by appropriately monitoring and treating patients at the greatest risk of respiratory failure. To help with the complexity of deciding whether a patient needs IMV, machine learning algorithms may help bring more prognostic value in a timely and systematic manner. Chest radiographs (CXRs) and electronic medical records (EMRs), typically obtained early in patients admitted with COVID-19, are the keys to deciding whether they need IMV.

OBJECTIVE

We aimed to evaluate the use of a machine learning model to predict the need for intubation within 24 hours by using a combination of CXR and EMR data in an end-to-end automated pipeline. We included historical data from 2481 hospitalizations at The Mount Sinai Hospital in New York City.

METHODS

CXRs were first resized, rescaled, and normalized. Then lungs were segmented from the CXRs by using a U-Net algorithm. After splitting them into a training and a test set, the training set images were augmented. The augmented images were used to train an image classifier to predict the probability of intubation with a prediction window of 24 hours by retraining a pretrained DenseNet model by using transfer learning, 10-fold cross-validation, and grid search. Then, in the final fusion model, we trained a random forest algorithm via 10-fold cross-validation by combining the probability score from the image classifier with 41 longitudinal variables in the EMR. Variables in the EMR included clinical and laboratory data routinely collected in the inpatient setting. The final fusion model gave a prediction likelihood for the need of intubation within 24 hours as well.

RESULTS

At a prediction probability threshold of 0.5, the fusion model provided 78.9% (95% CI 59%-96%) sensitivity, 83% (95% CI 76%-89%) specificity, 0.509 (95% CI 0.34-0.67) F1-score, 0.874 (95% CI 0.80-0.94) area under the receiver operating characteristic curve (AUROC), and 0.497 (95% CI 0.32-0.65) area under the precision recall curve (AUPRC) on the holdout set. Compared to the image classifier alone, which had an AUROC of 0.577 (95% CI 0.44-0.73) and an AUPRC of 0.206 (95% CI 0.08-0.38), the fusion model showed significant improvement (P<.001). The most important predictor variables were respiratory rate, C-reactive protein, oxygen saturation, and lactate dehydrogenase. The imaging probability score ranked 15th in overall feature importance.

CONCLUSIONS

We show that, when linked with EMR data, an automated deep learning image classifier improved performance in identifying hospitalized patients with severe COVID-19 at risk for intubation. With additional prospective and external validation, such a model may assist risk assessment and optimize clinical decision-making in choosing the best care plan during the critical stages of COVID-19.

Semi-Automatic Graphical Tool for Measuring Coronary Artery Spatially Weighted Calcium Score from Gated Cardiac Computed Tomography Images

The current standard for measuring coronary artery calcification to determine the extent of atherosclerosis is by calculating the Agatston score from computed tomography (CT). However, the Agatston score disregards pixel values less than 130 Hounsfield Units (HU) and calcium regions less than 1 mm2. Due to this thresholding, the score is not sensitive to small, weakly attenuating regions of calcium deposition and may not detect nascent micro-calcification. A recently proposed metric called the spatially weighted calcium score (SWCS) also utilizes CT but does not include a threshold for HU and does not require elevated signals in contiguous pixels. Thus, the SWCS is sensitive to weakly attenuating, smaller calcium deposits and may improve the measurement of coronary heart disease risk. Currently, the SWCS is underutilized owing to the added computational complexity. To promote translation of the SWCS into clinical research and reliable, repeatable computation of the score, the aim of this study was to develop a semi-automatic graphical tool that calculates both the SWCS and the Agatston score. The program requires gated cardiac CT scans with a calcium hydroxyapatite phantom in the field of view. The phantom allows for deriving a weighting function, from which each pixel's weight is adjusted, allowing for the mitigation of signal variations and variability between scans. With all three anatomical views visible simultaneously, the user traces the course of the four main coronary arteries by placing points or regions of interest. Features such as scroll-to-zoom, double-click to delete, and brightness/contrast adjustment, along with written guidance at every step, make the program user-friendly and easy to use. Once tracing the arteries is complete, the program generates reports, which include the scores and snapshots of any visible calcium. The SWCS may reveal the presence of subclinical disease, which may be used for early intervention and lifestyle changes.

Resolving sepsis-induced immunoparalysis via trained immunity by targeting interleukin-4 to myeloid cells

Immunoparalysis is a compensatory and persistent anti-inflammatory response to trauma, sepsis or another serious insult, which increases the risk of opportunistic infections, morbidity and mortality. Here, we show that in cultured primary human monocytes, interleukin-4 (IL4) inhibits acute inflammation, while simultaneously inducing a long-lasting innate immune memory named trained immunity. To take advantage of this paradoxical IL4 feature in vivo, we developed a fusion protein of apolipoprotein A1 (apoA1) and IL4, which integrates into a lipid nanoparticle. In mice and non-human primates, an intravenously injected apoA1-IL4-embedding nanoparticle targets myeloid-cell-rich haematopoietic organs, in particular, the spleen and bone marrow. We subsequently demonstrate that IL4 nanotherapy resolved immunoparalysis in mice with lipopolysaccharide-induced hyperinflammation, as well as in ex vivo human sepsis models and in experimental endotoxemia. Our findings support the translational development of nanoparticle formulations of apoA1-IL4 for the treatment of patients with sepsis at risk of immunoparalysis-induced complications.

Arrhythmic Mitral Valve Prolapse With Only Mild or Moderate Mitral Regurgitation: Characterization of Myocardial Substrate

BACKGROUND

Sustained ventricular tachycardia and sudden cardiac death due to degenerative mitral valve prolapse (MVP) can occur in the absence of severe mitral regurgitation (MR). A significant percentage of patients with MVP-related sudden death do not have any evidence of replacement fibrosis, suggesting other unrecognized proarrhythmic factors may place these patients at risk.

OBJECTIVES

This study aims to characterize myocardial fibrosis/inflammation and ventricular arrhythmia complexity in patients with MVP and only mild or moderate MR.

METHODS

Prospective observational study of patients with MVP and only mild or moderate MR underwent ventricular arrhythmia characterization and hybrid positron emission tomography (PET)/magnetic resonance imaging (MRI). Coregistered hybrid 18F-fluorodeoxyglucose (18F-FDG)-PET and MRI late gadolinium enhancement images were assessed and categorized. Recruitment occurred in the cardiac electrophysiology clinic.

RESULTS

In 12 patients with degenerative MVP with only mild or moderate MR, of which a majority had complex ventricular ectopy (n = 10, 83%), focal (or focal-on-diffuse) uptake of 18F-FDG (PET-positive) was detected in 83% (n = 10) of patients. Three-quarters of the patients (n = 9, 75%) had FDG uptake that coexisted with areas of late gadolinium enhancement (PET/MRI-positive). Abnormal T1, T2 and extracellular volume (ECV) values were observed in 58% (n = 7), 25% (n = 3), and 16% (n = 2), respectively.

CONCLUSIONS

Most patients with degenerative MVP, ventricular ectopy, and mild or moderate MR show myocardial inflammation that is concordant with myocardial scar. Further study is needed to determine whether these findings contribute to the observation that most MVP-related sudden deaths occur in patients with less than severe MR.

Deep Learning for Automated Measurement of Patellofemoral Anatomic Landmarks

BACKGROUND

Patellofemoral anatomy has not been well characterized. Applying deep learning to automatically measure knee anatomy can provide a better understanding of anatomy, which can be a key factor in improving outcomes.

METHODS

483 total patients with knee CT imaging (April 2017-May 2022) from 6 centers were selected from a cohort scheduled for knee arthroplasty and a cohort with healthy knee anatomy. A total of 7 patellofemoral landmarks were annotated on 14,652 images and approved by a senior musculoskeletal radiologist. A two-stage deep learning model was trained to predict landmark coordinates using a modified ResNet50 architecture initialized with self-supervised learning pretrained weights on RadImageNet. Landmark predictions were evaluated with mean absolute error, and derived patellofemoral measurements were analyzed with Bland-Altman plots. Statistical significance of measurements was assessed by paired t-tests.

RESULTS

Mean absolute error between predicted and ground truth landmark coordinates was 0.20/0.26 cm in the healthy/arthroplasty cohort. Four knee parameters were calculated, including transepicondylar axis length, transepicondylar-posterior femur axis angle, trochlear medial asymmetry, and sulcus angle. There were no statistically significant parameter differences (p > 0.05) between predicted and ground truth measurements in both cohorts, except for the healthy cohort sulcus angle.

CONCLUSION

Our model accurately identifies key trochlear landmarks with ~0.20-0.26 cm accuracy and produces human-comparable measurements on both healthy and pathological knees. This work represents the first deep learning regression model for automated patellofemoral annotation trained on both physiologic and pathologic CT imaging at this scale. This novel model can enhance our ability to analyze the anatomy of the patellofemoral compartment at scale.

A machine learning approach to determine resilience utilizing wearable device data: analysis of an observational cohort

Objective

To assess whether an individual's degree of psychological resilience can be determined from physiological metrics passively collected from a wearable device.

Materials and Methods

Data were analyzed in this secondary analysis of the Warrior Watch Study dataset, a prospective cohort of healthcare workers enrolled across 7 hospitals in New York City. Subjects wore an Apple Watch for the duration of their participation. Surveys were collected measuring resilience, optimism, and emotional support at baseline.

Results

We evaluated data from 329 subjects (mean age 37.4 years, 37.1% male). Across all testing sets, gradient-boosting machines (GBM) and extreme gradient-boosting models performed best for high- versus low-resilience prediction, stratified on a median Connor-Davidson Resilience Scale-2 score of 6 (interquartile range = 5-7), with an AUC of 0.60. When predicting resilience as a continuous variable, multivariate linear models had a correlation of 0.24 (P = .029) and RMSE of 1.37 in the testing data. A positive psychological construct, comprised of resilience, optimism, and emotional support was also evaluated. The oblique random forest method performed best in estimating high- versus low-composite scores stratified on a median of 32.5, with an AUC of 0.65, a sensitivity of 0.60, and a specificity of 0.70.

Discussion

In a post hoc analysis, machine learning models applied to physiological metrics collected from wearable devices had some predictive ability in identifying resilience states and a positive psychological construct.

Conclusions

These findings support the further assessment of psychological characteristics from passively collected wearable data in dedicated studies.

Multiparametric Immunoimaging Maps Inflammatory Signatures in Murine Myocardial Infarction Models

In the past 2 decades, research on atherosclerotic cardiovascular disease has uncovered inflammation to be a key driver of the pathophysiological process. A pressing need therefore exists to quantitatively and longitudinally probe inflammation, in preclinical models and in cardiovascular disease patients, ideally using non-invasive methods and at multiple levels. Here, we developed and employed in vivo multiparametric imaging approaches to investigate the immune response following myocardial infarction. The myocardial infarction models encompassed either transient or permanent left anterior descending coronary artery occlusion in C57BL/6 and Apoe-/-mice. We performed nanotracer-based fluorine magnetic resonance imaging and positron emission tomography (PET) imaging using a CD11b-specific nanobody and a C-C motif chemokine receptor 2-binding probe. We found that immune cell influx in the infarct was more pronounced in the permanent occlusion model. Further, using 18F-fluorothymidine and 18F-fluorodeoxyglucose PET, we detected increased hematopoietic activity after myocardial infarction, with no difference between the models. Finally, we observed persistent systemic inflammation and exacerbated atherosclerosis in Apoe-/- mice, regardless of which infarction model was used. Taken together, we showed the strengths and capabilities of multiparametric imaging in detecting inflammatory activity in cardiovascular disease, which augments the development of clinical readouts.

Reduced Stress-Related Neural Network Activity Mediates the Effect of Alcohol on Cardiovascular Risk

BACKGROUND

Chronic stress associates with major adverse cardiovascular events (MACE) via increased stress-related neural network activity (SNA). Light/moderate alcohol consumption (ACl/m) has been linked to lower MACE risk, but the mechanisms are unclear.

OBJECTIVES

The purpose of this study was to evaluate whether the association between ACl/m and MACE is mediated by decreased SNA.

METHODS

Individuals enrolled in the Mass General Brigham Biobank who completed a health behavior survey were studied. A subset underwent 18F-fluorodeoxyglucose positron emission tomography, enabling assessment of SNA. Alcohol consumption was classified as none/minimal, light/moderate, or high (<1, 1-14, or >14 drinks/week, respectively).

RESULTS

Of 53,064 participants (median age 60 years, 60% women), 23,920 had no/minimal alcohol consumption and 27,053 ACl/m. Over a median follow-up of 3.4 years, 1,914 experienced MACE. ACl/m (vs none/minimal) associated with lower MACE risk (HR: 0.786; 95% CI: 0.717-0.862; P < 0.0001) after adjusting for cardiovascular risk factors. In 713 participants with brain imaging, ACl/m (vs none/minimal) associated with decreased SNA (standardized beta -0.192; 95% CI: -0.338 to -0.046; P = 0.01). Lower SNA partially mediated the beneficial effect of ACl/m on MACE (log OR: -0.040; 95% CI: -0.097 to -0.003; P < 0.05). Further, ACl/m associated with larger decreases in MACE risk among individuals with (vs without) prior anxiety (HR: 0.60 [95% CI: 0.50-0.72] vs 0.78 [95% CI: 0.73-0.80]; P interaction = 0.003).

CONCLUSIONS

ACl/m associates with reduced MACE risk, in part, by lowering activity of a stress-related brain network known for its association with cardiovascular disease. Given alcohol's potential health detriments, new interventions with similar effects on SNA are needed.

Systems immunology-based drug repurposing framework to target inflammation in atherosclerosis

The development of new immunotherapies to treat the inflammatory mechanisms that sustain atherosclerotic cardiovascular disease (ASCVD) is urgently needed. Herein, we present a path to drug repurposing to identify immunotherapies for ASCVD. The integration of time-of-flight mass cytometry and RNA sequencing identified unique inflammatory signatures in peripheral blood mononuclear cells stimulated with ASCVD plasma. By comparing these inflammatory signatures to large-scale gene expression data from the LINCS L1000 dataset, we identified drugs that could reverse this inflammatory response. Ex vivo screens, using human samples, showed that saracatinib-a phase 2a-ready SRC and ABL inhibitor-reversed the inflammatory responses induced by ASCVD plasma. In Apoe-/- mice, saracatinib reduced atherosclerosis progression by reprogramming reparative macrophages. In a rabbit model of advanced atherosclerosis, saracatinib reduced plaque inflammation measured by [18F] fluorodeoxyglucose positron emission tomography-magnetic resonance imaging. Here we show a systems immunology-driven drug repurposing with a preclinical validation strategy to aid the development of cardiovascular immunotherapies.

Exploring the Utility of Cardiovascular Magnetic Resonance Radiomic Feature Extraction for Evaluation of Cardiac Sarcoidosis

BACKGROUND

The aim of this study is to explore the utility of cardiac magnetic resonance (CMR) imaging of radiomic features to distinguish active and inactive cardiac sarcoidosis (CS).

METHODS

Subjects were classified into active cardiac sarcoidosis (CS_active) and inactive cardiac sarcoidosis (CS_inactive) based on PET-CMR imaging. CS_active was classified as featuring patchy [¹⁸F]fluorodeoxyglucose ([¹⁸F]FDG) uptake on PET and presence of late gadolinium enhancement (LGE) on CMR, while CS_inactive was classified as featuring no [¹⁸F]FDG uptake in the presence of LGE on CMR. Among those screened, thirty CS_active and thirty-one CS_inactive patients met these criteria. A total of 94 radiomic features were subsequently extracted using PyRadiomics. The values of individual features were compared between CS_active and CS_inactive using the Mann-Whitney U test. Subsequently, machine learning (ML) approaches were tested. ML was applied to two sub-sets of radiomic features (signatures A and B) that were selected by logistic regression and PCA, respectively.

RESULTS

Univariate analysis of individual features showed no significant differences. Of all features, gray level co-occurrence matrix (GLCM) joint entropy had a good area under the curve (AUC) and accuracy with the smallest confidence interval, suggesting it may be a good target for further investigation. Some ML classifiers achieved reasonable discrimination between CS_active and CS_inactive patients. With signature A, support vector machine and k-neighbors showed good performance with AUC (0.77 and 0.73) and accuracy (0.67 and 0.72), respectively. With signature B, decision tree demonstrated AUC and accuracy around 0.7; Conclusion: CMR radiomic analysis in CS provides promising results to distinguish patients with active and inactive disease.

Influence of thoracic radiology training on classification of interstitial lung diseases

INTRODUCTION

Interpretation of high-resolution CT images plays an important role in the diagnosis and management of interstitial lung diseases. However, interreader variation may exist due to varying levels of training and expertise. This study aims to evaluate interreader variation and the role of thoracic radiology training in classifying interstitial lung disease (ILD).

METHODS

This is a retrospective study where seven physicians (radiologists, thoracic radiologists, and a pulmonologist) classified the subtypes of ILD of 128 patients from a tertiary referral center, all selected from the Interstitial Lung Disease Registry which consists of patients from November 2014 to January 2021. Each patient was diagnosed with a subtype of interstitial lung disease by a consensus diagnosis from pathology, radiology, and pulmonology. Each reader was provided with only clinical history, only CT images, or both. Reader sensitivity and specificity and interreader agreements using Cohen's κ were calculated.

RESULTS

Interreader agreement based only on clinical history, only on radiologic information, or combination of both was most consistent amongst readers with thoracic radiology training, ranging from fair (Cohen's κ: 0.2-0.46), moderate to almost perfect (Cohen's κ: 0.55-0.92), and moderate to almost perfect (Cohen's κ: 0.53-0.91) respectively. Radiologists with any thoracic training showed both increased sensitivity and specificity for NSIP as compared to other radiologists and the pulmonologist when using only clinical history, only CT information, or combination of both (p < 0.05).

CONCLUSIONS

Readers with thoracic radiology training showed the least interreader variation and were more sensitive and specific at classifying certain subtypes of ILD.

SUMMARY SENTENCE

Thoracic radiology training may improve sensitivity and specificity in classifying ILD based on HRCT images and clinical history.

Interstitial lung disease diagnosis and prognosis using an AI system integrating longitudinal data

For accurate diagnosis of interstitial lung disease (ILD), a consensus of radiologic, pathological, and clinical findings is vital. Management of ILD also requires thorough follow-up with computed tomography (CT) studies and lung function tests to assess disease progression, severity, and response to treatment. However, accurate classification of ILD subtypes can be challenging, especially for those not accustomed to reading chest CTs regularly. Dynamic models to predict patient survival rates based on longitudinal data are challenging to create due to disease complexity, variation, and irregular visit intervals. Here, we utilize RadImageNet pretrained models to diagnose five types of ILD with multimodal data and a transformer model to determine a patient's 3-year survival rate. When clinical history and associated CT scans are available, the proposed deep learning system can help clinicians diagnose and classify ILD patients and, importantly, dynamically predict disease progression and prognosis.

The combined effect of air and transportation noise pollution on atherosclerotic inflammation and risk of cardiovascular disease events

BACKGROUND

Air pollution and noise exposures individually associate with major adverse cardiovascular events (MACE) via a mechanism involving arterial inflammation (ArtI); however, their combined impact on ArtI and MACE remains unknown. We tested whether dual (vs. one or neither) exposure associates with greater ArtI and MACE risk and whether MACE risk is mediated via ArtI.

METHODS

Individuals (N = 474) without active cancer or known cardiovascular disease with clinical 18F-FDG-PET/CT imaging were followed for 5 years for MACE. ArtI was measured. Average air pollution (particulate matter ≤ 2.5 μm, PM2.5) and transportation noise exposure were determined at individual residences. Higher exposures were defined as noise > 55 dBA (World Health Organization cutoff) and PM2.5 ≥ sample median.

RESULTS

At baseline, 46%, 46%, and 8% were exposed to high levels of neither, one, or both pollutants; 39 experienced MACE over a median 4.1 years. Exposure to an increasing number of pollutants associated with higher ArtI (standardized β [95% CI: .195 [.052, .339], P = .008) and MACE (HR [95% CI]: 2.897 [1.818-4.615], P < .001). In path analysis, ArtI partially mediated the relationship between pollutant exposures and MACE (P < .05).

CONCLUSION

Air pollution and transportation noise exposures contribute incrementally to ArtI and MACE. The mechanism linking dual exposure to MACE involves ArtI.

Sleep apnea and carotid atherosclerosis in the Multi-Ethnic Study of Atherosclerosis (MESA): leveraging state-of-the-art vascular imaging

PURPOSE

To further characterize the relationship between obstructive sleep apnea (OSA) and carotid atherosclerosis, we examined the structural and metabolic features of carotid plaque using hybrid 18-F-fluorodeoxyglucose (FDG) Positron Emission Tomography/Magnetic Resonance Imaging (PET/MRI) in the Multi-Ethnic Study of Atherosclerosis (MESA).

METHODS

We studied 46 individuals from the MESA-PET and MESA-Sleep ancillary studies. OSA was defined as an apnea hypopnea index [AHI] ≥ 15 events per hour (4% desaturation). PET/MRI was used to measure carotid plaque inflammation (using target-to-background-ratios [TBR]) and carotid wall thickness (CWT). Linear regression was used to assess the associations between OSA, CWT and TBR.

RESULTS

The mean age was 67.9 years (SD 8.53) and the mean BMI was 28.9 kg/m² (SD 4.47). There was a trend toward a higher mean CWT in the OSA (n = 11) vs. non-OSA group (n = 35), 1.51 vs. 1.41 (p = 0.098). TBR did not differ by OSA groups, and there was no significant association between OSA and carotid plaque inflammation (TBR) in adjusted analyses. Although there was a significant interaction between OSA and obesity, there were no statistically significant associations between OSA and vascular inflammation in stratified analysis by obesity.

CONCLUSION

Despite a trend toward a higher carotid wall thickness in OSA vs. non-OSA participants, we did not find an independent association between OSA and carotid plaque inflammation using PET/MRI in MESA. Our findings suggest that simultaneous assessments of structural and metabolic features of atherosclerosis may fill current knowledge gaps pertaining to the influence of OSA on atherosclerosis prevalence and progression.

PET/MR imaging of inflammation in atherosclerosis

Myocardial infarction, stroke, mental disorders, neurodegenerative processes, autoimmune diseases, cancer and the human immunodeficiency virus impact the haematopoietic system, which through immunity and inflammation may aggravate pre-existing atherosclerosis. The interplay between the haematopoietic system and its modulation of atherosclerosis has been studied by imaging the cardiovascular system and the activation of haematopoietic organs via scanners integrating positron emission tomography and resonance imaging (PET/MRI). In this Perspective, we review the applicability of integrated whole-body PET/MRI for the study of immune-mediated phenomena associated with haematopoietic activity and cardiovascular disease, and discuss the translational opportunities and challenges of the technology.

Performance of spiral UTE-MRI of the lung in post-COVID patients

Patients recovered from COVID-19 may develop long-COVID symptoms in the lung. For this patient population (post-COVID patients), they may benefit from longitudinal, radiation-free lung MRI exams for monitoring lung lesion development and progression. The purpose of this study was to investigate the performance of a spiral ultrashort echo time MRI sequence (Spiral-VIBE-UTE) in a cohort of post-COVID patients in comparison with CT and to compare image quality obtained using different spiral MRI acquisition protocols. Lung MRI was performed in 36 post-COVID patients with different acquisition protocols, including different spiral sampling reordering schemes (line in partition or partition in line) and different breath-hold positions (inspiration or expiration). Three experienced chest radiologists independently scored all the MR images for different pulmonary structures. Lung MR images from spiral acquisition protocol that received the highest image quality scores were also compared against corresponding CT images in 27 patients for evaluating diagnostic image quality and lesion identification. Spiral-VIBE-UTE MRI acquired with the line in partition reordering scheme in an inspiratory breath-holding position achieved the highest image quality scores (score range = 2.17-3.69) compared to others (score range = 1.7-3.29). Compared to corresponding chest CT images, three readers found that 81.5% (22 out of 27), 81.5% (22 out of 27) and 37% (10 out of 27) of the MR images were useful, respectively. Meanwhile, they all agreed that MRI could identify significant lesions in the lungs. The Spiral-VIBE-UTE sequence allows for fast imaging of the lung in a single breath hold. It could be a valuable tool for lung imaging without radiation and could provide great value for managing different lung diseases including assessment of post-COVID lesions.

Somatostatin Receptor PET/MR Imaging of Inflammation in Patients With Large Vessel Vasculitis and Atherosclerosis

BACKGROUND

Assessing inflammatory disease activity in large vessel vasculitis (LVV) can be challenging by conventional measures.

OBJECTIVES

We aimed to investigate somatostatin receptor 2 (SST2) as a novel inflammation-specific molecular imaging target in LVV.

METHODS

In a prospective, observational cohort study, in vivo arterial SST2 expression was assessed by positron emission tomography/magnetic resonance imaging (PET/MRI) using 68Ga-DOTATATE and 18F-FET-βAG-TOCA. Ex vivo mapping of the imaging target was performed using immunofluorescence microscopy; imaging mass cytometry; and bulk, single-cell, and single-nucleus RNA sequencing.

RESULTS

Sixty-one participants (LVV: n = 27; recent atherosclerotic myocardial infarction of ≤2 weeks: n = 25; control subjects with an oncologic indication for imaging: n = 9) were included. Index vessel SST2 maximum tissue-to-blood ratio was 61.8% (P < 0.0001) higher in active/grumbling LVV than inactive LVV and 34.6% (P = 0.0002) higher than myocardial infarction, with good diagnostic accuracy (area under the curve: ≥0.86; P < 0.001 for both). Arterial SST2 signal was not elevated in any of the control subjects. SST2 PET/MRI was generally consistent with 18F-fluorodeoxyglucose PET/computed tomography imaging in LVV patients with contemporaneous clinical scans but with very low background signal in the brain and heart, allowing for unimpeded assessment of nearby coronary, myocardial, and intracranial artery involvement. Clinically effective treatment for LVV was associated with a 0.49 ± 0.24 (standard error of the mean [SEM]) (P = 0.04; 22.3%) reduction in the SST2 maximum tissue-to-blood ratio after 9.3 ± 3.2 months. SST2 expression was localized to macrophages, pericytes, and perivascular adipocytes in vasculitis specimens, with specific receptor binding confirmed by autoradiography. SSTR2-expressing macrophages coexpressed proinflammatory markers.

CONCLUSIONS

SST2 PET/MRI holds major promise for diagnosis and therapeutic monitoring in LVV. (PET Imaging of Giant Cell and Takayasu Arteritis [PITA], NCT04071691; Residual Inflammation and Plaque Progression Long-Term Evaluation [RIPPLE], NCT04073810).

Self-gated, dynamic contrast-enhanced magnetic resonance imaging with compressed-sensing reconstruction for evaluating endothelial permeability in the aortic root of atherosclerotic mice

High-risk atherosclerotic plaques are characterized by active inflammation and abundant leaky microvessels. We present a self-gated, dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) acquisition with compressed sensing reconstruction and apply it to assess longitudinal changes in endothelial permeability in the aortic root of Apoe-/- atherosclerotic mice during natural disease progression. Twenty-four, 8-week-old, female Apoe-/- mice were divided into four groups (n = 6 each) and imaged with self-gated DCE-MRI at 4, 8, 12, and 16 weeks after high-fat diet initiation, and then euthanized for CD68 immunohistochemistry for macrophages. Eight additional mice were kept on a high-fat diet and imaged longitudinally at the same time points. Aortic-root pseudo-concentration curves were analyzed using a validated piecewise linear model. Contrast agent wash-in and washout slopes (b1 and b2 ) were measured as surrogates of aortic root endothelial permeability and compared with macrophage density by immunohistochemistry. b2 , indicating contrast agent washout, was significantly higher in mice kept on an high-fat diet for longer periods of time (p = 0.03). Group comparison revealed significant differences between mice on a high-fat diet for 4 versus 16 weeks (p = 0.03). Macrophage density also significantly increased with diet duration (p = 0.009). Spearman correlation between b2 from DCE-MRI and macrophage density indicated a weak relationship between the two parameters (r = 0.28, p = 0.20). Validated piecewise linear modeling of the DCE-MRI data showed that the aortic root contrast agent washout rate is significantly different during disease progression. Further development of this technique from a single-slice to a 3D acquisition may enable better investigation of the relationship between in vivo imaging of endothelial permeability and atherosclerotic plaques' genetic, molecular, and cellular makeup in this important model of disease.

Multidimensional immunoimaging to characterize the inflammatory reaction after permanent and temporary coronary artery occlusion in mice

Background

The immune response following acute myocardial infarction encompasses a delicate balance between inflammatory and reparative programs. Our knowledge of these complex mechanisms is mainly derived from studies using a murine model of permanent coronary artery occlusion. In this study we developed, validated and implemented multiparametric imaging methods to investigate cardiac function and the systemic immune response in transient or permanent coronary artery occlusion mouse models.

Methods

The myocardial infarction models encompassed either transient (40 min) or permanent LAD occlusion in C57BL/6 mice, and non-infarcted mice were used as controls. Two or seven days later, the animals subjected to systemic immunoimaging of the bone marrow, spleen and myocardium with late gadolinium enhancement cardiac MRI (LGE cMRI), 18F-fluorodeoxyglucose (18F-FDG) PET, 18F-Fluorothymidine (18F-FLT) PET, 64Cu-CCR2 PET targeting inflammatory monocytes, 89Zr-CD11b nanobody PET and 19F-HDL-PERFECTA MRI, both targeting myeloid cells. In addition, the same myocardial infarction models were applied to atherosclerosis-prone Apoe−/− and systemic inflammation and plaque progression were assessed by flow cytometry and immunohistochemistry four weeks after infarction.

Results

Through LGE cMRI and 18F-FDG PET, we observed that temporary coronary occlusion resulted in a smaller infarct size, better cardiac function and viability compared to permanent occlusion. Multiparametric immunoimaging targeting CD11b+ cells by 89Zr-CD11b nanobody PET and Ly6Chi inflammatory monocytes by 64Cu-CCR2 PET demonstrated that mice subjected to transient coronary occlusion had less immune cell influx to the ischemic myocardium. This finding was confirmed by flow cytometry analysis of the infarct zone. In contrast, both myocardial infarction models cause a similar systemic immune response in the bone marrow and spleen as observed with multimodal imaging with subsequent similar numbers of CD11b+ cells in the blood. Both permanent and temporary coronary artery occlusion aggravate atherosclerosis in Apoe−/− mice with higher macrophage and Ly6Chi monocyte numbers in aortas and larger plaque size compared to Apoe−/− mice without myocardial infarction.

Conclusions

We developed and employed multimodal, multiparametric imaging protocols to characterize the immune response in the heart, bone marrow and spleen in two models of myocardial infarction. While cardiac function was superior in the ischemia reperfusion model, both types of myocardial infarction accelerated atherosclerosis.

Funding Acknowledgement

Type of funding sources: Other. Main funding source(s): NIH and DFG

FSH-blocking therapeutic for osteoporosis

Pharmacological and genetic studies over the past decade have established the follicle-stimulating hormone (FSH) as an actionable target for diseases affecting millions, namely osteoporosis, obesity, and Alzheimer's disease. Blocking FSH action prevents bone loss, fat gain and neurodegeneration in mice. We recently developed a first-in-class, humanized, epitope-specific FSH-blocking antibody, MS-Hu6, with a K_D of 7.52 nM. Using a GLP-compliant platform, we now report the efficacy of MS-Hu6 in preventing and treating osteoporosis in mice and parameters of acute safety in monkeys. Biodistribution studies using ⁸⁹Zr-labelled, biotinylated or unconjugated MS-Hu6 in mice and monkeys showed localization to bone and bone marrow. MS-Hu6 displayed a β phase t_½ of 7.5 days (180 hours) in humanized Tg32 mice. We tested 217 variations of excipients using the protein thermal shift assay to generate a final formulation that rendered MS-Hu6 stable in solution upon freeze-thaw and at different temperatures, with minimal aggregation, and without self-, cross-, or hydrophobic interactions or appreciable binding to relevant human antigens. MS-Hu6 showed the same level of 'humanness' as human IgG1 in silico and was non-immunogenic in ELISPOT assays for IL-2 and IFNg in human peripheral blood mononuclear cell cultures. We conclude that MS-Hu6 is efficacious, durable, and manufacturable, and is therefore poised for future human testing.

U-Net Based Segmentation and Characterization of Gliomas

(1) Background: Gliomas are the most common primary brain neoplasms accounting for roughly 40−50% of all malignant primary central nervous system tumors. We aim to develop a deep learning-based framework for automated segmentation and prediction of biomarkers and prognosis in patients with gliomas. (2) Methods: In this retrospective two center study, patients were included if they (1) had a diagnosis of glioma with known surgical histopathology and (2) had preoperative MRI with FLAIR sequence. The entire tumor volume including FLAIR hyperintense infiltrative component and necrotic and cystic components was segmented. Deep learning-based U-Net framework was developed based on symmetric architecture from the 512 × 512 segmented maps from FLAIR as the ground truth mask. (3) Results: The final cohort consisted of 208 patients with mean ± standard deviation of age (years) of 56 ± 15 with M/F of 130/78. DSC of the generated mask was 0.93. Prediction for IDH-1 and MGMT status had a performance of AUC 0.88 and 0.62, respectively. Survival prediction of <18 months demonstrated AUC of 0.75. (4) Conclusions: Our deep learning-based framework can detect and segment gliomas with excellent performance for the prediction of IDH-1 biomarker status and survival.

Inflammatory Diseases of the Aorta: JACC Focus Seminar, Part 2

Inflammatory aortitis is most often caused by large vessel vasculitis (LVV), including giant cell arteritis, Takayasu's arteritis, immunoglobulin G4-related aortitis, and isolated aortitis. There are distinct differences in the clinical presentation, imaging findings, and natural history of LVV that are important for the cardiovascular provider to know. If possible, histopathologic specimens should be obtained to aide in accurate diagnosis and management of LVV. In most cases, corticosteroids are utilized in the acute phase, with the addition of steroid-sparing agents to achieve disease remission while sparing corticosteroid toxic effects. Endovascular and surgical procedures have been described with success but should be delayed until disease control is achieved whenever possible. Long-term management should include regular follow-up with rheumatology and surveillance imaging for sequelae of LVV.

Scan-rescan measurement repeatability of 18F-FDG PET/MR imaging of vascular inflammation

Non-invasive positron emission tomography (PET) of vascular inflammation and atherosclerotic plaque by identifying increased uptake of 18F-fluordeoxyglucose (18F-FDG) is a powerful tool for monitoring disease activity, progression, and its response to therapy. 18F-FDG PET/computed tomography (PET/CT) of the aorta and carotid arteries has become widely used to assess changes in inflammation in clinical trials. However, the recent advent of hybrid PET/magnetic resonance (PET/MR) scanners has advantages for vascular imaging due to the reduction in radiation exposure and improved soft tissue contrast of MR compared to CT. Important for research and clinical use is an understanding of the scan-rescan repeatability of the PET measurement. While this has been studied for PET/CT, no data is currently available for vascular PET/MR imaging. In this study, we determined the scan-rescan measurement repeatability of 18F-FDG PET/MR in the aorta and carotid arteries was less than 5%, comparable to similar findings for 18F-FDG PET/CT.

RadImageNet: An Open Radiologic Deep Learning Research Dataset for Effective Transfer Learning

Purpose

To demonstrate the value of pretraining with millions of radiologic images compared with ImageNet photographic images on downstream medical applications when using transfer learning.

Materials and Methods

This retrospective study included patients who underwent a radiologic study between 2005 and 2020 at an outpatient imaging facility. Key images and associated labels from the studies were retrospectively extracted from the original study interpretation. These images were used for RadImageNet model training with random weight initiation. The RadImageNet models were compared with ImageNet models using the area under the receiver operating characteristic curve (AUC) for eight classification tasks and using Dice scores for two segmentation problems.

Results

The RadImageNet database consists of 1.35 million annotated medical images in 131 872 patients who underwent CT, MRI, and US for musculoskeletal, neurologic, oncologic, gastrointestinal, endocrine, abdominal, and pulmonary pathologic conditions. For transfer learning tasks on small datasets—thyroid nodules (US), breast masses (US), anterior cruciate ligament injuries (MRI), and meniscal tears (MRI)—the RadImageNet models demonstrated a significant advantage (P < .001) to ImageNet models (9.4%, 4.0%, 4.8%, and 4.5% AUC improvements, respectively). For larger datasets—pneumonia (chest radiography), COVID-19 (CT), SARS-CoV-2 (CT), and intracranial hemorrhage (CT)—the RadImageNet models also illustrated improved AUC (P < .001) by 1.9%, 6.1%, 1.7%, and 0.9%, respectively. Additionally, lesion localizations of the RadImageNet models were improved by 64.6% and 16.4% on thyroid and breast US datasets, respectively.

Conclusion

RadImageNet pretrained models demonstrated better interpretability compared with ImageNet models, especially for smaller radiologic datasets.

Keywords: CT, MR Imaging, US, Head/Neck, Thorax, Brain/Brain Stem, Evidence-based Medicine, Computer Applications–General (Informatics)

Supplemental material is available for this article.

Published under a CC BY 4.0 license.

See also the commentary by Cadrin-Chênevert in this issue.

Subclinical Atherosclerosis in Young, Socioeconomically Vulnerable Hispanic and Non-Hispanic Black Adults

BACKGROUND

Non-Hispanic Black persons are at greater risk of cardiovascular (CV) events than other racial/ethnic groups; however, their differential vulnerability to early subclinical atherosclerosis is poorly understood.

OBJECTIVES

This work aims to study the impact of race/ethnicity on early subclinical atherosclerosis in young socioeconomically disadvantaged adults.

METHODS

Bilateral carotid and femoral 3-dimensional vascular ultrasound examinations were performed on 436 adults (parents/caregivers and staff) with a mean age of 38.0 ± 11.1 years, 82.3% female, 66% self-reported as Hispanic, 34% self-reported as non-Hispanic Black, and no history of CV disease recruited in the FAMILIA (Family-Based Approach in a Minority Community Integrating Systems-Biology for Promotion of Health) trial from 15 Head Start preschools in Harlem (neighborhood in New York, New York, USA). The 10-year Framingham CV risk score was calculated, and the relationship between race/ethnicity and the presence and extent of subclinical atherosclerosis was analyzed with multivariable logistic and linear regression models.

RESULTS

The mean 10-year Framingham CV risk was 4.0%, with no differences by racial/ethnic category. The overall prevalence of subclinical atherosclerosis was significantly higher in the non-Hispanic Black (12.9%) than in the Hispanic subpopulation (6.6%). After adjusting for 10-year Framingham CV risk score, body mass index, fruit and vegetable consumption, physical activity, and employment status, non-Hispanic Black individuals were more likely than Hispanic individuals to have subclinical atherosclerosis (OR: 3.45; 95% CI: 1.44-8.29; P = 0.006) and multiterritorial disease (P = 0.026).

CONCLUSIONS

After adjustment for classic CV risk, lifestyle, and socioeconomic factors, non-Hispanic Black younger adults seem more vulnerable to early subclinical atherosclerosis than their Hispanic peers, suggesting that the existence of emerging or undiscovered CV factors underlying the residual excess risk (Family-Based Approach in a Minority Community Integrating Systems-Biology for Promotion of Health [FAMILIA (Project 2)]; NCT02481401).

Evaluation of a machine learning approach utilizing wearable data for prediction of SARS-CoV-2 infection in healthcare workers

Objective

To determine whether a machine learning model can detect SARS-CoV-2 infection from physiological metrics collected from wearable devices.

Materials and Methods

Health care workers from 7 hospitals were enrolled and prospectively followed in a multicenter observational study. Subjects downloaded a custom smart phone app and wore Apple Watches for the duration of the study period. Daily surveys related to symptoms and the diagnosis of Coronavirus Disease 2019 were answered in the app.

Results

We enrolled 407 participants with 49 (12%) having a positive nasal SARS-CoV-2 polymerase chain reaction test during follow-up. We examined 5 machine-learning approaches and found that gradient-boosting machines (GBM) had the most favorable validation performance. Across all testing sets, our GBM model predicted SARS-CoV-2 infection with an average area under the receiver operating characteristic (auROC) = 86.4% (confidence interval [CI] 84-89%). The model was calibrated to value sensitivity over specificity, achieving an average sensitivity of 82% (CI ±∼4%) and specificity of 77% (CI ±∼1%). The most important predictors included parameters describing the circadian heart rate variability mean (MESOR) and peak-timing (acrophase), and age.

Discussion

We show that a tree-based ML algorithm applied to physiological metrics passively collected from a wearable device can identify and predict SARS-CoV-2 infection.

Conclusion

Applying machine learning models to the passively collected physiological metrics from wearable devices may improve SARS-CoV-2 screening methods and infection tracking.

Clonally expanded CD8 T cells characterize amyotrophic lateral sclerosis-4

Amyotrophic lateral sclerosis (ALS) is a heterogenous neurodegenerative disorder that affects motor neurons and voluntary muscle control¹. ALS heterogeneity includes the age of manifestation, the rate of progression and the anatomical sites of symptom onset. Disease-causing mutations in specific genes have been identified and define different subtypes of ALS¹. Although several ALS-associated genes have been shown to affect immune functions², whether specific immune features account for ALS heterogeneity is poorly understood. Amyotrophic lateral sclerosis-4 (ALS4) is characterized by juvenile onset and slow progression³. Patients with ALS4 show motor difficulties by the time that they are in their thirties, and most of them require devices to assist with walking by their fifties. ALS4 is caused by mutations in the senataxin gene (SETX). Here, using Setx knock-in mice that carry the ALS4-causative L389S mutation, we describe an immunological signature that consists of clonally expanded, terminally differentiated effector memory (T_EMRA) CD8 T cells in the central nervous system and the blood of knock-in mice. Increased frequencies of antigen-specific CD8 T cells in knock-in mice mirror the progression of motor neuron disease and correlate with anti-glioma immunity. Furthermore, bone marrow transplantation experiments indicate that the immune system has a key role in ALS4 neurodegeneration. In patients with ALS4, clonally expanded T_EMRA CD8 T cells circulate in the peripheral blood. Our results provide evidence of an antigen-specific CD8 T cell response in ALS4, which could be used to unravel disease mechanisms and as a potential biomarker of disease state.

Systematically evaluating DOTATATE and FDG as PET immuno-imaging tracers of cardiovascular inflammation

In recent years, cardiovascular immuno-imaging by positron emission tomography (PET) has undergone tremendous progress in preclinical settings. Clinically, two approved PET tracers hold great potential for inflammation imaging in cardiovascular patients, namely FDG and DOTATATE. While the former is a widely applied metabolic tracer, DOTATATE is a relatively new PET tracer targeting the somatostatin receptor 2 (SST2). In the current study, we performed a detailed, head-to-head comparison of DOTATATE-based radiotracers and [18F]F-FDG in mouse and rabbit models of cardiovascular inflammation. For mouse experiments, we labeled DOTATATE with the long-lived isotope [64Cu]Cu to enable studying the tracer's mode of action by complementing in vivo PET/CT experiments with thorough ex vivo immunological analyses. For translational PET/MRI rabbit studies, we employed the more widely clinically used [68Ga]Ga-labeled DOTATATE, which was approved by the FDA in 2016. DOTATATE's pharmacokinetics and timed biodistribution were determined in control and atherosclerotic mice and rabbits by ex vivo gamma counting of blood and organs. Additionally, we performed in vivo PET/CT experiments in mice with atherosclerosis, mice subjected to myocardial infarction and control animals, using both [64Cu]Cu-DOTATATE and [18F]F-FDG. To evaluate differences in the tracers' cellular specificity, we performed ensuing ex vivo flow cytometry and gamma counting. In mice subjected to myocardial infarction, in vivo [64Cu]Cu-DOTATATE PET showed higher differential uptake between infarcted (SUVmax 1.3, IQR, 1.2-1.4, N = 4) and remote myocardium (SUVmax 0.7, IQR, 0.5-0.8, N = 4, p = 0.0286), and with respect to controls (SUVmax 0.6, IQR, 0.5-0.7, N = 4, p = 0.0286), than [18F]F-FDG PET. In atherosclerotic mice, [64Cu]Cu-DOTATATE PET aortic signal, but not [18F]F-FDG PET, was higher compared to controls (SUVmax 1.1, IQR, 0.9-1.3 and 0.5, IQR, 0.5-0.6, respectively, N = 4, p = 0.0286). In both models, [64Cu]Cu-DOTATATE demonstrated preferential accumulation in macrophages with respect to other myeloid cells, while [18F]F-FDG was taken up by macrophages and other leukocytes. In a translational PET/MRI study in atherosclerotic rabbits, we then compared [68Ga]Ga-DOTATATE and [18F]F-FDG for the assessment of aortic inflammation, combined with ex vivo radiometric assays and near-infrared imaging of macrophage burden. Rabbit experiments showed significantly higher aortic accumulation of both [68Ga]Ga-DOTATATE and [18F]F-FDG in atherosclerotic (SUVmax 0.415, IQR, 0.338-0.499, N = 32 and 0.446, IQR, 0.387-0.536, N = 27, respectively) compared to control animals (SUVmax 0.253, IQR, 0.197-0.285, p = 0.0002, N = 10 and 0.349, IQR, 0.299-0.423, p = 0.0159, N = 11, respectively). In conclusion, we present a detailed, head-to-head comparison of the novel SST2-specific tracer DOTATATE and the validated metabolic tracer [18F]F-FDG for the evaluation of inflammation in small animal models of cardiovascular disease. Our results support further investigations on the use of DOTATATE to assess cardiovascular inflammation as a complementary readout to the widely used [18F]F-FDG.

18F-fluoride PET/MR in cardiac amyloid: A comparison study with aortic stenosis and age- and sex-matched controls

OBJECTIVES

Cardiac MR is widely used to diagnose cardiac amyloid, but cannot differentiate AL and ATTR subtypes: an important distinction given their differing treatments and prognoses. We used PET/MR imaging to quantify myocardial uptake of 18F-fluoride in ATTR and AL amyloid patients, as well as participants with aortic stenosis and age/sex-matched controls.

METHODS

In this prospective multicenter study, patients were recruited in Edinburgh and New York and underwent 18F-fluoride PET/MR imaging. Standardized volumes of interest were drawn in the septum and areas of late gadolinium enhancement to derive myocardial standardized uptake values (SUV) and tissue-to-background ratio (TBRMEAN) after correction for blood pool activity in the right atrium.

RESULTS

53 patients were scanned: 18 with cardiac amyloid (10 ATTR and 8 AL), 13 controls, and 22 with aortic stenosis. No differences in myocardial TBR values were observed between participants scanned in Edinburgh and New York. Mean myocardial TBRMEAN values in ATTR amyloid (1.13 ± 0.16) were higher than controls (0.84 ± 0.11, P = .0006), aortic stenosis (0.73 ± 0.12, P < .0001), and those with AL amyloid (0.96 ± 0.08, P = .01). TBRMEAN values within areas of late gadolinium enhancement provided discrimination between patients with ATTR (1.36 ± 0.23) and all other groups (e.g., AL [1.06 ± 0.07, P = .003]). A TBRMEAN threshold >1.14 in areas of LGE demonstrated 100% sensitivity (CI 72.25 to 100%) and 100% specificity (CI 67.56 to 100%) for ATTR compared to AL amyloid (AUC 1, P = .0004).

CONCLUSION

Quantitative 18F-fluoride PET/MR imaging can distinguish ATTR amyloid from other similar phenotypes and holds promise in improving the diagnosis of this condition.

Exploring the Utility of Radiomic Feature Extraction to Improve the Diagnostic Accuracy of Cardiac Sarcoidosis Using FDG PET

Background

This study aimed to explore the radiomic features from PET images to detect active cardiac sarcoidosis (CS).

Methods

Forty sarcoid patients and twenty-nine controls were scanned using FDG PET-CMR. Five feature classes were compared between the groups. From the PET images alone, two different segmentations were drawn. For segmentation A, a region of interest (ROI) was manually delineated for the patients' myocardium hot regions with standardized uptake value (SUV) higher than 2.5 and the controls' normal myocardium region. A second ROI was drawn in the entire left ventricular myocardium for both study groups, segmentation B. The conventional metrics and radiomic features were then extracted for each ROI. Mann-Whitney U-test and a logistic regression classifier were used to compare the individual features of the study groups.

Results

For segmentation A, the SUVmin had the highest area under the curve (AUC) and greatest accuracy among the conventional metrics. However, for both segmentations, the AUC and accuracy of the TBRmax were relatively high, &gt;0.85. Twenty-two (from segmentation A) and thirty-five (from segmentation B) of 75 radiomic features fulfilled the criteria: P-value &lt; 0.00061 (after Bonferroni correction), AUC &gt;0.5, and accuracy &gt;0.7. Principal Component Analysis (PCA) was conducted, with five components leading to cumulative variance higher than 90%. Ten machine learning classifiers were then tested and trained. Most of them had AUCs and accuracies ≥0.8. For segmentation A, the AUCs and accuracies of all classifiers are &gt;0.9, but k-neighbors and neural network classifiers were the highest (=1). For segmentation B, there are four classifiers with AUCs and accuracies ≥0.8. However, the gaussian process classifier indicated the highest AUC and accuracy (0.9 and 0.8, respectively).

Conclusions

Radiomic analysis of the specific PET data was not proven to be necessary for the detection of CS. However, building an automated procedure will help to accelerate the analysis and potentially lead to more reproducible findings across different scanners and imaging centers and consequently improve standardization procedures that are important for clinical trials and development of more robust diagnostic protocols.

Assessing the qualitative and quantitative impacts of simple two-class vs multiple tissue-class MR-based attenuation correction for cardiac PET/MR

BACKGROUND

Hybrid PET/MR imaging has significant potential in cardiology due to its combination of molecular PET imaging and cardiac MR. Multi-tissue-class MR-based attenuation correction (MRAC) is necessary for accurate PET quantification. Moreover, for thoracic PET imaging, respiration is known to lead to misalignments of MRAC and PET data that result in PET artifacts. These factors can be addressed by using multi-echo MR for tissue segmentation and motion-robust or motion-gated acquisitions. However, the combination of these strategies is not routinely available and can be prone to errors. In this study, we examine the qualitative and quantitative impacts of multi-class MRAC compared to a more widely available simple two-class MRAC for cardiac PET/MR.

METHODS AND RESULTS

In a cohort of patients with cardiac sarcoidosis, we acquired MRAC data using multi-echo radial gradient-echo MR imaging. Water-fat separation was used to produce attenuation maps with up to 4 tissue classes including water-based soft tissue, fat, lung, and background air. Simultaneously acquired 18F-fluorodeoxyglucose PET data were subsequently reconstructed using each attenuation map separately. PET uptake values were measured in the myocardium and compared between different PET images. The inclusion of lung and subcutaneous fat in the MRAC maps significantly affected the quantification of 18F-fluorodeoxyglucose activity in the myocardium but only moderately altered the appearance of the PET image without introduction of image artifacts.

CONCLUSION

Optimal MRAC for cardiac PET/MR applications should include segmentation of all tissues in combination with compensation for the respiratory-related motion of the heart. Simple two-class MRAC is adequate for qualitative clinical assessment.

Cardiovascular 18F-fluoride positron emission tomography-magnetic resonance imaging: A comparison study

BACKGROUND

18F-Fluoride uptake denotes calcification activity in aortic stenosis and atherosclerosis. While PET/MR has several advantages over PET/CT, attenuation correction of PET/MR data is challenging, limiting cardiovascular application. We compared PET/MR and PET/CT assessments of 18F-fluoride uptake in the aortic valve and coronary arteries.

METHODS AND RESULTS

18 patients with aortic stenosis or recent myocardial infarction underwent 18F-fluoride PET/CT followed immediately by PET/MR. Valve and coronary 18F-fluoride uptake were evaluated independently. Both standard (Dixon) and novel radial GRE) MR attenuation correction (AC) maps were validated against PET/CT with results expressed as tissue-to-background ratios (TBRs). Visually, aortic valve 18F-fluoride uptake was similar on PET/CT and PET/MR. TBRMAX values were comparable with radial GRE AC (PET/CT 1.55±0.33 vs. PET/MR 1.58 ± 0.34, P = 0.66; 95% limits of agreement - 27% to + 25%) but performed less well with Dixon AC (1.38 ± 0.44, P = 0.06; bias (-)14%; 95% limits of agreement - 25% to + 53%). In native coronaries, 18F-fluoride uptake was similar on PET/MR to PET/CT regardless of AC approach. PET/MR identified 28/29 plaques identified on PET/CT; however, stents caused artifact on PET/MR making assessment of 18F-fluoride uptake challenging.

CONCLUSION

Cardiovascular PET/MR demonstrates good visual and quantitative agreement with PET/CT. However, PET/MR is hampered by stent-related artifacts currently limiting clinical application.

Pericoronary and periaortic adipose tissue density are associated with inflammatory disease activity in Takayasu arteritis and atherosclerosis

AIMS

To examine pericoronary adipose tissue (PCAT) and periaortic adipose tissue (PAAT) density on coronary computed tomography angiography for assessing arterial inflammation in Takayasu arteritis (TAK) and atherosclerosis.

METHODS AND RESULTS

PCAT and PAAT density was measured in coronary (n = 1016) and aortic (n = 108) segments from 108 subjects [TAK + coronary artery disease (CAD), n = 36; TAK, n = 18; atherosclerotic CAD, n = 32; matched controls, n = 22]. Median PCAT and PAAT densities varied between groups (mPCAT: P < 0.0001; PAAT: P = 0.0002). PCAT density was 7.01 ± standard error of the mean (SEM) 1.78 Hounsfield Unit (HU) higher in coronary segments from TAK + CAD patients than stable CAD patients (P = 0.0002), and 8.20 ± SEM 2.04 HU higher in TAK patients without CAD than controls (P = 0.0001). mPCAT density was correlated with Indian Takayasu Clinical Activity Score (r = 0.43, P = 0.001) and C-reactive protein (r = 0.41, P < 0.0001) and was higher in active vs. inactive TAK (P = 0.002). mPCAT density above -74 HU had 100% sensitivity and 95% specificity for differentiating active TAK from controls [area under the curve = 0.99 (95% confidence interval 0.97-1)]. The association of PCAT density and coronary arterial inflammation measured by 68Ga-DOTATATE positron emission tomography (PET) equated to an increase of 2.44 ± SEM 0.77 HU in PCAT density for each unit increase in 68Ga-DOTATATE maximum tissue-to-blood ratio (P = 0.002). These findings remained in multivariable sensitivity analyses adjusted for potential confounders.

CONCLUSIONS

PCAT and PAAT density are higher in TAK than atherosclerotic CAD or controls and are associated with clinical, biochemical, and PET markers of inflammation. Owing to excellent diagnostic accuracy, PCAT density could be useful as a clinical adjunct for assessing disease activity in TAK.

Factors Associated with Longitudinal Psychological and Physiological Stress in Health Care Workers During the COVID-19 Pandemic: Observational Study Using Apple Watch Data

Background

The COVID-19 pandemic has resulted in a high degree of psychological distress among health care workers (HCWs). There is a need to characterize which HCWs are at an increased risk of developing psychological effects from the pandemic. Given the differences in the response of individuals to stress, an analysis of both the perceived and physiological consequences of stressors can provide a comprehensive evaluation of its impact.

Objective

This study aimed to determine characteristics associated with longitudinal perceived stress in HCWs and to assess whether changes in heart rate variability (HRV), a marker of autonomic nervous system function, are associated with features protective against longitudinal stress.

Methods

HCWs across 7 hospitals in New York City, NY, were prospectively followed in an ongoing observational digital study using the custom Warrior Watch Study app. Participants wore an Apple Watch for the duration of the study to measure HRV throughout the follow-up period. Surveys measuring perceived stress, resilience, emotional support, quality of life, and optimism were collected at baseline and longitudinally.

Results

A total of 361 participants (mean age 36.8, SD 10.1 years; female: n=246, 69.3%) were enrolled. Multivariate analysis found New York City’s COVID-19 case count to be associated with increased longitudinal stress (P=.008). Baseline emotional support, quality of life, and resilience were associated with decreased longitudinal stress (P<.001). A significant reduction in stress during the 4-week period after COVID-19 diagnosis was observed in the highest tertial of emotional support (P=.03) and resilience (P=.006). Participants in the highest tertial of baseline emotional support and resilience had a significantly different circadian pattern of longitudinally collected HRV compared to subjects in the low or medium tertial.

Conclusions

High resilience, emotional support, and quality of life place HCWs at reduced risk of longitudinal perceived stress and have a distinct physiological stress profile. Our findings support the use of these characteristics to identify HCWs at risk of the psychological and physiological stress effects of the pandemic.

Predictive Approaches for Acute Dialysis Requirement and Death in COVID-19

BACKGROUND AND OBJECTIVES

AKI treated with dialysis initiation is a common complication of coronavirus disease 2019 (COVID-19) among hospitalized patients. However, dialysis supplies and personnel are often limited.

DESIGN, SETTING, PARTICIPANTS, & MEASUREMENTS

Using data from adult patients hospitalized with COVID-19 from five hospitals from the Mount Sinai Health System who were admitted between March 10 and December 26, 2020, we developed and validated several models (logistic regression, Least Absolute Shrinkage and Selection Operator (LASSO), random forest, and eXtreme GradientBoosting [XGBoost; with and without imputation]) for predicting treatment with dialysis or death at various time horizons (1, 3, 5, and 7 days) after hospital admission. Patients admitted to the Mount Sinai Hospital were used for internal validation, whereas the other hospitals formed part of the external validation cohort. Features included demographics, comorbidities, and laboratory and vital signs within 12 hours of hospital admission.

RESULTS

A total of 6093 patients (2442 in training and 3651 in external validation) were included in the final cohort. Of the different modeling approaches used, XGBoost without imputation had the highest area under the receiver operating characteristic (AUROC) curve on internal validation (range of 0.93-0.98) and area under the precision-recall curve (AUPRC; range of 0.78-0.82) for all time points. XGBoost without imputation also had the highest test parameters on external validation (AUROC range of 0.85-0.87, and AUPRC range of 0.27-0.54) across all time windows. XGBoost without imputation outperformed all models with higher precision and recall (mean difference in AUROC of 0.04; mean difference in AUPRC of 0.15). Features of creatinine, BUN, and red cell distribution width were major drivers of the model's prediction.

CONCLUSIONS

An XGBoost model without imputation for prediction of a composite outcome of either death or dialysis in patients positive for COVID-19 had the best performance, as compared with standard and other machine learning models.

PODCAST

This article contains a podcast at https://www.asn-online.org/media/podcast/CJASN/2021_07_09_CJN17311120.mp3.