Citations

AbstractBackgroundLarge neuroimaging studies like ADNI‐4 now include T1‐weighted, T2‐weighted and FLAIR images at isotropic1mm resolution for investigation of both neurodegenerative and cerebrovascular disease. These three input types allow segmentation of GM, WM, DeepGM, WMH, and CSF compartments. Additionally, this supports automated assessment of enlarged perivascular spaces (PVS). Our objective was to propose a unified framework for segmentation of intracranial classes and PVS while reducing human effort.MethodsFLAIR and T2‐weighted images are aligned to the T1‐weighted image.. Standard SPM12 unified segmentation using all three input images establishes deformations between subject and template space and bias field corrections. Extended tissue priors including deep GM are deformed into subject space. Adding mean intensities from high probability deep GM voxels as seeds and using the extended prior set, segmentation is re‐started, refining model parameters, deformations and bias‐field estimates. The procedure repeats, adding a WMH channel, resulting in a segmentation including GM, WM, CSF, deep GM, WMH and nuisance classes. Frangi filters are applied to the T1‐ and T2‐ weighted images. Filter response images are smoothed, averaged, masked and inserted in place of the CSF prior. Probability maps of vessel‐like shapes with CSF‐like contrast are estimated to identify PVS. Visual grading was also performed for comparison.ResultsThe method was applied to 300 participants. Visual quality control found two instances of poor WMH over‐segmentation. Perivascular space maps were all visually reasonable. Standard segmentation classes were visibly superior to single T1‐based segmentation. Figures 1,2 present examples. Pearson correlations of PVS fractional volume with multiple human raters’ PVS counts in CSO/CR was over 0.7 in a subset of 125 participants (Figure 3).ConclusionModern dementia protocols include high‐resolution T1‐weighted, T2‐weighted, and FLAIR images that can be used together as inputs to extend probabilistic segmentation adding deep grey and WMH classes with automated PVS assessment.

AbstractBackgroundAdults with Down Syndrome (DS) develop Alzheimer disease (AD)‐like pathology and dementia as they age. This is attributed to triplication of the amyloid precursor protein gene. Recent work demonstrated that amyloid is elevated in DS and accumulates in a similar topography to autosomal dominant AD (ADAD). Tau accumulation is the second hallmark of AD and may be measured using tau positron emission tomography (PET) imaging. The pattern of tau accumulation remains unknown in DS.MethodIn a cross‐sectional analysis we evaluated amyloid and tau positron emission tomography (PET) in 124 participants with DS (109 cognitively stable, 8 symptomatic, 7 no consensus), 29 ADAD mutation carriers (26 asymptomatic, 3 symptomatic), and 201 cognitively normal non‐mutation carriers (CN). Regional cortical standard uptake value ratios (SUVR) for tau imaging were obtained and compared by cognitive status. Similarly, regional cortical amyloid SUVR were obtained and converted to centiloids. We then compared the magnitude of tau burden as well as spatial spread of tau across groups relative to amyloid burden.ResultCompared to CN, tau PET binding was elevated in participants with (µDS = 2.12 ± 0.22 SUVR, µADAD = 2.52 ± 0.37 SUVR) and without (µDS = 1.24 ± 0.15 SUVR, µADAD = 1.16 ± 0.19 SUVR) cognitive changes. For both groups, tau increased with increasing amyloid (centiloids). In some regions, DS had greater tau burden for a given amyloid level (centiloids) (Figure 1). There was increased spatial distribution of tau throughout the brain for participants with DS compared to CN. In contrast, tau burden was more focal in participants with ADAD with tau primarily elevated in posterior regions (Figure 2).ConclusionWhile tau burden was higher with symptomatic disease for both groups, significant differences existed in the overall amount and spatial spread. Tau spread was greatest in participants with DS and more diffuse compared to ADAD. These results indicate a substantial tau burden is present in DS, and suggest not only anti‐amyloid but also tau‐reducing agents should be considered for treating cognitive decline in adults with DS.

AbstractBackgroundLewy body dementia (LBD) is one of the most common causes of dementia and is characterized by the deposition of extracellular amyloid plaques and intracellular accumulation of alpha‐synuclein in the form of Lewy bodies and neurites. Many of these individuals also have concomitant neuropathological changes such as vascular disease, tangles, and TDP‐43. Like in Alzheimer’s disease, APOE4 is the most significant genetic risk factor for the development of LBD. We sought to survey the landscape of transcriptional changes in the cingulate cortex from a large cohort of LBD cases (>400) compared to normal controls of both sexes.MethodWe performed bulk tissue RNAseq and aligned reads to a human reference genome that is informed on the sex chromosome complement of the sample. Differentially expressed genes were determined after genome‐wide correction.ResultWe found a core set of transcriptional changes that were shared between both sexes in pathways related to neuronal functions and inflammation. However, we also identified numerous changes in genetic female samples that were not significantly altered in genetic males, or that were significantly altered but in the opposite direction. Similar male‐specific changes were also found. These alterations were present for genes on both sex chromosomes as well as autosomes. Though cohort sizes were smaller, we also compared this dataset to Alzheimer’s disease and found shared but unique transcriptional alterations.ConclusionOverall, genetic males and females exhibit sex‐shared and sex‐specific gene expression patterns in dementia with Lewy bodies. We are currently assessing splice alterations, running causal gene network analysis, and performing histological validation.

AbstractBackgroundLarge neuroimaging studies like ADNI‐4 now include T1‐weighted, T2‐weighted and FLAIR images at isotropic1mm resolution for investigation of both neurodegenerative and cerebrovascular disease. These three input types allow segmentation of GM, WM, DeepGM, WMH, and CSF compartments. Additionally, this supports automated assessment of enlarged perivascular spaces (PVS). Our objective was to propose a unified framework for segmentation of intracranial classes and PVS while reducing human effort.MethodFLAIR and T2‐weighted images are aligned to the T1‐weighted image.. Standard SPM12 unified segmentation using all three input images establishes deformations between subject and template space and bias field corrections. Extended tissue priors including deep GM are deformed into subject space. Adding mean intensities from high probability deep GM voxels as seeds and using the extended prior set, segmentation is re‐started, refining model parameters, deformations and bias‐field estimates. The procedure repeats, adding a WMH channel, resulting in a segmentation including GM, WM, CSF, deep GM, WMH and nuisance classes. Frangi filters are applied to the T1‐ and T2‐ weighted images. Filter response images are smoothed, averaged, masked and inserted in place of the CSF prior. Probability maps of vessel‐like shapes with CSF‐like contrast are estimated to identify PVS. Visual grading was also performed for comparison.ResultThe method was applied to 300 participants. Visual quality control found two instances of poor WMH over‐segmentation. Perivascular space maps were all visually reasonable. Standard segmentation classes were visibly superior to single T1‐based segmentation. Figures 1,2 present examples. Pearson correlations of PVS fractional volume with multiple human raters’ PVS counts in CSO/CR was over 0.7 in a subset of 125 participants (Figure 3).ConclusionModern dementia protocols include high‐resolution T1‐weighted, T2‐weighted, and FLAIR images that can be used together as inputs to extend probabilistic segmentation adding deep grey and WMH classes with automated PVS assessment.

Abstract Background and Aims Severe aortic stenosis (AS) is the guideline-based indication for aortic valve replacement (AVR), which has markedly increased with transcatheter approaches, suggesting possible increasing AS incidence. However, reported secular trends of AS incidence remain contradictory and lack quantitative Doppler echocardiographic ascertainment. Methods All adults residents in Olmsted County (MN, USA) diagnosed over 20 years (1997–2016) with incident severe AS (first diagnosis) based on quantitatively defined measures (aortic valve area ≤ 1 cm2, aortic valve area index ≤ 0.6 cm2/m2, mean gradient ≥ 40 mmHg, peak velocity ≥ 4 m/s, Doppler velocity index ≤ 0.25) were counted to define trends in incidence, presentation, treatment, and outcome. Results Incident severe AS was diagnosed in 1069 community residents. The incidence rate was 52.5 [49.4–55.8] per 100 000 patient-year, slightly higher in males vs. females and was almost unchanged after age and sex adjustment for the US population 53.8 [50.6–57.0] per 100 000 residents/year. Over 20 years, severe AS incidence remained stable (P = .2) but absolute burden of incident cases markedly increased (P = .0004) due to population growth. Incidence trend differed by sex, stable in men (incidence rate ratio 0.99, P = .7) but declining in women (incidence rate ratio 0.93, P = .02). Over the study, AS clinical characteristics remained remarkably stable and AVR performance grew and was more prompt (from 1.3 [0.1–3.3] years in 1997–2000 to 0.5 [0.2–2.1] years in 2013–16, P = .001) but undertreatment remained prominent (>40%). Early AVR was associated with survival benefit (adjusted hazard ratio 0.55 [0.42–0.71], P < .0001). Despite these improvements, overall mortality (3-month 8% and 3-year 36%), was swift, considerable and unabated (all P ≥ .4) throughout the study. Conclusions Over 20 years, the population incidence of severe AS remained stable with increased absolute case burden related to population growth. Despite stable severe AS presentation, AVR performance grew notably, but while declining, undertreatment remained substantial and disease lethality did not yet decline. These population-based findings have important implications for improving AS management pathways.

Background: Atrial septal defects (ASD) and ventricular septal defects (VSD) are among the most common forms of congenital heart disease (CHD). Early detection and intervention for these defects are crucial to prevent complications. Given advancements in AI technology, the development of AI-enabled ECG models for detecting these defects is of significant interest. Aims: This study aimed to create AI models using pediatric ECG data for the detection of ASD, VSD, and a combined model detecting either ASD or VSD. Methods: We analyzed a dataset of 7,795 pediatric patients from Mayo Clinic who had undergone echocardiography between 2002 and 2022, with corresponding 10-second, 12-lead surface ECGs taken within 14 days of echocardiography. This dataset included 259 patients with ASDs and 139 patients with VSDs (single or multiple). The control group was comprised of 7,397 patients without significant or complex CHD. For cases and controls, we included patients with all forms of electrocardiographic abnormalities and those with valvular lesions who had less than moderate stenosis or regurgitation. The dataset was divided into training, validation, and testing subsets to develop convolutional neural network (CNN) models for ASD, VSD, and either ASD or VSD detection. Results: Our ASD and VSD detection models achieved test AUCs of 0.82 (95% CI 0.76, 0.88) and 0.80 (95% CI 0.70, 0.90), respectively. Meanwhile, a combined model that detected either ASD or VSD demonstrated a test AUC of 0.81 (95% CI, 0.76, 0.86), sensitivity of 0.73, specificity of 0.76, PPV of 0.14, and NPV of 0.98 (Figure 1). Conclusions: AI-enabled ECG models demonstrate promising performance in detecting ASD and VSD in pediatric patients. A combined model that detects either ASD or VSD also shows robust results. Future research should focus on multi-center validation of these models, refining their diagnostic capacities, and investigating their potential for early intervention in CHDs.

Introduction: Left and right ventricular systolic dysfunction (LVSD and RVSD respectively) contribute to considerable pediatric morbidity. Both LVSD and RVSD can occur in children before symptom onset, emphasizing the need for early detection. Existing deep-learning algorithms can identify LVSD in adults using 12-lead ECG; however, their efficacy in children is uncertain. Aims: We aimed to create a deep-learning model for LVSD and RVSD detection in pediatric patients using an AI-enabled ECG and compare novel pediatric models to a previously validated adult LVSD model. Methods: We identified 10,214 pediatric patients at the Mayo Clinic with a transthoracic echocardiogram (TTE) and a 10-second, 12-lead surface ECG within 14 days of TTE, performed from 2002 -2022. Our cohort comprised the first TTE-ECG pair from each patient, totaling 138 with EF ≤ 35%, 312 with EF ≤ 50%, 482 with RVSD, and 9,567 without any LVSD or RVSD. The cohort was split into training, validation, and testing datasets to develop neural networks for systolic dysfunction detection. Results: The cohort was 52.3% male, with an average age of 10.3 years. Novel models generated in the pediatric cohort demonstrated excellent performance in systolic dysfunction detection (Figure 1). Pediatric models for EF ≤ 35% and EF ≤ 50% achieved test AUCs of 0.92 (95% CI 0.89, 0.95) and 0.89 (95% CI 0.85, 0.94) respectively. The RVSD model reached a test AUC of 0.91 (95% CI 0.88, 0.94). A combined model for EF ≤ 35% or RVSD exhibited a test AUC of 0.90 (95% CI 0.87, 0.93), sensitivity 0.88, specificity 0.82, PPV 0.20, and NPV 0.95. A combined model for EF ≤ 50% or RVSD exhibited a test AUC of 0.88 (95% CI 0.85, 0.91). The previously validated adult data-derived model achieved an AUC of 0.86 (95% CI 0.83, 0.89) for EF ≤ 50% in children. Conclusions: AI-enabled ECGs demonstrate efficient detection of both LVSD and RVSD in pediatric patients. Future multi-site validation is needed to further establish the reliability of these models.

Introduction: Biologic sex and hormonal concentrations shape ECG parameters due to sex hormone effects on cardiac function. AI-enabled ECGs have previously been shown to accurately predict patient sex in adults and correlate with sex hormone levels. Aims: We aimed to test the ability of AI-enabled ECGs to predict pediatric patient sex and explore the influence of pubertal development. Methods: Two AI-enabled ECG models were created using a convolutional neural network trained on pediatric 10-second, 12-lead surface ECGs. The first model was trained de novo using pediatric data. The second model used transfer learning from a previously validated adult data-derived algorithm. We analyzed the first ECG from 89,063 pediatric patients (aged <19 years) recorded from 1988-2022, and divided the cohort into training, validation, and testing datasets. Subgroup analysis was performed on prepubertal (0-7 years), pubertal (8-14 years), and post-pubertal (15-18 years) patients. Results: The cohort consisted of 46.6% males, with 21,434 prepubertal, 26,300 pubertal, and 41,329 post-pubertal children. No significant performance difference was observed between the two models. Both models demonstrated 81% accuracy and an overall AUC of 0.91 in the testing cohorts, with high discriminatory ability in post-pubertal teenagers (AUC = 0.98). (Figure 1) Conclusions: AI-ECG predicts pediatric patient sex with varying performance across pubertal stages. The highest discriminatory ability is observed post-puberty, decreasing in pubertal and prepubertal children. The results suggest hormonal and physiological changes during puberty may influence cardiac electrophysiology measurably. More research is required to interpret CNN-predicted sex discordance with actual sex and its implications for future cardiovascular risk.

Introduction: Hospitalization for COVID-19 is often associated with myocardial injury (elevated cardiac troponin [cTn]). In the initial pandemic waves, patients with myocardial injury had increased likelihood of cardiovascular complications and inpatient mortality; however, the differences in impact of distinct subtypes of myocardial injury on mortality is unclear. In addition, little is known about how these outcomes compare with the subsequent, milder, Omicron wave. Methods: Utilizing Mayo Clinic enterprise patients we developed algorithms based on diagnostic/procedural codes, cTn levels, and temporal relationships to positive COVID-19 PCR tests, to determine 5 common subtypes of myocardial injury: stress cardiomyopathy, myocarditis, pericarditis, pulmonary embolism, and myocardial infarction. We examined hospitalized adults with COVID-19; those with elevated cTn designated as cases and others as controls. We fitted Cox models using the following covariates: spline transformed age, sex, race, ethnicity, index date and clinical comorbidities. Patients were analyzed based on admission dates coinciding with either the Alpha-Delta or Omicron waves. Results: Across both wave cohorts, hospitalized patients with acute COVID-19 complicated by myocardial injury had an increased risk of mortality compared to those without ( Figure ). Patients with one of the 5 subtypes examined tended to have higher risk than those in the “other myocardial injury” category. Admission during the Omicron wave, compared to the Alpha-Delta waves, tended to have higher mortality risk in all subtypes except myocarditis. Conclusions: Our algorithms successfully identified subgroups of myocardial injury in patients hospitalized with acute COVID-19 leading to stratification of mortality risk. Surprisingly, among hospitalized patients during the Omicron wave, myocardial injury of most subtypes was tended to have increased risk of mortality compared to the Alpha-Delta waves.

Background: Heart failure (HF) is a complex trait influenced by environmental and genetic factors. Genome-wide association studies (GWAS) have found common genetic variants associated with HF risk, but the extent that common and rare variants explain risk and implicate similar genes and pathways in HF remains less understood. Methods: For common variants, we meta-analyze existing GWAS of all-cause HF with data from the Million Veteran Program, including 207,346 individuals with HF and 2,151,210 without in diverse populations. We examine rare coding variants in the Penn Medicine BioBank (PMBB) using gene-collapsing association studies and burden heritability regression. We test for an interaction between a polygenic risk score (PRS) based on our GWAS meta-analysis and truncating variants in TTN (TTNtv). Results: GWAS meta-analysis identified 176 risk loci (146 previously unreported), implicating known risk factors for HF and genes involved in cardiac development. New signals at GIPR/ GLP1R highlight potential contributions of dysregulated glucose homeostasis and metabolism in HF. Gene collapsing studies identified associations between HF and rare loss-of-function variants in TTN (OR 2.94, 95% CI 2.43-3.56) and MYBPC3 (OR 7.71, 95% CI 3.97-14.97) at exome-wide significance. HF heritability mediated through common variants (4.5%) was significantly higher than for rare variants (1.8%), but rare variant heritability was further enriched in cardiomyopathy genes compared to common variants (27-fold vs. 3.4-fold). We found a significant interaction between a new multi-ancestry HF-associated PRS (AUC 0.72, 95% CI 0.72-0.73) and TTNtv carrier status (interaction OR 1.32, 95% CI 1.06-1.64). Conclusions: Common and rare variants contribute to HF heritability. Common variant HF heritability is highly polygenic, while rare variant burden heritability appears more concentrated in core cardiomyopathy genes. HF risk conferred by TTNtv carrier status is further increased by an individual’s polygenic contribution. Our results support the continued inclusion of cardiomyopathy genes in rare-variant genetic testing panels for HF but suggest a meaningful polygenic component to HF exists that is not currently captured by clinical genetic testing.