Citations

Abstract Background Right ventricular (RV) systolic function is an independent predictor of outcome, but a robust automated method of quantitation is lacking. Purpose To develop and validate a deep learning model for quantifying RV systolic function from 2D echo clips and assess its prognostic value in various loading states and left-heart disease. Methods A deep learning convolutional neural network (CNN) was developed to quantitate RV systolic function using all available apical 4 chamber (A4C) views from 13,039 patients (61,187 A4C video clips) in a US institution. Videos were converted to frames of 256x256 grayscale pixels, with graphics, including the ECG, removed. The model used the S3D architecture, a 3D-CNN with many of the spatial and temporal convolutions separated. RV free wall longitudinal strain (RVFWLS) was used to define RV function: normal (≤-25%); mild (-20% to –24%); moderate (-15% to –19%) and severe RV dysfunction (0% to –14%). The AI-estimate was the average of multiple 32-frame segments for each clip identified as an A4C, including both standard and RV-focused variations. The model was validated in four distinct clinical cohorts from independent data sources: 1) 1365 patients referred to an external diagnostic echocardiography lab in a different US state; 2) 569 patients recruited prospectively with known or suspected pulmonary hypertension (PH); 3) 416 patients with moderate or greater tricuspid valve regurgitation (TR); 4) 872 patients imaged at an Eastern European University with left-sided cardiac diseases (valvular disease, cardiomyopathies, etc.), athletes, or healthy volunteers. Results Model output was strongly correlated with the presence of RV dysfunction by manual measurement [AUC 0.907 (95% CI: 0.888-0.926)] in the internal validation set. Results were similar in an external validation US cohort [N=1365; AUC 0.919 (95% CI: 0.899-0.939)]. When the model was applied to 569 patients with known or suspected PH, presence and severity of the AI measures of RV dysfunction were strongly associated with functional class, 6 min walk distance, NT-proBNP levels and right heart failure (p<0.001). The presence and severity of RV dysfunction (as categorized by the AI model) was strongly associated with all-cause mortality in patients with PH (Figure 1A) and in patients with significant TR (Figure 1B). The European cohort was a geographically and institutionally distinct cohort with a low prevalence of >mild TR (10.8%) or PH (6.4%). In this cohort the AI model predicted a reduced RVEF (< 45%) with an AUC of 0.864 (95% CI: 0.831-0.898) and associated with both all-cause mortality (Figure 2A) and a composite endpoint of heart failure hospitalization or death (Figure 2B). Conclusions A fully automated, AI-based assessment of RV systolic function, using 2D apical 4 chamber echo views alone, is highly feasible and can accurately categorize disease severity and predict outcome across a wide spectrum of cardiovascular conditions.Figure 1 Figure 2

Objective: To evaluate hearing outcomes after treatment with stereotactic radiosurgery (SRS) for sporadic vestibular schwannoma (VS) among patients with baseline 100% word recognition scores (WRS). Study design: Historical cohort. Setting: Tertiary academic medical center. Patients: Patients with 100% WRS on audiometric testing at initial treatment with SRS. Interventions: Gamma Knife SRS. Main outcome measures: Rate of maintaining serviceable hearing (SH), defined as American Academy of Otolaryngology-Head and Neck Surgery hearing class A or B. Results: In total, 125 patients had 100% WRS at SRS for sporadic VS. Median age at SRS was 56 years (IQR: 51 to 62) and median pure tone average (PTA) at SRS was 20 dB HL (IQR: 13 to 31). Sixty-eight patients progressed to non-SH at a median of 3.4 years after SRS (IQR: 1.0 to 8.6); the median duration of follow-up for the 57 patients who maintained SH was 5.1 years (IQR: 2.8 to 9.6). Rates of maintaining SH (95% CI, number still at risk) at 1, 3, 5, 7, and 10 years after SRS were 88% (82 to 94, 101), 75% (68 to 84, 81), 61% (53 to 71, 55), 58% (49 to 68, 43), and 44% (35 to 56, 26), respectively. Each 10-dB HL increase in PTA at SRS was associated with a 52% increased risk of progression to non-SH (hazard ratio: 1.52, 95% CI: 1.24-1.87, P <0.001). Conclusions: Despite having 100% pretreatment WRS, most patients with sporadic VS treated with SRS experienced clinically significant hearing decline over time. Even among this cohort of patients with excellent baseline hearing, at 10 years after treatment, only 44% of patients maintained SH.

Objective Age of symptom onset is highly variable in familial frontotemporal lobar degeneration (f‐FTLD). Accurate prediction of onset would inform clinical management and trial enrollment. Prior studies indicate that individualized maps of brain atrophy can predict conversion to dementia in f‐FTLD. We used a Bayesian linear mixed‐effect (BLME) prediction method for identifying accelerated brain volume loss to predict conversion to dementia. Methods Participants included 234 asymptomatic or prodromal carriers of C9orf72 , GRN , or MAPT mutations (including 21 dementia converters) with ≥3 longitudinal magnetic resonance imaging (MRI) T1‐weighted scans. The BLME models established individual voxel‐wise gray matter trajectories using the first 2 scans. Person‐specific clusters of accelerated volume loss were estimated in subsequent scans and tested as predictors of dementia conversion compared with other approaches in time‐varying Cox proportional hazard models covarying for age. Receiver‐operating characteristic (ROC) curves estimated utility of cluster volume in discriminating which participants converted to dementia within 24 months. Results The BLME cluster volume predicted conversion to dementia in f‐FTLD mutation carriers overall and separately in C9orf72 , GRN , and MAPT , with comparable hazard ratios observed for atrophy W‐maps and regional volumes. Within a 24‐month timeframe, BLME cluster volume discriminated dementia converters from non‐converters with larger areas under the curve (AUCs) than other approaches. Interpretation Bayesian‐modeled individualized atrophy scores predict dementia progression among asymptomatic f‐FTLD mutation carriers and may have increased utility compared with other structural imaging methods when studying individuals over shorter timeframes that align with clinical trial design. ANN NEUROL 20269999:n/a–n/a

Introduction: Plasma biomarkers of Alzheimer’s disease (AD), including amyloid-β (Aβ42/Aβ40), phosphorylated tau (p-tau181), neurofilament light (NfL), and glial fibrillary acidic protein (GFAP) have shown strong reliability in predicting dementia risk. White matter hyperintensities (WMH) represent a major vascular contributor to dementia. We investigated how plasma biomarkers of AD interact with WMH on MRI to influence incident dementia risk. Methods: We included dementia-free participants from the Atherosclerosis Risk in Communities (ARIC) Study who underwent MRI and plasma biomarker assessment at Visit 5 (2011–2013). WMH was quantified and considered continuously (log-transformed and standardized), and in tertiles. Plasma biomarkers were log-base-2-transformed. Cox proportional hazards models were used to estimate dementia risk over 10 years associated with plasma biomarkers and WMH, independently and multiplicatively. Models were adjusted for demographic variables (age, sex, race, and education), vascular risk factors (hypertension, diabetes, smoking status, body mass index, and estimated glomerular filtration rate), and APOE ε4 status. Results: Among 1,538 participants (aged 67- 90 years, 29% Black), 30% ( N = 462) developed dementia over the 10-year follow-up period. For every doubling of plasma Aβ42/Aβ40 , indicating lower amyloid burden, dementia risk decreased, while doubling of p-tau 181 , NfL, and GFAP were each associated with increased dementia risk, in separate models; each of these associations was independent of WMH (Table 1). Additionally, dementia risk increased per SD increase in WMH, independent of each plasma biomarker. Although no multiplicative interaction terms reached statistical significance (p<0.05), there was suggestive evidence that Aβ42/Aβ40 and GFAP have smaller effect sizes on dementia risk in subgroups at the highest WMH tertile (p-interactions<0.10; see Table 2). Discussion: Plasma biomarkers and WMH independently predict dementia risk over 10 years, though there remained limited evidence of multiplicative effects. Plasma biomarkers may be less important for dementia risk in individuals who also have extensive WMH. These findings emphasize the importance of incorporating both AD-specific plasma markers as well as indicators of cerebral small vessel disease, such as WMH, into dementia risk profiling.

Background: Cardiac CT acquired during the acute stroke imaging protocol is an emerging diagnostic modality to detect cardiac thrombi. We aimed to determine its diagnostic yield in a large, international, multicenter study. Methods: We performed an individual patient data meta-analysis by pooling prospective and retrospective observational data of patients with acute ischemic stroke (AIS) who underwent cardiac CT (ECG-gated or non-ECG-gated) as part of the acute imaging protocol between May 2018 and June 2024. We excluded patients with a TIA or stroke mimic. The primary outcome was the proportion of patients with a cardiac thrombus detected on cardiac CT. Secondary outcomes were additional scan time and radiation dose, comparison of yield with echocardiography, and 90-day mortality and recurrent ischemic stroke rates. Results: We included 3940 patients from 4 hospitals (median age 74 [IQR 63-82], 58.0% male, median NIHSS 6 [IQR 3-12], median time onset to hospital arrival 100 minutes [IQR 62-185]). Atrial fibrillation at baseline was present in 23.1% patients, and 29.6% had a large vessel occlusion. Cardiac CT was non-diagnostic in 17 (0.4%) patients. In total, cardiac CT detected 254 cardiac thrombi in 243/3923 (6.2%) patients. Thrombi were located in the left atrial appendage (n=193), left atrium (n=19), and left ventricle (n=42). Of 95/243 (39.1%) patients with a cardiac thrombus that underwent echocardiography (median interval cardiac CT-echocardiography: 2 days), transthoracic echocardiography detected a thrombus in 18/92 (19.6%) patients and transesophageal echocardiography in 2/5 (40.0%) patients. The diagnostic yield in patients that underwent both modalities was higher for cardiac CT (OR 7.5 [CI 4.0-15.3], p<0.001). Median additional scan time was 6 minutes (IQR 5-7) for ECG-gated and 17 seconds (IQR 15-63) for non-ECG-gated cardiac CT. Median additional radiation dose was 5.4mSv (IQR 3.0-7.6). Patients with a cardiac thrombus on cardiac CT had higher 90-day mortality than those without a thrombus (33.3% vs 14.6%, p<0.001). We found no difference in 90-day recurrent stroke rate (5.4% vs 4.4%, p=0.50). Conclusion: Routine cardiac CT performed during the acute stroke imaging protocol detects a cardiac thrombus in 1 out of 16 patients with AIS, with a 7-times higher yield than routine echocardiography. Mortality rate was doubled in patients with a cardiac thrombus, but 90-day stroke recurrence risk did not differ.

Objectives: Cerebral hemodynamic impairment in patients with asymptomatic high-grade internal carotid artery (ICA) stenosis confers a risk of stroke and cognitive decline, but correlations between degree of stenosis and hemodynamic impairment are derived from small case series only. Using baseline data from 282 participants in the Carotid Revascularization and Medical Management for Asymptomatic Carotid Stenosis – Hemodynamics (CREST-H) study, we hypothesized that degree of stenosis, measured by peak systolic velocity (PSV) and end diastolic velocity (EDV) in the ipsilateral ICA, would correlate with time-to-peak (TTP) delay using MR (n=275) or CT (n=7) perfusion scans, adjusting for demographic and cardiovascular risk variables. Methods: Using a 7-10MHz Doppler probe, ICA PSV and EDV were assessed. Doppler results were adjudicated centrally. Dynamic contrast perfusion imaging was standardized across 61 CREST-H sites, using sequential T2*-weighted time-sequence scanning (e.g. TR=1500, TE=25ms, flip angle=60deg, phase=128, slice thickness=5mm for GE 3T magnet). CTP was acquired with standard clinical protocols. TTP delay was calculated in the ipsilateral versus contralateral MCA/ACA territories using OLEA software. Linear regression used TTP delay as the outcome variable and PSV and EDV as predictor variables, adjusting for covariates.The 282 with complete data for all variables were included in the analysis. Results: Among 392 CREST-H participants, 63% were men, 7% Black, 3% Hispanic, 33% with less than college education. Tables 1a/b show the regression results for PSV and EDV as main predictor variables. PSV and EDV on the index side significantly correlated with TTP delay (p=0.004, p=0.002 respectively). Neither PSV nor EDV on the non-index side correlated with TTP delay (p=0.49, p=0.82). Compared with moderate exercise, vigorous exercise correlated with lower TTP delay in both models (p=0.05. p=0.04). Compared with college graduates, those with HS education had longer TTP delays in both models (p=0.03. p=0.02). Conclusions: In this large cohort of patients with high-grade asymptomatic carotid stenosis, higher flow velocities correlated with greater TTP delays, suggesting an increasing hemodynamic impairment with greater degree of stenosis above 70%. Our findings have implications for management of asymptomatic carotid stenosis which are being tested in the CREST-H and CREST-2 studies. Exercise and education may independently impact cerebral hemodynamics.

Background: The rate of intracardiac thrombus detection in the hyperacute phase of stroke varies from 5% to 12% between studies. Our aim was to compare the thrombus detection rate between ECG-gated and non-ECG-gated cardiac CT protocols in acute stroke patients, using data from a large, international, multicentre registry. Methods: We conducted an individual patient data meta-analysis by pooling prospective and retrospective observational data from stroke patients assessed at four comprehensive stroke centres who underwent either ECG-gated or non-ECG-gated cardiac CT (extended CT angiography or two-phase non-ECG-gated cardiac CT) as part of the acute stroke imaging protocol between May 2018 and June 2024. The primary outcome was the proportion of patients with a cardiac thrombus in gated vs non gated techniques. Results: Of 3,940 patients, 1,474 (37.4%) underwent ECG-gated and 2,466 (62.6%) non-ECG-gated cardiac CT. Patients in the ECG-gated group were younger (median age 72 [IQR 62–81] vs. 74 [IQR 64–82] years, p = 0.000) and had a lower NIHSS at baseline (median 5 [IQR 2–12] vs. 6 [IQR 3–13], p < 0.00). There was a higher rate of thrombus detection with ECG-gated CT compared to non-gated CT (7.5% vs. 5.5%, p = 0.011. The association remained significant after adjusting for age, sex, baseline NIHSS score, and baseline atrial fibrillation (adjusted OR 1.87, 95% CI 1.42–2.47, p = 0.000). ECG-gated CT was associated with lower radiation dose (3.6 vs. 6.3 mSv, p = 0.000) but longer acquisition time (5.8 vs. 0.28 min, p = 0.000). Conclusion: ECG-gated cardiac CT has a higher detection rate for intracardiac thrombus detection compared to non-ECG-gated CT in patients with acute ischemic stroke.

BACKGROUND: Cerebral hemodynamic impairment in patients with asymptomatic high-grade (>70%) internal carotid artery (ICA) stenosis is associated with risk of stroke and cognitive decline, but correlations between the degree of stenosis and hemodynamic impairment are derived from small case series. Using baseline data from 242 participants in the CREST-H study (Carotid Revascularization and Medical Management for Asymptomatic Carotid Stenosis–Hemodynamics), we hypothesized that the degree of stenosis in the ICA would correlate with time-to-peak (TTP) delay in ipsilateral cerebral blood flow using MR/CT perfusion scans, adjusting for demographic and cardiovascular risk variables. METHODS: From all participants, ICAs were insonated using a 7 to 10 MHz probe for peak systolic velocity (PSV) and end diastolic velocity. Dynamic contrast perfusion imaging was standardized across 61 CREST-H sites, using standard sequential T2*-weighted perfusion imaging. CT perfusion used standard clinical protocols. TTP delay was calculated in the ipsilateral versus contralateral hemispheres. In cross-sectional analysis, linear regression was used to model TTP delay as the outcome variable and PSV, end diastolic velocity, and PSV ICA/common carotid artery ratio on the index and nonindex side as primary predictor variables in 6 separate models, adjusting for covariates, followed by automated backward elimination model reduction. RESULTS: Among 392 CREST-H participants, the 242 with complete data for all variables were included in the regression analysis (age, 70±7.6; 62% M). End diastolic velocity on the index side correlated with TTP delay (β=0.003, P =0.005). PSV had a similar correlation but did not reach significance (β=0.001, P =0.099), nor did ICA/common carotid artery ratio (β=0.003, P =0.126). Nonindex side PSV, end diastolic velocity, and ICA/common carotid artery ratio showed no correlation with TTP delay ( P =0.268, P =0.495, P =0.380, respectively). Circle of Willis completeness did not correlate with TTP. CONCLUSIONS: In this large cohort of patients with high-grade asymptomatic carotid stenosis, higher end-diastolic flow velocities correlated with greater TTP delays, supporting a definition of hemodynamically significant stenosis. Our findings have implications for the management of asymptomatic carotid stenosis, which will be tested in the CREST-H study. REGISTRATION: URL: https://www.clinicaltrials.gov ; Unique identifier: NCT03121209.

ABSTRACT Background Although kidney biopsy remains the gold standard for diagnosing immune checkpoint inhibitors–associated acute interstitial nephritis (ICI-AIN), several promising urinary biomarkers have emerged as potential non-invasive diagnostics. Notably, recent studies have highlighted tumor necrosis factor alpha (TNF-α), CXC motif chemokine ligand 9 (CXCL-9) and others as potential biomarkers to help distinguish ICI-AIN from other causes of acute kidney injury (AKI). However, a direct systematic comparison between these two biomarkers in a cohort of patients undergoing immunotherapy has not been performed. Methods We prospectively enrolled a cohort of ICI-treated cancer patients who developed AKI (including both ICI-AIN and ICI-Non-AIN AKI), as well as those without AKI, referred to as the internal validation cohort. Urinary CXCL-9 and TNF-α were measured across all groups using the human Quantikine ELISA Kit. We externally validated our findings in a cohort of biopsy proven ICI-AIN patients from Spain, referred to as the external validation cohort. Results Urinary CXCL-9 levels were significantly higher in ICI-AIN patients from Mayo (n = 40) compared with those with ICI-Non-AIN AKI (n = 36) and ICI-treated patients without AKI (n = 16) (both P ≤ .0001); these findings were confirmed in the validation cohort from Spain (n = 19) with P-values of .0008 and &lt;.0001, respectively. The area under the curve (AUC) for urinary CXCL-9 distinguishing ICI-AIN biopsy proven (combined from both cohorts, n = 45) from ICI-Non-AIN AKI was 0.83 (95% confidence interval 0.74–0.92), while for urinary TNF-α the AUC was 0.63 (0.51–0.75). Conclusions Increased urinary CXCL-9 was superior to TNF-α as a biomarker for the non-invasive diagnosis of ICI-AIN, even after prior corticosteroid exposure, highlighting its promise as clinical diagnostic tool.