Daily Cardiology Research Analysis
Three impactful cardiology studies stand out today: a multicenter prospective study shows a wearable, AI-enabled system can estimate pulmonary capillary wedge pressure with accuracy approaching invasive sensors; a national registry analysis links earlier cardiac resynchronization therapy after medical stabilization to improved outcomes; and a deep-learning model on chest radiographs accurately detects pulmonary hypertension and its CHD-associated subtype with right heart catheterization validati
Summary
Three impactful cardiology studies stand out today: a multicenter prospective study shows a wearable, AI-enabled system can estimate pulmonary capillary wedge pressure with accuracy approaching invasive sensors; a national registry analysis links earlier cardiac resynchronization therapy after medical stabilization to improved outcomes; and a deep-learning model on chest radiographs accurately detects pulmonary hypertension and its CHD-associated subtype with right heart catheterization validation.
Research Themes
- Noninvasive hemodynamic assessment using wearable sensors and AI
- Optimization of device therapy timing in heart failure
- AI-enabled imaging for early detection of pulmonary hypertension
Selected Articles
1. Noninvasive Pulmonary Capillary Wedge Pressure Estimation in Heart Failure Patients With the Use of Wearable Sensing and AI.
In a multicenter prospective study of 310 HFrEF patients, a wearable sensor combining ECG, seismocardiography, and photoplethysmography with machine learning estimated PCWP with an error of 1.04 ± 5.57 mmHg versus right heart catheterization. Performance was consistent across demographics, suggesting a scalable, noninvasive alternative to implantable hemodynamic monitoring.
Impact: This study demonstrates clinically relevant accuracy for noninvasive, AI-based hemodynamic assessment, potentially broadening access to wedge pressure–guided heart failure management without invasive sensors.
Clinical Implications: If validated in home/ambulatory settings and tied to outcome improvements, this approach could enable more widespread hemodynamics-guided titration of therapies in HFrEF, potentially reducing hospitalizations without implantable devices.
Key Findings
- Wearable multimodal signals (ECG, seismocardiography, photoplethysmography) with ML estimated PCWP versus RHC with an error of 1.04 ± 5.57 mmHg.
- Limits of agreement were −9.9 to 11.9 mmHg with consistent performance across sex, race, ethnicity, and BMI.
- Prospective multicenter design with blinded core-lab adjudicated PCWP labels supports methodological rigor.
- Accuracy approaches that of implantable hemodynamic sensors, suggesting a cost-effective noninvasive alternative.
Methodological Strengths
- Prospective multicenter design with blinded core-lab adjudication of RHC PCWP
- Held-out testing set and demographic subgroup analyses to assess generalizability
Limitations
- Evaluated in HFrEF only; external validation beyond study sites and at-home use not reported
- No assessment of outcome impact from management changes based on noninvasive estimates
Future Directions: Test clinical workflows that incorporate wearable PCWP estimates to guide therapy, evaluate at-home longitudinal monitoring, and conduct randomized trials to link noninvasive hemodynamics-guided care to outcome reductions.
BACKGROUND: Remote hemodynamics-guided management of heart failure (HF) with implantable pulmonary artery pressure sensors has been shown to reduce HF hospitalizations. The widespread clinical adoption of this procedure is constrained by its invasive nature and high cost. We present a noninvasive technology based on a wearable sensor (CardioTag; Cardiosense) and machine learning (ML) for estimating pulmonary capillary wedge pressure (PCWP) in patients with heart failure with reduced ejection fraction (HFrEF). OBJECTIVES: The authors developed and evaluated (against right heart catheterization [RHC]) an ML model to estimate PCWP with the use of electrocardiography, seismocardiography, and photoplethysmography signals from CardioTag. METHODS: A multicenter prospective study was performed, and 310 patients with HFrEF (EF ≤40%) were recruited in both inpatient and outpatient settings. A blinded core laboratory adjudicated the RHC PCWP tracings to yield criterion-standard PCWP labels against which the model was trained and tested. The data were separated into 2 sets: a training set for model training and fine-tuning, and a held-out testing set unseen until final evaluation. RESULTS: The patients were 61± 13 years of age, 38% female, 44% White, and 39% African American, and had a PCWP of 18.1 ± 9.45 mm Hg. The model estimated PCWP values in the held-out test set with error of 1.04 ± 5.57 mm Hg (limits of agreement of -9.9 to 11.9 mm Hg), with consistent performance across sex, race, ethnicity, and body mass index. CONCLUSIONS: The CardioTag and its ML algorithm estimate PCWP with accuracy approaching implantable hemodynamic sensors, potentially offering a more accessible and cost-effective option for hemodynamics-guided management in HFrEF patients.
2. Deep Learning-Enhanced Noninvasive Detection of Pulmonary Hypertension and Subtypes via Chest Radiographs, Validated by Catheterization.
Using 4,576 chest radiographs with catheterization-linked labels, deep-learning models achieved high sensitivity and strong AUCs for detecting pulmonary hypertension (AUC up to 0.964) and CHD-associated PAH, with validation in internal and external RHC cohorts. Performance remained favorable even in mild disease, supporting use as a screening and triage tool.
Impact: This work demonstrates expert-level, catheterization-validated PH detection from ubiquitous imaging, enabling scalable screening, especially where echocardiography and invasive testing are limited.
Clinical Implications: CXR-based AI screening could identify patients needing echocardiography or right heart catheterization earlier, improving triage in resource-limited settings and potentially reducing diagnostic delays for PH and CHD-PAH.
Key Findings
- CXR-PH-Net achieved AUC 0.964 (internal), and 0.872 (RHC-confirmed internal), with sensitivity ~0.90; external RHC AUC 0.811 with sensitivity 0.803.
- CXR-CHD-PAH-Net achieved AUC 0.908 (internal) and 0.860 (external) with sensitivities ~0.86.
- Models maintained favorable sensitivity for CHD-PAH in mild PH cases (0.813–0.846).
Methodological Strengths
- Large dataset with catheterization-linked validation and external cohort testing
- Separate models for PH and CHD-PAH, with performance maintained in mild disease
Limitations
- Retrospective design; relatively small external RHC cohort (n=90) limits generalizability
- Specificity and calibration across diverse clinical environments require further evaluation
Future Directions: Prospective multicenter implementation studies comparing CXR-AI triage to standard care, with impact on time-to-diagnosis, resource use, and patient outcomes across diverse populations.
BACKGROUND: Pulmonary hypertension (PH) is a complex, life-threatening condition requiring noninvasive, accessible, and accurate diagnostic tools, particularly in resource-limited settings. Early and precise identification of PH and its subtypes is critical for effective management and timely intervention. RESEARCH QUESTION: Can deep learning (DL) methods applied to chest radiography (CXR) accurately detect PH and its subtype, congenital heart disease-associated pulmonary arterial hypertension (CHD-PAH)? STUDY DESIGN AND METHODS: A retrospective cohort study was conducted with 4,576 patients, including 2,288 patients with PH, who underwent CXR followed by right heart catheterization (RHC) or transthoracic echocardiography. DL models were developed and validated for detecting PH (CXR-PH-Net model) and CHD-PAH (CXR-CHD-PAH-Net model). Internal testing used a data set of 2,140 patients (1,070 patients with PH), and additional validation included an RHC-confirmed internal cohort (1,158 patients) and an external RHC cohort (90 patients) from 2 independent hospitals. Model performance was evaluated primarily using the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity. RESULTS: The CXR-PH-Net model achieved a sensitivity of 0.902 and an AUC of 0.964 for PH detection in the internal test set. In the RHC-confirmed cohort, sensitivity was 0.902 (AUC, 0.872) internally and 0.803 (AUC, 0.811) externally. The CXR-CHD-PAH-Net model demonstrated sensitivities of 0.859 and 0.870 with AUCs of 0.908 and 0.860 in the internal and external data sets, respectively. Meanwhile, the CXR-CHD-PAH-Net model showed favorable sensitivity in detecting CHD-PAH among patients with mild PH, with values of 0.813 and 0.846 in the internal and external datasets, respectively. INTERPRETATION: The CXR-PH-Net and CXR-CHD-PAH-Net models demonstrated high sensitivity as screening tools for PH and CHD-PAH, potentially facilitating early detection and triage for further evaluation, particularly in resource-limited settings. Further validation in diverse populations is warranted to enhance clinical generalizability. CLINICAL TRIAL REGISTRATION: ClinicalTrials.gov; No.: NCT05566002; URL: www. CLINICALTRIALS: gov.
3. Timing of Cardiac Resynchronization Therapy Following Stable Medical Therapy in Patients With Heart Failure.
In 9,409 Swedish registry patients, earlier CRT implantation (<3 months after achieving stable medical therapy) was associated with a lower adjusted risk of cardiovascular death versus 3–9 months, while delays >9 months were associated with higher risks of cardiovascular death, HF hospitalization, and the composite outcome.
Impact: This large, real-world analysis provides timely evidence that earlier CRT after GDMT optimization may confer survival benefits, informing practice patterns beyond guideline eligibility alone.
Clinical Implications: When CRT criteria are met after GDMT optimization, unnecessary delays beyond 3 months should be avoided. Systems-level pathways should prioritize timely referral and implantation to minimize avoidable cardiovascular mortality and HF hospitalizations.
Key Findings
- Among 9,409 patients, 43.8% received CRT <3 months after SMT, 34.9% at 3–9 months, and 21.3% >9 months, with decreasing time to CRT over years.
- CRT <3 months vs 3–9 months was associated with a 9% lower adjusted risk of cardiovascular death (P = 0.045).
- Delay >9 months vs 3–9 months was associated with a 13% higher risk of CV death/HF hospitalization, 12% higher CV death, and 11% higher first HF hospitalization.
- Determinants of earlier CRT included recent HF hospitalization, prior defibrillator, and greater GDMT use.
Methodological Strengths
- Large national registry with detailed timing relative to medical stabilization
- Multivariable logistic and Cox regression analyses adjusting for confounders
Limitations
- Observational design with potential residual confounding and selection bias
- Definition and ascertainment of stable medical therapy may vary; no randomization
Future Directions: Prospective studies and pragmatic trials to test expedited CRT pathways after GDMT optimization, including impacts on mortality, hospitalizations, and quality of life.
BACKGROUND: Guidelines' recommendations for cardiac resynchronization therapy (CRT) implantation in selected patients with heart failure (HF) exist. However, data on the best timing for CRT implantation after the achievement of stable medical therapy (SMT) and its association with outcomes are currently lacking. OBJECTIVES: The aim of this study was to investigate the timing of CRT implantation after the achievement of SMT, associated patient profiles, and clinical outcomes in a real-world HF population. METHODS: Patients with HF treated with SMT derived from the Swedish ICD and Pacemaker Registry who received CRT between 2007 and 2020 were included in the study. Patient characteristics associated with a shorter or longer time to CRT implantation were assessed using multivariable logistic regression, and associations between the time from SMT to CRT implantation and clinical outcomes (mortality and morbidity) were analyzed using multivariable Cox regression. RESULTS: Of the 9,409 patients, 43.8% received CRT at <3 months of achieving SMT, 34.9% between 3 and 9 months, and 21.3% after 9 months. The time from SMT to CRT implantation decreased significantly over the study period. Independent determinants of shorter time to implantation included recent HF hospitalization, previous implantation of a defibrillator, and greater use of guideline-directed medical therapy, whereas a history of HF >6 months and ischemic heart disease were associated with a longer time. After adjustments, there was a 9% lower risk of cardiovascular death with a shorter time from SMT to CRT implantation of <3 months vs 3-9 months (P = 0.045). A delayed time of >9 months vs 3-9 months was associated with a 13% higher risk of cardiovascular death/HF hospitalization, a 12% higher risk of cardiovascular death (P = 0.040), and an 11% higher risk of first HF hospitalization (P = 0.013). CONCLUSIONS: Time from the achievement of SMT to CRT implantation decreased over the study period. Delayed CRT implantation beyond 3 months was associated with higher cardiovascular mortality compared with earlier implantation after GDMT optimization.