Publications by authors named "Sumukh Vasisht Shankar"
Aims: Artificial intelligence (AI)-enhanced 12-lead electrocardiogram (ECG) can detect a range of structural heart diseases (SHDs); however, it has a limited role in community-based screening. We developed and externally validated a noise-resilient single-lead AI-ECG algorithm that can detect SHDs and predict the risk of their development using wearable/portable devices.
Methods And Results: Using 266 740 ECGs from 99 205 patients with paired echocardiographic data at Yale New Haven Hospital, we developed AI Deep learning for Adapting Portable Technology in HEART disease detection (ADAPT-HEART), a noise-resilient, deep learning algorithm, to detect SHDs using lead I ECG.
Aims: Rich data in cardiovascular diagnostic testing are often sequestered in unstructured reports, limiting their use.
Methods And Results: We sequentially deployed generative and interpretative open-source large language models (LLMs; Llama2-70b, Llama2-13b). Using Llama2-70b, we generated varying formats of transthoracic echocardiogram (TTE) reports from 3000 real-world reports with paired structured elements.
Background And Aims: The diagnosis of transthyretin amyloid cardiomyopathy (ATTR-CM) requires advanced imaging, precluding large-scale preclinical testing. Artificial intelligence (AI)-enabled transthoracic echocardiography (TTE) and electrocardiography (ECG) may provide a scalable strategy for preclinical monitoring.
Methods: This was a retrospective analysis of individuals referred for nuclear cardiac amyloid testing at the Yale-New Haven Health System (YNHHS, internal cohort) and Houston Methodist Hospitals (HMH, external cohort).
Background: Accurate aortic stenosis (AS) phenotyping requires access to multimodality imaging which has limited availability. The Digital Aortic Stenosis Severity Index (DASSi), an AI biomarker of AS-related remodeling on 2D echocardiography, predicts AS progression independent of Doppler measurements. Whether DASSi-enhanced echocardiography provides a scalable alternative to multimodality AS imaging remains unknown.
View Article and Find Full Text PDFBackground: Identifying structural heart diseases (SHDs) early can change the course of the disease, but their diagnosis requires cardiac imaging, which is limited in accessibility.
Objectives: The purpose of this study was to leverage images of 12-lead electrocardiograms (ECGs) for automated detection and prediction of multiple SHDs using an ensemble deep learning approach.
Methods: We developed a series of convolutional neural network models for detecting a range of individual SHDs from images of ECGs with SHDs defined by transthoracic echocardiograms performed within 30 days of the ECG at the Yale New Haven Hospital (YNHH).
Development and Multinational Validation of an Ensemble Deep Learning Algorithm for Detecting and Predicting Structural Heart Disease Using Noisy Single-lead Electrocardiograms.
medRxiv
October 2024
Background And Aims: AI-enhanced 12-lead ECG can detect a range of structural heart diseases (SHDs) but has a limited role in community-based screening. We developed and externally validated a noise-resilient single-lead AI-ECG algorithm that can detect SHD and predict the risk of their development using wearable/portable devices.
Methods: Using 266,740 ECGs from 99,205 patients with paired echocardiographic data at Yale New Haven Hospital, we developed ADAPT-HEART, a noise-resilient, deep-learning algorithm, to detect SHD using lead I ECG.
Background: Identifying structural heart diseases (SHDs) early can change the course of the disease, but their diagnosis requires cardiac imaging, which is limited in accessibility.
Objective: To leverage images of 12-lead ECGs for automated detection and prediction of multiple SHDs using an ensemble deep learning approach.
Methods: We developed a series of convolutional neural network models for detecting a range of individual SHDs from images of ECGs with SHDs defined by transthoracic echocardiograms (TTEs) performed within 30 days of the ECG at the Yale New Haven Hospital (YNHH).
Article Synopsis
- Rich cardiovascular data is often hidden in unstructured reports, making it difficult to use in real-time patient care and research due to manual abstraction requirements.
- A two-step process was developed using generative and interpretative large language models (LLMs) to convert these unstructured reports into usable formats, specifically focusing on transthoracic echocardiograms (TTE).
- The HeartDX-LM model demonstrated impressive accuracy, extracting 98.7% of values from unstructured reports across various datasets, proving its effectiveness in improving data accessibility for clinical analysis.
Article Synopsis
- Assessment of stroke risk in patients with atrial fibrillation (AF) is important for anticoagulation therapy, and the CHADS-VASc score is commonly used for this purpose, though it often relies on manual calculations or simplified data from electronic health records (EHR).
- Researchers developed a Retrieval-Augmented Generation (RAG) approach using the Llama3.1 language model to effectively extract CHADS-VASc risk factors from unstructured clinical notes, improving the accuracy of risk assessment.
- The RAG model outperformed traditional structured data in identifying key risk factors such as hypertension and diabetes, leading to higher and more accurate CHADS-VASc scores when these unstructured insights were incorporated into the assessment.
Article Synopsis
- Researchers explored using artificial intelligence (AI) to improve the diagnosis of transthyretin amyloid cardiomyopathy (ATTR-CM) through echocardiograms (TTE) and electrocardiograms (ECG), potentially allowing for earlier detection of the disease.
- They trained deep learning models to identify ATTR-CM patterns, achieving high accuracy in recognizing these signatures from cardiac data in two large patient groups.
- The study found that AI can effectively predict the likelihood of ATTR-CM in individuals up to three years before a formal diagnosis, suggesting that it could help identify patients who might benefit from early treatment options.
Article Synopsis
- * A machine learning model, called ARISE, was developed to improve screening for elevated Lp(a) levels using data from the UK Biobank and validated in three other large studies.
- * ARISE significantly reduces the number of tests needed to identify someone with high Lp(a), making it easier to utilize common clinical data and potentially implement in electronic health records for better screening outcomes.
Article Synopsis
- Randomized clinical trials (RCTs) help inform medical practice, but their applicability to different populations can be unclear.
- The RCT-Twin-GAN model was developed to create digital twins of RCTs that simulate treatment effects using data from varying patient populations.
- The model successfully reproduced treatment effects from two notable studies, SPRINT and ACCORD, demonstrating its potential to bridge gaps in understanding how different populations might respond to medical interventions.
Article Synopsis
- - The study aims to assess the effectiveness of an AI-based algorithm in detecting structural heart disease (SHD) using portable ECG devices, focusing on its real-world application compared to established echocardiogram methods.
- - It will involve enrolling 585 patients who will undergo a single-lead ECG with both an Apple Watch and another portable device during their echocardiogram routine visits, linking their ECG data to electronic health records for analysis.
- - Ethical considerations include secure access to patient data and maintaining confidentiality, ensuring that the study adheres to strict guidelines for handling protected health information.
Article Synopsis
- Accurate assessment of ECGs is vital for patient diagnosis and care, but current automated systems lack flexibility and reliability, especially in low-resource areas where specialists review each ECG manually.
- AI systems show promise for improved accuracy but often have limitations in the variety of conditions they can assess and require raw data not typically available to doctors.
- The ECG-GPT model was developed to generate expert-level diagnosis directly from ECG images, demonstrating strong performance across diverse healthcare settings and providing an accessible web application for accurate triage of patients.
Article Synopsis
- Randomized clinical trials (RCTs) often fail to represent key populations adequately, making it challenging to assess their real-world effectiveness; the study proposes using digital twins of RCTs to better understand these effects using electronic health records (EHR).
- A new model called RCT-Twin-GAN generates RCT-like datasets by simulating data from EHRs, allowing researchers to analyze the impact of treatments like spironolactone on heart failure patients more accurately.
- The results showed that the simulated RCT-Twin data closely mirrored real RCT results, with a balanced representation of covariates and similar treatment effects, suggesting that this approach could enhance the relevance of RCT findings in broader clinical contexts.
In the rapidly evolving landscape of modern healthcare, the integration of wearable and portable technology provides a unique opportunity for personalized health monitoring in the community. Devices like the Apple Watch, FitBit, and AliveCor KardiaMobile have revolutionized the acquisition and processing of intricate health data streams that were previously accessible only through devices only available to healthcare providers. Amidst the variety of data collected by these gadgets, single-lead electrocardiogram (ECG) recordings have emerged as a crucial source of information for monitoring cardiovascular health.
View Article and Find Full Text PDFArticle Synopsis
- * Researchers created a machine learning model called ARISE, which effectively predicts elevated Lp(a) levels using various health data, outperforming existing cardiovascular risk assessments.
- * In a study involving over 456,000 participants, the ARISE model demonstrated a much lower number-needed-to-test (NNT) for identifying elevated Lp(a) compared to traditional methods, suggesting it could streamline screening.