Open-source hubert-ecg-large model, based on 9.1 million ECG data, helps diagnose 164 cardiovascular diseases!

Hubert Ecg Large

Developed by Edoardo-BS

A self-supervised foundational model for broadly scalable cardiac applications, trained on 9.1 million 12-lead ECGs covering 164 cardiovascular diseases

Molecular Model

Transformers

#ECG Analysis #Self-supervised Pretraining #Multi-disease Detection

Downloads 168

Release Time : 12/2/2024

Model Overview

An ECG foundational model based on self-supervised pretraining, fine-tunable for various cardiovascular disease diagnosis and event prediction tasks

Model Features

Large-scale Pretraining

Self-supervised pretraining using 9.1 million 12-lead ECGs

Multi-disease Detection

Capable of detecting 164 cardiovascular diseases, with AUROC exceeding 90% for 140 diseases

Event Prediction Capability

Predicts death events within 2 years (AUROC reaching 93.4%)

Transfer Learning Friendly

Only requires adding an output layer for fine-tuning on various downstream tasks

Model Capabilities

ECG Analysis

Cardiovascular Disease Diagnosis

Cardiovascular Event Prediction

Multi-disease Simultaneous Detection

Use Cases

Medical Diagnosis

Cardiovascular Disease Detection

Detects 164 cardiovascular diseases

AUROC>90% for 140 diseases, >95% for 94 diseases

Prognostic Prediction

Mortality Risk Prediction

Predicts death events within 2 years

AUROC reaches 93.4%

🚀 HuBERT-ECG: A Self-Supervised Foundation Model for Broad and Scalable Cardiac Application

HuBERT-ECG is a self - supervised pre - trained foundation model for electrocardiogram (ECG) analysis. It can be fine - tuned for various downstream tasks and shows excellent performance across diverse real - world scenarios.

🚀 Quick Start

Original code is available at https://github.com/Edoar-do/HuBERT-ECG. The project is licensed under CC BY - NC 4.0.

✨ Features

Deep learning models have shown remarkable performance in electrocardiogram (ECG) analysis. However, their success has been limited by the scarcity and size of ECG datasets, leading to systems that are more task - specific rather than versatile. HuBERT - ECG is pre - trained in a self - supervised manner on a large and diverse dataset of 9.1 million 12 - lead ECGs covering 164 cardiovascular conditions. By simply adding an output layer, it can be fine - tuned for a wide range of downstream tasks, such as diagnosing diseases and predicting future cardiovascular events.

High Performance: Across diverse real - world scenarios, HuBERT - ECG achieves AUROCs from 84.3% in low - data settings to 99% in large - scale setups.
Multi - condition Detection: When trained to detect 164 overlapping conditions simultaneously, the model delivers AUROCs above 90% and 95% for 140 and 94 diseases, respectively.
Death Event Prediction: HuBERT - ECG predicts death events within a 2 - year follow - up with an AUROC of 93.4%.

📦 Installation

No specific installation steps are provided in the original README.

💻 Usage Examples

Basic Usage

from transformers import AutoModel
size = 'small' # any size from small, base, large
hubert_ecg = AutoModel.from_pretrained(f"Edoardo-BS/hubert-ecg-{size}", trust_remote_code=True)

📚 Documentation

Models

This repository contains the self - supervised pre - trained hubert - ecg - large.

Citation

If you use our models or find our work useful, please consider citing us:

doi: https://doi.org/10.1101/2024.11.14.24317328

📄 License

The project is licensed under CC BY - NC 4.0.

⚠️ Important Note

Don't forget to pre - process your data! Read the paper to know more about it

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご