Development and Validation of a Deep-Learning Model to Screen for Hyperkalemia From the Electrocardiogram
Problem
Hyperkalemia is common, potentially fatal, and often asymptomatic in chronic kidney disease patients, while guideline-directed blood potassium monitoring is underused.
Why it matters
Undetected hyperkalemia can cause fatal arrhythmias in chronic kidney disease patients.
What the AI did
A deep convolutional neural network was trained on 1,576,581 ECGs from 449,380 patients to noninvasively detect hyperkalemia from ECG leads, then validated on retrospective data across three Mayo Clinic sites.
Results
not stated in source
Impact
Model developed and validated across three Mayo Clinic sites; performance metrics and deployment not in abstract.
Record
| Venue | JAMA Cardiology |
|---|---|
| Published | 2019 |
| Work type | article |
| Authors | Conner D. Galloway, Alexander V. Valys, Jacqueline Baras Shreibati, Daniel Treiman, Frank L. Petterson, Vivek P. Gundotra, D.J. Albert, Zachi Itzhak Attia et al. |
| Citations | 318 (OpenAlex, at import time) |
| References | 38 |
| Model or method | deep convolutional neural network (DNN) |
| Organization | Mayo Clinic |
| Primary source | https://doi.org/10.1001/jamacardio.2019.0640 |
| Open access | bronze |
| Keywords | hyperkalemia, ECG, deep learning, screening, chronic kidney disease |
Importance rating
significant — thousands to millions affected
Abstract states hyperkalemia is common in CKD, associated with fatal arrhythmias, often asymptomatic, and monitoring is underused; a life-threatening detection gap addressed via noninvasive ECG screening.