Novel Bloodless Potassium Determination Using a Signal-Processed Single-Lead ECG



Zachi I Attia1,2, Christopher V DeSimone1, John J Dillon3, Yehu Sapir2, Virend K Somers1, Jennifer Dugan1, Charles Bruce1, michael Ackerman1, Samuel Asirvatham1, Bryan Striemer4, Jan Bukartyk1, Christopher Scott5, Kevin E Bennet4, Dorothy Ladewig6, Emily Gilles6, Dan Sadot2, Amir B Geva2, Paul A Friedman1
1 Division of Cardiovascular Diseases, Mayo Clinic, USA
2 Electrical and Computer Engineering, Ben-Gurion University of the Negev, Israel
3 Nephrology and Hypertension, Mayo Clinic, USA
4 Division of Engineering, Mayo Clinic, USA
5 Biomedical Statistics and Informatics, Mayo Clinic, USA
6 Mayo Clinic Ventures, Mayo Clinic, USA

Aims: Potassium homeostasis is critical for electrically active cells; its aberration can result in life-threatening arrhythmias. Our aim is to predict potassium from a single surface ECG recording, providing significant improvement in the ability to monitor potassium levels in high-risk patients in remote monitored settings.

Methods and results:

Two groups of hemodialysis patients (development group, n=26 and validation group, n=20) underwent digital ECG along with 2-3 arterial blood draws to obtain potassium levels over three dialysis sessions. ECG T wave features were extracted using advanced signal processing methods. Personalized prediction models were developed for each patient using the first session to predict potassium values during the second and third dialysis sessions using only the processed single channel ECG. Additionally, by analyzing the entire development group’s first visit data, we created a global model for all patients that was validated against subsequent sessions in the development group and in a separate validation group. For the personalized model we successfully predicted potassium values with an absolute error of 0.36 ± 0.34 mmol/L (or 10% of the measured blood potassium). For the global model, potassium prediction was also accurate, with an absolute error of 0.44 ± 0.47 mmol/L for the training group (or 11% of the measured blood potassium) and 0.5 ± 0.42 for the validation set (or 12% of the measured blood potassium).

Conclusions:

These findings suggest that a signal processed ECG recorded from a single lead can be used for potassium prediction using a personalized or a global template to yield a clinically meaningful resolution. These early studies will need to be tested in a larger cohort, to demonstrate the potential for clinical applications in remote monitored settings to provide more effective, safe, and personalized drug titration and care.