Artificial intelligence for detecting electrolyte imbalance using electrocardiography.

Area of Science:

• Biomedical Engineering
• Cardiology
• Artificial Intelligence in Medicine

Background:

• Electrolyte imbalances are critical in managing metabolic diseases but lack reliable, noninvasive detection tools.
• Current methods for electrolyte monitoring are invasive and may not be suitable for continuous assessment.

Purpose of the Study:

• To develop and validate a deep learning model (DLM) for noninvasive detection of electrolyte imbalances using electrocardiography (ECG).
• To assess the DLM's performance across multiple centers for detecting various electrolyte abnormalities.

Main Methods:

• A retrospective cohort study involving 92,140 patients from two hospitals.
• Development of a DLM using 83,449 ECGs, with internal validation on 12,091 ECGs.
• External validation of the DLM using 31,693 ECGs from a separate hospital.

Main Results:

• The DLM achieved high Area Under the Curve (AUC) values for detecting hyperkalemia (0.945 internal, 0.873 external) and other imbalances.
• External validation demonstrated robust performance, with AUCs ranging from 0.813 to 0.873 for different electrolyte conditions.
• The model visualized key ECG regions (P wave, QRS, T wave) crucial for specific electrolyte imbalance detection.

Conclusions:

• The developed DLM demonstrates significant potential for accurate and noninvasive detection of electrolyte imbalances.
• The findings suggest that DLMs utilizing ECG data can be employed for daily monitoring of electrolyte status.
• This approach offers a promising noninvasive alternative for managing patients with metabolic diseases requiring electrolyte monitoring.