Cardiorenal Interorgan Assessment via a Novel Clustering Method Using Dynamic Time Warping on Electrocardiogram:...

Area of Science:

• Cardiology
• Nephrology
• Medical Informatics

Background:

• The heart and kidneys have a reciprocal physiological relationship, where dysfunction in one organ can negatively impact the other.
• Heart failure with preserved ejection fraction (HFpEF) affects over 50% of heart failure patients, and 1 in 6 with chronic kidney disease (CKD) also have HF.
• Predicting and understanding the cardiorenal link between HFpEF and CKD is crucial for patient management.

Purpose of the Study:

• To develop an electrocardiogram (ECG)-based model for stratifying HFpEF patients and identifying CKD-enriched subgroups.
• To derive a minimal set of significant ECG features for accessible precision diagnostics.
• To validate the cardiorenal relationship between HFpEF and CKD using ECG data for enhanced biological insight.

Main Methods:

• Unsupervised clustering of ECG features from FinnGen data for patients with HFpEF (LVEF ≥50%, NT-proBNP >450 pg/mL).
• Isolation of significant ECG features (PR interval, QRS duration) for phenogrouping and risk analysis.
• Comparison of clustering methods: k-means (all features), k-means (minimal features), and dynamic time warping (DTW) on raw ECG signals, followed by CKD risk assessment.

Main Results:

• Dynamic time warping (DTW) clustering of lead II ECG waveforms yielded the most stable patient clusters.
• ANOVA analysis revealed significant deviations in CKD risk among several HFpEF clusters, indicated by creatinine levels.
• DTW clusters demonstrated the highest concordance with baseline clusters formed using creatinine levels, validating the cardiorenal relationship.

Conclusions:

• The study validates the cardiorenal relationship between HFpEF and CKD using ECG data.
• DTW clustering of ECG lead II waveforms proved effective for clinically meaningful patient stratification in HFpEF.
• This ECG-based clustering methodology holds potential for broader applications in patient stratification beyond HFpEF.