A novel method for analysis of transient morphological changes in quasiperiodic physiological signals and their...

Transient changes in electrocardiographic (ECG) morphology often accompany fluctuations in heart rhythm and may provide clinically relevant information about cardiac function. However, conventional ECG visualization and analysis methods typically emphasize either waveform morphology or rhythm variability, making it difficult to assess their temporal interplay in long-term recordings and, consequently, to identify rhythm-related morphological changes and exploit their potential diagnostic value. This work presents a visualization framework for quasiperiodic physiological signals that enables simultaneous assessment of beat-to-beat morphological changes and rhythm dynamics in a single representation.

The proposed method converts quasiperiodic signals into two-dimensional carpet plots. Characteristic events (e.g., ECG R peaks) are used to align consecutive signal segments, which are transformed into color-coded rows and stacked in chronological order. The resulting image preserves both the temporal evolution of signal morphology and variations in cycle duration. The method was evaluated using ECG recordings from multiple publicly available databases containing healthy subjects and patients with diverse cardiac abnormalities, as well as synchronized multimodal physiological recordings.

Carpet plots enabled rapid visualization of transient morphological changes alongside heart rate dynamics across recordings ranging from several minutes to hours. The representation highlighted clinically relevant phenomena, including ST segment alterations, QT interval variability, changes in T wave morphology, atrial fibrillation episodes, premature ventricular complexes, Wenckebach periodicity, and stress-test phase transitions. The image-based representation was also shown to be suitable for automated analysis using convolutional neural network feature extraction.

Carpet plots provide a compact representation of quasiperiodic physiological signals, jointly visualizing rhythm and morphology across long-term recordings. The proposed framework facilitates both expert interpretation and image-based computational analysis, offering a general approach for investigating transient physiological phenomena in ECG and other synchronized quasiperiodic signals.