To evaluate whether a single, unified model can achieve comparable performance to a multiple-model pipeline in ECG reconstruction without the need for denoising.
Approach:
Study Design: The study compares the multiple-model pipeline (MMP) and the single-model pipeline (SMP) using 10,000 normal ECG records from the CODE-15% database.
Data Processing: ECG records were trimmed to 10 seconds, resampled to 500 Hz, and denoised to serve as a reference for added artificial noise.
Key Findings:
The SMP performs comparably to the MMP, achieving average correlations above 0.85.
SMP maintains stable performance across varying noise levels with approximately 0.04 variation.
Fine-tuning the SMP through transfer learning can restore performance on abnormal ECG with minimal data.
Interpretation:
Limitations:
The SMP performs poorly on abnormal ECG without fine-tuning.
The study primarily focuses on normal ECG records, limiting generalizability to diverse patient populations.
