A Hybrid Lossless ECG Compressor with Morphological Detection and QRS Prediction

This paper proposes a hybrid approach to lossless ECG signal compression, capitalizing on the repetitive patterns inherent in the QRS complex. The proposed processor integrates a morphological filter QRS detection channel and 12 lead channels for linear prediction and packaging. By subtracting the detected QRS complexes using predictions based on previous buffers, the proposed algorithm generates Non-QRS signals with a smaller dynamic range. The Non-QRS signals undergo further processing using a linear predictor before being formatted into 32-bit packaged outputs. The proposed method achieves an average compression ratio of 2.32, showcasing a 3% improvement over the MIT-BIH dataset compared to methods without QRS removal. The proposed work was implemented with 40nm CMOS technology. The 12-lead ECG compression chip consumes a power of 179nW under 0.9V and occupies a silicon area of 16,032µm² per channel. These results yield a 19× higher FoM than prior works, underscoring the superior efficiency and effectiveness of the proposed approach.