Inter-patient ECG Arrhythmia Classification with LGNs and LUTNs

TL;DR

This work tackles inter-patient ECG arrhythmia classification with ultra-low-power logic-gate-based networks. It introduces rate-coded LGNs and LUTNs, a novel MUX-based LUT training method, and a refined preprocessing pipeline, benchmarking on the MIT-BIH dataset. Across inter-patient experiments, accuracies exceed 94% and a $j\kappa$-index up to 0.683 are achieved with shallow LGNs and LUTNs, while consuming orders of magnitude fewer FLOPs than state-of-the-art methods. FPGA-based estimates show low power (5–7 mW) and modest LUT counts, supporting potential deployment in wearables or implants for rapid, energy-efficient arrhythmia detection.

Abstract

Deep Differentiable Logic Gate Networks (LGNs) and Lookup Table Networks (LUTNs) are demonstrated to be suitable for the automatic classification of electrocardiograms (ECGs) using the inter-patient paradigm. The methods are benchmarked using the MIT-BIH arrhythmia data set, achieving up to 94.28% accuracy and a $jκ$ index of 0.683 on a four-class classification problem. Our models use between 2.89k and 6.17k FLOPs, including preprocessing and readout, which is three to six orders of magnitude less compared to SOTA methods. A novel preprocessing method is utilized that attains superior performance compared to existing methods for both the mixed-patient and inter-patient paradigms. In addition, a novel method for training the Lookup Tables (LUTs) in LUTNs is devised that uses the Boolean equation of a multiplexer (MUX). Additionally, rate coding was utilized for the first time in these LGNs and LUTNs, enhancing the performance of LGNs. Furthermore, it is the first time that LGNs and LUTNs have been benchmarked on the MIT-BIH arrhythmia dataset using the inter-patient paradigm. Using an Artix 7 FPGA, between 2000 and 2990 LUTs were needed, and between 5 to 7 mW (i.e. 50 pJ to 70 pJ per inference) was estimated for running these models. The performance in terms of both accuracy and $jκ$-index is significantly higher compared to previous LGN results. These positive results suggest that one can utilize LGNs and LUTNs for the detection of arrhythmias at extremely low power and high speeds in heart implants or wearable devices, even for patients not included in the training set.

Figures (7)

• Figure 1: Excerpt of the ECG data highlighting some of the features to be extracted for the heartbeat that is indicated in gray.
• Figure 2: An example of a 3-layer LGN.
• Figure 3: Equivalence between an 8:1 MUX and a 3-input LUT.
• Figure 4: An example of a 4 layer 6-LUTN.
• Figure 5: Comparison between single threshold coding and rate coding on the MNIST data set (a) and Fashion-MNIST data set (b), for LGNs using 8000 gates per layer.