Electrocardiogram (ECG) signals are widely used by cardiologists for the early detection of cardiovascular diseases (CVDs). In the early detection of CVDs, long-term ECG data is used for analysis. Healthcare devices used for the acquisition of long-term ECG data require an efficient ECG data compression algorithm. But compression of ECG signal with maintaining its quality is a challenge. Hence, this chapter presents a quality-controlled compression method that compresses the ECG data efficiently with retaining its quality up to a certain mark. For this, a distortion measure is used with specifying its value in a tolerable range. The compression performance of the proposed algorithm is evaluated using ECG records of the MIT-BIH arrhythmia database. In performance assessment, it is found that the compression algorithm performs well. The compressed ECG data are also used for normal and arrhythmia beat classification. The classification performance for ECG beats obtained from the compressed ECG data is good. It denotes the better diagnostic quality of the compressed ECG data.
TopIntroduction
In cardiology, electrocardiogram (ECG) signals are widely used as a non-invasive diagnostic tool to detect abnormal behavior of heart. In hospitals, usually, ECG signals are recorded using 12-lead ECG acquisition machine. These ECG signals are analyzed by cardiologist to diagnose cardiovascular diseases (CVD’s). In early detection of CVD’s, ECG signals are recorded for long-term which may vary from 12-72 hours (Jha & Kolekar, 2021a). ECG data acquired in this duration requires large memory space for storage. Large ECG data also consumes large bandwidth to transmit it from one terminal to other. It results in increased transmission cost. An ECG data acquisition and transmission model used in remote healthcare is shown in Figure 1. This model is used for cardiac patient monitoring from remote. In ambulatory monitoring of cardiac patients, wearable ECG devices are used which can acquire ECG data during normal activity of patients. These ECG devices use wireless ECG sensors for which handling of large ECG data increases power consumption. For efficient performance of these devices, compression of ECG data size is essential to reduce the memory space requirement, transmission cost and power consumption (Lee et al., 2011), (Jha & Kolekar, 2016).
Figure 1. Cardiac patient monitoring model in remote healthcare
In past, several techniques have been developed for ECG data compression. These techniques are broadly categorized into two types (Manikandan & Dandapat, 2014): lossless and lossy. Lossless ECG data compression techniques offer no any distortion of signal, but compression ratio obtained by these techniques are very low. Examples of lossless ECG data compression techniques are Huffman encoding, run-length encoding, entropy coding, and linear prediction. In order to achieve higher compression ratio, lossy techniques have been developed with more focus. Further, lossy compression techniques are classified into three types. These types are (Jalaleddine et al., 1990): direct-time domain, parameter extraction and transform domain.
In direct time-domain techniques, ECG data is analyzed and compressed directly without changing it in another domain (Alvarado et al., 2012) For ECG data, many direct time-domain techniques have been developed such as turning-point (TP), scan along polynomial approximation (SAPA), amplitude zone time epoch coding (AZTEC), coordinate reduction time encoding system (CORTES), differential pulse code modulation (DPCM) and entropy coding. These techniques involve simple steps to implement but achieved compression ratio by these techniques are poor. In parameter extraction techniques, important features of ECG signals are extracted and based on those features ECG signals are reconstructed again. Examples of parameter extraction techniques are neural network-based method, peak picking and linear prediction method (Deepu & Lian, 2014), (Adamo et al., 2015). Transform domain compression methods analyze energy distribution of signal by converting it to another domain.
In past, many ECG data compression techniques have been developed based on transform domain methods. These techniques widely use discrete Fourier transform (DFT) (Sadhukhan et al., 2015), discrete cosine transform (DCT) (Jha & Kolekar, 2017), (Lee et al., 2011) and discrete wavelet transform (DWT) (Hossain, 2011), (Abo-Zahhad et al., 2013; Jha & Kolekar, 2015, 2019a, 2019c, 2021c, 2021b; Kolekar et al., 2021; Motinath et al., 2016). An ECG data compression technique is proposed in (Sadhukhan et al., 2015) which uses adaptive bit encoding of DFT coefficients. In this technique, DFT coefficients are calculated using sine and cosine basis functions. It avoids to get complex coefficients values. Further, DFT coefficients are encoded using fixed and adaptive strategies. In (Lee et al., 2011), an ECG data compression algorithm based on DCT is proposed. It utilizes DCT to compact energy of the signal. This algorithm is suitable for real-time compression and transmission of ECG data between e-health terminals. A DCT based ECG data compression algorithm along with dual encoding technique is proposed in (Jha & Kolekar, 2017). It compresses ECG data very well and it is suitable for tele-monitoring of cardiac patients.