A Novel 2-Lead to 12-Lead ECG Reconstruction Methodology for Remote Health Monitoring Applications

Naresh Vemishetty1, Vishnuvardhan Gundlapalle2, Amit Acharyya3, Bhudeb Chakravarti4
1IIT Hyderabad, 2Indian Institute of Technology Hyderabad, 3Indian Institute of Technology, Hyderabad, 4Sense Health Technologies private Limited


Cardiovascular diseases (CVD) today are one of the prime causes of human mortality worldwide, and therefore tremendous research has been done for the prevention of CVD. Advancement in technology has made remote ECG monitoring possible with considerably small equipment. Standard clinical 12-lead ECG is the setup most commonly used by the doctors for the reliable diagnosis. However, placing all the electrodes on the body will be inconvenient and affect the daily activities of the patient. In addition, transmitting all the lead information will require high power and large memory. Accounting this, we propose here a novel 2 to 12 lead reduction system.

There are few research articles available for reduced 3 lead to 12 lead ECG reconstruction predominantly by authors themselves. Reduction from the state of the art 3 leads to proposed 2 lead minuscule from the theoretical perspective. However, it will have a significant impact on patient psychology. In the proposed method, the coefficient values of all the leads are generated as the first step using the Least-Square (LS) fit method and Heart-Vector Projection (HVP) (Fig.1). For the 12 lead ECG reconstruction, lead I and V2 are taken as the basis leads (Fig.2). As a part of the reconstruction, Lead II is derived initially from these two basis leads (Lead I, V2) using the HVP computation. After Lead II derivation, these three (I, II, V2) leads acts as basis leads and will derive remaining leads (V1, V3, V4, V5, V6) by repeating the HVP computation.

ECG database of healthy (80 subjects) and unhealthy (333 MI and 16 BBB subjects) has been taken from PTBDB to validate the proposed method. The reconstruction results (Fig.3) are evaluated using R2 statistics, correlation and regression coefficients and achieved significant accuracy for each lead (Table I) without significant loss of any medical features.