12-Lead ECG-Driven Cardiac TMP Recovery VIA a Joint Physiological Model

Accurate recovery of cardiac transmembrane potential (TMP) from body surface potential (BSP) plays an important role in non-invasive diagnosis of heart disease. However, most existing TMP recovery methods are established based on multi-lead ECG (64-256 leads), which requires expensive equipment with complex signal acquisition processes. In this work, we proposed to recover TMP from the cheaper and more convenient 12-lead ECG. To achieve this, we develop a joint physiological model to address the serious ill-conditioned inverse problem caused by information asymmetry between 12-lead signal and TMP. To be specific, we introduce the self-attention module to capture the long-term, periodic dynamic activation process characteristics of cardiac electrophysiol-ogy, and the cross-attention module to capture the dynamic characteristics of the inverse relationship. The above process is completed in multiple subspace at the same time, which expands the model’s ability to learn different features of the sequence. With a series of synthetic and clinical experiments, the method is proved to be effective and convincing.