Cardiac eigen imaging: a novel method to isolate cardiac activity in thoracic electrical impedance tomography

As the global burden of cardiovascular disease increases, proactive cardiovascular healthcare by means of accurate, precise, continuous, and non-invasive monitoring is becoming crucial. However, no current device is able to provide cardiac hemodynamic monitoring with the aforementioned criterion. Electrical impedance tomography (EIT) is an inexpensive, non-invasive imaging modality that can provide real-time images of internal conductivity distributions that describe physiological activity. This work explores and compares a standard approach of regular cardiac gated averaging (RCGA) and a newly developed method, cardiac eigen-imaging (CEI), based on the singular value decomposition (SVD) to isolate cardiac activity in thoracic EIT.

EIT and heart-rate (HR) data were collected from 20 heart-failure patients preceding echocardiography. Features from RCGA and CEI images were correlated with stroke volume (SV) from echocardiography and image reconstruction parameters were optimized using leave-one-out (LOO) cross-validation.

CEI per-pixel-based features achieved a Pearson correlation coefficient of 0.80 with SV relative to 0.72 with RCGA. CEI had 33 high-correlating pixels while RCGA had 8. High-correlating pixels tend to concentrate in the right-ventricle (RV) when referenced to a general chest model.

While both RCGA and CEI images had high-correlating pixels, CEI had higher correlations, a larger number of high-correlating pixels, and unlike RCGA is not dependent on the quality of the HR data collected. The observed performance of the CEI approach represents a promising step forward for EIT-based cardiac monitoring in either clinical or ambulatory settings.