Current Source Density Imaging Using Regularized Inversion of Acoustoelectric Signals

Acoustoelectric (AE) imaging can potentially image biological currents at high spatial (similar to mm) and temporal (similar to ms) resolution. However, it does not directly map the current field distribution due to signal modulation by the acoustic field and electric lead fields. Here we present a new method for current source density (CSD) imaging. The fundamental AE equation is inverted using truncated singular value decomposition (TSVD) combined with Tikhonov regularization, where the optimal regularization parameter is found based on a modified L-curve criterion with TSVD. After deconvolution of acoustic fields, the current field can be directly reconstructed from lead field projections and the CSD image computed from the divergence of that field. A cube phantom model with a single dipole source was used for both simulation and bench-top phantom studies, where 2D AE signals generated by a 0.6 MHz 1.5D array transducer were recorded by orthogonal leads in a 3D Cartesian coordinate system. In simulations, the CSD reconstruction had significantly improved image quality and current source localization compared to AE images, and performance further improved as the fractional bandwidth (BW) increased. Similar results were obtained in the phantom with a time-varying current injected. Finally, a feasibility study using an in vivo swine heart model showed that optimally reconstructed CSD images better localized the current source than AE images over the cardiac cycle.

Automated summary

Acoustoelectric (AE) imaging has the potential to image biological currents with high spatial and temporal resolution. However, it does not directly map the current field distribution. In this study, a new method for current source density (CSD) imaging is proposed, which involves inverting the fundamental AE equation using truncated singular value decomposition (TSVD) combined with Tikhonov regularization. The feasibility of the method is demonstrated through simulation, bench-top phantom studies, and in vivo swine heart models.