Optimization of Transcranial Acoustoelectric Brain Imaging of Neuronal Currents in a Human Head Model

Abstract

Mapping brain function in humans requires high spatial and temporal resolution when resolving fast, localized neural activity. While scalp electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) have advanced neurological research, EEG has poor spatial resolution and fMRI lacks temporal sensitivity. These inherent limitations create a critical gap in non-invasive electrical brain mapping at the neuronal scale, limiting progress in diagnosing and treating disorders such as epilepsy, Alzheimer’s disease, Parkinson’s disease, and depression.Transcranial Acoustoelectric Brain Imaging (tABI) with neuronavigation is an emerging method for mapping neuronal currents through the human skull. In tABI, ultrasound (US) is focused and steered within the brain while surface electrodes record the resulting acoustoelectric (AE) interaction signal, enabling the direct mapping of current density. A key advantage of tABI is its ability to achieve millimeter-scale spatial resolution with millisecond-scale temporal resolution, determined primarily by the US focal size and acquisition rate, addressing a major limitation of conventional electrophysiological and hemodynamic imaging techniques. However, several challenges remain. Evoked neuronal current densities in the human brain are weak (typically 0.01–0.5 mA/cm² under normal conditions) with a spatial resolution of 5 mm, requiring high sensitivity for reliable detection, and overlapping activity—such as ictal and background signals—demands high selectivity to accurately distinguish and localize sources at the voxel level. These constraints highlight the need for advanced signal processing strategies to improve both detection sensitivity and source separability in tABI. The three main objectives of my dissertation are to: 1) Improve tABI sensitivity and selectivity under seizure activity using data-driven methods, including Singular Value Decomposition (SVD) and Wiener filtering (WF), 2) Validate tABI in a human head model with a real skull for detecting and mapping weak, neuronal-like currents, and 3) Develop real-time, continuous tABI scanning as a translational step toward human application. Across all studies, the proposed framework significantly improved the sensitivity and selectivity of tABI in a human head model. Spatial Wiener filtering increased SNR depending on waveform type, with optimal performance at a ~5.2 mm kernel size, while preserving spatial resolution. Singular Value Decomposition (SVD) showed that the first 2–4 components captured >90% of the signal variance, enabling effective separation of overlapping dipole sources under both 200 Hz and broadband EEG stimulation. Detection limit analysis demonstrated reliable identification of weak currents down to ~0.1 mA, with reported sensitivity reaching ~71 µA for long-duration seizure-like waveforms. Corresponding current density sensitivity was on the order of ~400 µA/cm²·MPa. Real-time tABI achieved acquisition rates of up to ~400 cross-sections per second, enabling continuous imaging of a 9 s seizure waveform and multi-channel mapping across EEG configurations. Overall, these results demonstrate millimeter-scale (~1–4 mm) spatial resolution and millisecond-scale temporal resolution, confirming the capability of tABI to localize weak, dynamic neuronal-like currents in realistic transcranial conditions.