Acoustoelectric Time-Reversal for Ultrasound Phase-Aberration Correction.

Acoustoelectric imaging (AEI) is a technique that combines ultrasound (US) with radio frequency recording to detect and map local current source densities. This study demonstrates a new method called acoustoelectric time reversal (AETR), which uses AEI of a small current source to correct for phase aberrations through a skull or other US-aberrating layers with applications to brain imaging and therapy. Simulations conducted at three different US frequencies (0.5, 1.5, and 2.5 MHz) were performed through media layered with different sound speeds and geometries to induce aberrations of the US beam. Time delays of the acoustoelectric (AE) signal from a monopole within the medium were calculated for each element to enable corrections using AETR. Uncorrected aberrated beam profiles were compared with those after applying AETR corrections, which demonstrated a strong recovery (29%-100%) of lateral resolution and increases in focal pressure up to 283%. To further demonstrate the practical feasibility of AETR, we further conducted bench-top experiments using a 2.5 MHz linear US array to perform AETR through 3-D-printed aberrating objects. These experiments restored lost lateral restoration up to 100% for the different aberrators and increased focal pressure up to 230% after applying AETR corrections. Cumulatively, these results highlight AETR as a powerful tool for correcting focal aberrations in the presence of a local current source with applications to AEI, US imaging, neuromodulation, and therapy.

High resolution transcranial acoustoelectric imaging of current densities from a directional deep brain stimulator.

OBJECTIVE: New innovations in deep brain stimulation (DBS) enable directional current steering-allowing more precise electrical stimulation of the targeted brain structures for Parkinson's disease, essential tremor and other neurological disorders. While intra-operative navigation through MRI or CT approaches millimeter accuracy for placing the DBS leads, no existing modality provides feedback of the currents as they spread from the contacts through the brain tissue. In this study, we investigate transcranial acoustoelectric imaging (tAEI) as a new modality to non-invasively image and characterize current produced from a directional DBS lead. tAEI uses ultrasound (US) to modulate tissue resistivity to generate detectable voltage signals proportional to the local currents. APPROACH: An 8-channel directional DBS lead (Infinity 6172ANS, Abbott Inc) was inserted inside three adult human skulls submerged in 0.9% NaCl. A 2.5 MHz linear array delivered US pulses through the transtemporal window and focused near the contacts on the lead, while a custom amplifier and acquisition system recorded the acoustoelectric (AE) interaction used to generate images. MAIN RESULTS: tAEI detected monopolar current with stimulation pulses as short as 100 µs with an SNR ranging from 10-27 dB when using safe US pressure (mechanical indices <0.78) and injected current of ~2 mA peak amplitude. Adjacent contacts were discernable along the length and within each ring of the lead with a mean radial separation between contacts of 2.10 and 1.34 mm, respectively. SIGNIFICANCE: These results demonstrate the feasibility of tAEI for high resolution mapping of directional DBS currents using clinically-relevant stimulation parameters. This new modality may improve the accuracy for placing the DBS leads, guide calibration and programming, and monitor long-term performance of DBS for treatment of Parkinson's disease.

Automated detection of geological landforms on Mars using Convolutional Neural Networks.

The large volume of high-resolution images acquired by the Mars Reconnaissance Orbiter has opened a new frontier for developing automated approaches to detecting landforms on the surface of Mars. However, most landform classifiers focus on crater detection, which represents only one of many geological landforms of scientific interest. In this work, we use Convolutional Neural Networks (ConvNets) to detect both volcanic rootless cones and transverse aeolian ridges. Our system, named MarsNet, consists of five networks, each of which is trained to detect landforms of different sizes. We compare our detection algorithm with a widely used method for image recognition, Support Vector Machines (SVMs) using Histogram of Oriented Gradients (HOG) features. We show that ConvNets can detect a wide range of landforms and has better accuracy and recall in testing data than traditional classifiers based on SVMs.