A deep-learning model for one-shot transcranial ultrasound simulation and phase aberration correction

Comprehensive Summary

In their research article, Naftchi-Ardebili et al. investigate the efficiency of the fast and precise simulation and error correction of phase aberration in transcranial ultrasound for the use of focused ultrasound therapies. They proposed the use of the TUSNet algorithm for the simulation of the two-dimensional focal pressure distribution and the phases of error correction for the ultrasound waves that travel through the human brain using CT scan maps. The deep network was trained with more than 180,000 simulated cranial CT scans and validated on 1,232 actual cranial CT scans, and its performance was measured and compared with the physics-based simulation tool k-Wave. TUSNet performed the simulation at a mean speed of 21 ms/slice. This illustrated that the computation times were 1,200 times faster than k-Wave and the absolute errors of the focal pressure were small. The researchers elaborated on the capabilities of deep learning-based simulation for focusing ultrasound waves, pointing out the need for a detailed experimental investigation of the appropriateness and accuracy of the approach presented.

Outcomes and Implications

This article is important since there is a need for accurate and fast ultrasound focusing for the safe and effective use of transcranial ultrasound therapies such as neuromodulation, opening the blood-brain barrier, and targeted drug delivery. With the increase in speed of computations performed, it was possible to apply the method for treatment planning for transcranial ultrasound therapy. This could be used within the current and upcoming ultrasound clinical trials for neurological disorders such as epilepsy, Alzheimer's disease, and brain tumors. Although the proposed model has not been validated to work in a real-world clinical setting, it may prove helpful in the future after validation and in analyzing the simulation in three-dimensional space.