Transcranial Phase Correction Using Pulse-Echo Ultrasound and Deep Learning: A 2-D Numerical Study.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Artificial Intelligence

Background:

• Transcranial ultrasound imaging quality is degraded by phase aberration from human skulls.
• Accurate skull profile and speed of sound (SOS) are crucial for aberration correction.
• Physics-based estimation of skull properties is challenging due to complex ultrasound-skull interactions.

Purpose of the Study:

• To develop and validate a deep learning approach for estimating skull profile and SOS from ultrasound RF signals.
• To assess the feasibility of using deep learning for transcranial phase aberration correction.

Main Methods:

• Constructed realistic numerical skull models from CT scans.
• Generated array-based ultrasound backscattered signals via acoustic simulations.
• Developed and trained a deep learning model to estimate skull thickness and SOS from RF data.

Main Results:

• The deep learning model achieved high accuracy: 0.15 mm MAE for thickness (2% error) and 13 m/s MAE for SOS (0.5% error).
• Pearson correlation coefficients were 0.99 for thickness and 0.95 for SOS.
• Aberration correction using estimated parameters significantly improved beam focusing and transcranial imaging quality.

Conclusions:

• The proposed deep learning approach is feasible for accurate skull property estimation.
• This method offers a viable solution for transcranial phase aberration correction.
• The findings pave the way for enhanced clinical applications of transcranial ultrasound.