Deep learning-based skull reconstruction and aberration correction method for transcranial ultrasound plane-wave imaging

The acoustic impedance mismatch between the skull and surrounding tissues can lead to a decrease in the quality of transcranial plane-wave imaging. A deep learning-based skull reconstruction technique was proposed, which combined fast marching method to achieve aberration correction in transcranial plane-wave imaging. In numerical simulations, an aberrator was set based on the CT image of the human skull, and results show that the time cost of the skull reconstruction is 0.97 s. The average error in the travel time of the plane-wave calculated based on the reconstruction result is 0.25%, while the average error in the travel time of the single point target scattered wave is 0.76%. In phantom experiments, aberration correction resulted in the reduction of the average position deviation of the point target image from 1.42 mm to 0.28 mm and the full-width-at-half-maximum from 1.82 mm to 0.99 mm. Additionally, the average contrast-to-noise ratio of circular targets was improved from 7.5 dB to 9.8 dB, and the eccentricity was simultaneously reduced from 0.31 to 0.24 after the correction. These results indicate that the proposed method can accurately calculate the travel time of transcranial ultrasound and significantly improve the quality of imaging, which holds the capability for high-quality plane-wave imaging of the brain through the skull, and exhibits the potential for application in transcranial Doppler imaging for cerebral blood flow.