Transcranial Ultrasound Imaging With Decomposition Descent Learning-Based Full Waveform Inversion

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Machine Learning

Background:

• Noninvasive brain diagnosis is crucial for predicting neurological conditions like stroke.
• Current 3-D quantitative ultrasonic brain imaging faces challenges due to complex anatomy and skull interference.
• Existing methods struggle with accuracy and computational efficiency.

Purpose of the Study:

• To develop a novel method for high-quality, quantitative 3-D ultrasonic imaging of the brain.
• To overcome limitations in current brain imaging techniques, particularly for detecting internal anomalies.
• To improve the efficiency and accuracy of brain imaging for clinical applications.

Main Methods:

• Proposed a decomposition descent learning-based full waveform inversion (DDL-FWI) technique.
• Utilized a linear residual decomposing technique to reduce computational load.
• Incorporated residual functions to enhance convergence in fast inversion tomography (FIT).

Main Results:

• Achieved high-quality quantitative imaging of brain tissues and skulls, even from a homogeneous background.
• Successfully reconstructed complex brain tissues and clots in both 2-D and 3-D with noisy and clean signals.
• Demonstrated robust 3-D clot resolution down to 18 mm and rapid 2-D reconstruction in 11.20 seconds.

Conclusions:

• DDL-FWI provides an efficient and accurate solution for noninvasive 3-D brain imaging.
• The method offers a safe alternative to traditional imaging techniques, avoiding ionizing radiation.
• Advanced ultrasonic hardware combined with DDL-FWI facilitates practical and efficient clinical brain imaging.