Spectrogram decomposition of ultrasonic guided waves for cortical thickness assessment using basis learning.

Area of Science:

• Biomedical Engineering
• Materials Science
• Acoustics

Background:

• Ultrasonic guided waves (UGWs) are valuable for bone characterization but are complicated by multimode and dispersive wave packets.
• Existing single-channel mode separation algorithms struggle with significantly overlapping modes in the time-frequency domain, limiting accuracy.
• Accurate separation of individual UGW modes and their dispersion curves is crucial for reliable bone property assessment.

Purpose of the Study:

• To develop an automated method for separating individual ultrasonic guided wave modes under severe overlapping and low signal-to-noise ratio (SNR) conditions.
• To utilize the extracted modes for accurate estimation of cortical bone thickness.
• To validate the proposed technique using both simulated and experimental bone data.

Main Methods:

• A spectrogram decomposition technique combining generalized separable nonnegative matrix factorization (GS-NMF) and adaptive basis learning was developed.
• The method focuses on automatic mode extraction from complex UGW signals.
• Extracted modes were employed for cortical thickness estimation.

Main Results:

• The proposed technique successfully separated multimodal UGWs even with severe overlapping and low SNR.
• Simulated data showed low relative errors (0.88%-1.43%) between extracted and theoretical dispersion curves.
• Cortical thickness estimation yielded low root-mean-square errors (0.039-0.052 mm) for bone samples.

Conclusions:

• The developed spectrogram decomposition technique effectively overcomes limitations of previous methods for UGW mode separation.
• The approach enables accurate bone characterization and cortical thickness assessment in challenging signal conditions.
• This advancement holds significant potential for non-destructive evaluation and medical diagnostics using UGWs.