Deep-learning-based approach for strain estimation in phase-sensitive optical coherence elastography.

Area of Science:

• Biomedical Optics
• Medical Imaging
• Machine Learning

Background:

• Phase-sensitive optical coherence elastography (OCE) is crucial for biomechanical property assessment.
• Accurate strain field estimation is vital for reliable OCE analysis.
• Current methods, like the vector method, face limitations in performance and signal-to-noise ratio.

Purpose of the Study:

• To propose a deep-learning-based approach for enhanced strain field estimation in OCE.
• To validate the proposed method using simulated and experimental data with varying deformations.
• To compare the performance of the deep learning method against the conventional vector method.

Main Methods:

• A convolution neural network (CNN) was trained using simulated wrapped phase and phase-gradient maps.
• The trained CNN was applied to measured phase-difference maps for strain field calculation.
• Two specimens, exhibiting homogeneous and heterogeneous deformations, were used for experimental validation.

Main Results:

• The deep learning approach demonstrated superior performance in estimating strain field distributions.
• Signal-to-noise ratio (SNR) of strain results was enhanced by 21.6 dB compared to the vector method.
• The method accurately characterized both homogeneous and heterogeneous deformation patterns.

Conclusions:

• The proposed deep-learning-based method offers a significant advancement for strain estimation in OCE.
• This technique provides more reliable and higher-quality strain imaging for biomedical applications.
• The enhanced SNR and accuracy make it a promising tool for quantitative elastography.