Convolutional Neural Networks Enable Direct Strain Estimation in Quasistatic Optical Coherence Elastography.

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Computational Biology

Background:

• Assessing tissue biomechanical properties is vital for disease diagnosis and monitoring treatment efficacy.
• Optical coherence elastography (OCE) is a promising imaging technique for non-invasively measuring these properties.
• Current OCE data processing is often time-consuming, requires manual adjustments, and handles large datasets.

Purpose of the Study:

• To develop a faster and more efficient method for processing raw OCE data.
• To leverage machine learning, specifically convolutional neural networks (CNNs), to streamline the conversion of OCE phase data to tissue strain.
• To improve the speed of biomechanical information extraction from OCE acquisitions.

Main Methods:

• A novel convolutional neural network (CNN) was designed to directly process raw OCE phase data.
• The CNN model translates phase data into strain maps for quasistatic OCE applications.
• The computational approach bypasses several conventional, intermediate data processing steps.

Main Results:

• The CNN-based method achieved a processing speed approximately 40 times faster than the traditional least squares approach.
• The results demonstrate accurate strain calculation from raw OCE data.
• The method shows potential for significantly reducing the time required for OCE data analysis.

Conclusions:

• Machine learning, particularly CNNs, offers a powerful tool for enhancing OCE data analysis.
• This approach enables fast, efficient, and accurate extraction of biomechanical information.
• The developed method facilitates clinical translation and broader application of OCE technology.