Quantification of Blood Flow Velocity in the Human Conjunctival Microvessels Using Deep Learning-Based Stabilization...

Area of Science:

• Ophthalmology
• Biomedical Engineering
• Medical Imaging

Background:

• Accurate quantification of blood flow velocity in the human conjunctiva is crucial for understanding microvascular hemodynamics.
• Eye motion during imaging introduces artifacts, compromising segmentation and velocity measurement accuracy.

Purpose of the Study:

• To introduce a novel optical imaging system with deep learning for conjunctival blood flow velocity measurement.
• To address motion artifacts in conjunctival imaging for enhanced accuracy.

Main Methods:

• Developed a customized optical imaging system incorporating deep learning-based segmentation (Attention-UNet) and a two-step motion correction technique (registration and template matching).
• Attention-UNet was employed for robust segmentation of conjunctival images affected by motion blur.
• Motion correction reduced image displacement to 4-7 μm (registration) and <1 μm (template matching).

Main Results:

• The novel system effectively corrects for motion artifacts in conjunctival imaging.
• Achieved significant reduction in image displacement, enabling more accurate blood flow velocity calculations.
• Blood flow velocity was calculated using corrected images, considering temporal signal variances and vessel lengths.

Conclusions:

• The developed deep learning-based system significantly improves the accuracy of conjunctival blood flow velocity quantification.
• This approach offers valuable insights for microvascular hemodynamics studies in the conjunctiva and other tissues.
• The methods provide a robust solution for resolving motion artifacts in sensitive biomedical imaging applications.