Development and quantitative assessment of deep learning-based image enhancement for optical coherence tomography

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Optical coherence tomography (OCT) is crucial for retinal imaging.
• Improving OCT image quality is essential for accurate diagnosis.
• Traditional methods like image averaging have limitations.

Purpose of the Study:

• To develop and evaluate a deep learning framework for enhancing OCT image quality.
• To compare the deep learning method against traditional image averaging.
• To assess the clinical utility of the enhanced OCT images.

Main Methods:

• A deep learning framework with high-resolution representation was developed.
• 359 normal eyes and 456 eyes with retinal conditions were analyzed.
• Quantitative comparisons included subjective ophthalmologist scores and objective metrics (SSIM, PSNR, CNR).

Main Results:

• The deep learning method consistently outperformed image averaging in SSIM and PSNR values.
• Enhanced tissue contrast (CNR) was observed with the deep learning method using fewer frames.
• Subjective image quality scores were highest for the deep learning method across all conditions.

Conclusions:

• The proposed deep learning framework offers superior OCT image enhancement.
• It provides a better trade-off between image quality and scanning time.
• This approach can potentially reduce the number of repeated OCT scans.