ADC-Net: An Open-Source Deep Learning Network for Automated Dispersion Compensation in Optical Coherence Tomography

Area of Science:

• Biomedical Optics
• Medical Imaging
• Deep Learning Applications

Background:

• Chromatic dispersion degrades image resolution in Optical Coherence Tomography (OCT).
• Accurate dispersion compensation is crucial for enhancing OCT system performance and diagnostic capabilities.

Purpose of the Study:

• To develop and validate a deep learning network, termed Automated Dispersion Compensation Network (ADC-Net), for automated chromatic dispersion compensation in OCT.
• To optimize the ADC-Net architecture and training parameters for robust performance.

Main Methods:

• A modified UNet architecture was employed, utilizing an encoder-decoder pipeline for OCT B-scan processing.
• The network was trained using partially compensated OCT B-scans as input and aiming for fully compensated scans as output.
• Performance was objectively assessed using Peak Signal to Noise Ratio (PSNR) and Multi-Scale Structural Similarity Index Metric (MS-SSIM).

Main Results:

• The ADC-Net achieved optimal performance with five input channels, demonstrating robust dispersion compensation.
• Objective assessment yielded high performance metrics: 29.95 ± 2.52 dB for PSNR and 0.97 ± 0.014 for MS-SSIM.
• Comparative analysis confirmed the superiority of the five-input channel model for ADC-Net training.

Conclusions:

• The developed ADC-Net effectively compensates for chromatic dispersion in OCT, significantly improving image quality.
• The deep learning approach offers an automated and efficient solution for enhancing OCT resolution.
• The findings highlight the potential of ADC-Net for advancing OCT imaging in clinical and research settings.