Deep learning ocular aberration retrieval from simulated retinal images under straylight.

Area of Science:

• Ophthalmology
• Optical Engineering
• Computational Imaging

Background:

• The human eye's Point Spread Function (PSF) is influenced by optical aberrations and straylight.
• Distinguishing between aberrations and straylight in PSF images is difficult with current methods.
• This limitation hinders accurate clinical assessment of ocular optical quality.

Purpose of the Study:

• To develop a deep learning model capable of retrieving wavefront aberrations from simulated PSF images containing straylight.
• To effectively separate the contributions of optical aberrations and straylight to the overall PSF.

Main Methods:

• Simulated PSF images were generated with incorporated straylight effects, modeled as random phase perturbations.
• A deep learning approach was employed to analyze these simulated PSFs.
• The model was trained to predict the underlying wavefront aberrations.

Main Results:

• The deep learning model achieved high accuracy in predicting wavefront aberrations.
• The method demonstrated rapid, one-shot inference in as little as 3 milliseconds.
• Successful separation of aberration and scatter components from simulated PSF data was achieved.

Conclusions:

• The proposed deep learning method offers a novel solution for disentangling optical aberrations and straylight in ocular imaging.
• This technique has the potential to significantly enhance the clinical assessment of the human eye's optical performance.
• Enables a more comprehensive evaluation of ocular optical quality using standard PSF measurements.