Pseudoaveraging for denoising of OCT angiography: a deep learning approach for image quality enhancement in healthy and diabetic eyes

Abstract

Background: This study aimed to develop a deep learning (DL) algorithm that enhances the quality of a single-frame enface OCTA scan to make it comparable to 4-frame averaged scan without the need for the repeated acquisitions required for averaging.

Methods: Each of the healthy eyes and eyes from diabetic subjects that were prospectively enrolled in this cross-sectional study underwent four repeated 6 × 6 mm macular scans (PLEX Elite 9000 SS-OCT), and the repeated scans of each eye were co-registered to produce 4-frame averages. This prospective dataset of original (single-frame) enface scans and their corresponding averaged scans was divided into a training dataset and a validation dataset. In the training dataset, a DL algorithm (named pseudoaveraging) was trained using original scans as input and 4-frame averages as target. In the validation dataset, the pseudoaveraging algorithm was applied to single-frame scans to produce pseudoaveraged scans, and the single-frame and its corresponding averaged and pseudoaveraged scans were all qualitatively compared. In a separate retrospectively collected dataset of single-frame scans from eyes of diabetic subjects, the DL algorithm was applied, and the produced pseudoaveraged scan was qualitatively compared against its corresponding original.

Results: This study included 39 eyes that comprised the prospective dataset (split into 5 eyes for training and 34 eyes for validating the DL algorithm), and 105 eyes that comprised the retrospective test dataset. Of the total 144 study eyes, 58% had any level of diabetic retinopathy (with and without diabetic macular edema), and the rest were from healthy eyes or eyes of diabetic subjects but without diabetic retinopathy and without macular edema. Grading results in the validation dataset showed that the pseudoaveraged enface scan ranked best in overall scan quality, background noise reduction, and visibility of microaneurysms (p < 0.05). Averaged scan ranked best for motion artifact reduction (p < 0.05). Grading results in the test dataset showed that pseudoaveraging resulted in enhanced small vessels, reduction of background noise, and motion artifact in 100%, 82%, and 98% of scans, respectively. Rates of false-positive/-negative perfusion were zero.

Conclusion: Pseudoaveraging is a feasible DL approach to more efficiently improve enface OCTA scan quality without introducing notable image artifacts.

Keywords: Averaging; Deep learning; Denoising; Diabetic retinopathy; Image artifact; Image quality; OCTA; Optical coherence tomography angiography; Pseudoaveraging.

Fig. 1

Schematic diagram illustrating the workflow of this study. The 39 prospectively collected eyes underwent 4 repeated 6 × 6 mm macular scans on the Zeiss PLEX Elite SS-OCT, and the repetitions were co-registered to generate an averaged volume. This dataset was split into a training dataset (5 eyes) and validation dataset (34 eyes). In the training dataset, the single-frame enface OCTA scan was used as an input and it was paired with its corresponding enface OCTA 4-frame average as a target to train the denoising or pseudoaveraging algorithm. Additional training of the algorithm (not shown here) was done by using 8 low-quality single-frame enface OCTA scans as an input and high-quality single-frame OCTA scan of the same eye as an output (peer-to-peer training). For the validation dataset, the pseudoaveraging algorithm was applied to the best-quality single-frame enface OCTA scan (out of the 4 repeated single-frame acquisitions), and the resulting pseudoaveraged scans were compared against their corresponding original single-frame scans as well as against their 4-frame averaged scans. In a separate test dataset of 105 eyes retrospectively collected, the algorithm was applied to single-frame enface OCTA scans, and the original and its corresponding pseudoaveraged enface scans were compared

Fig. 2

Example of a single (top), an averaged (middle), and a pseudoaveraged (bottom) full-thickness retinal enface OCTA scan in a healthy eye. It is noted how the overall scan quality, small vessels continuity, and background noise reduction are best in the pseudoaveraged scan. It is noted how the averaged scan is the best for motion (line) artifact correction;s however, the small vessels appear blurry in this scan (insets)

Fig. 3

Left panel: 6 × 6 mm enface OCTA scan demonstrating partial attenuation of FAZ contour in the pseudoaveraged scan at about 6 o’clock (middle) compared to the original scan (top). However, after adjusting the brightness/contrast histogram in image J, the vessel is no longer as attenuated and FAZ contour becomes more visible (bottom image) indicating that FAZ contour attenuation is at least partially due to inadequate brightness/contrast adjustment and not only due to pseudoaveraging. Right panel: cropped 6 × 6 mm enface OCTA scan from an eye with diabetic retinopathy. The arrows in the bottom image point to subtle residual FAZ noise in the pseudoaveraged scan (bottom scan)