Abstract
Purpose :
The measurement and analysis of retinal tissue biomechanics is important for understanding the pathophysiology of glaucoma. The dynamics of anatomical changes in the eye can uncover information that is not available in static Optical Coherence Tomography (OCT) images. However, noise in OCT images hampers the detailed and accurate analysis of retinal tissue movement. Here, we propose a new stabilizing and denoising method to improve the quality of retinal OCT videos, which exploits temporal redundancy.
Methods :
After spatial registration of the images, synchronization of the OCT acquisition with the measurement of the patient's pulse allows phase-wrapping each measurement using the heart frequency to average the frames of the video that correspond to the same instant in the cardiac cycle. This transforms a multi-cycle video into a single one-heart-cycle movie with reduced noise and enhanced visual clarity. A summary of this workflow is represented in figure 1.
Results :
Using the proposed workflow on the OCT videos of 15 patients (30 seconds at 100Hz), the CNR, SNR and the Mattes Mutual Information (MMI), which measure signal quality and registration accuracy, all increased significantly (Wilcoxon test, p ≤ 0.005). The CNR and SNR were computed in the RPE, the choroid and the RNFL, and showed a fold-change increase of respectively 1.70, 2.01 and 1.70 for the CNR and 1.63, 1.82 and 1.61 for the SNR. The average MMI showed a fold-change improvement of 3.51. Figure 2 illustrates an example of the application of our workflow on one subject.
Conclusions :
Videos of minimized noise allow local tissue displacement to be seen more clearly and analyzed more precisely than with standard image registration methods. Apart from the pulse signal hardware, all the necessary equipment is present in most ophthalmology clinics. As such, this technique has great potential for use in the clinical setting.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.