Abstract
Purpose: :
To assess qualitative and quantitative changes in OCT data resulting from noise removal process.
Methods: :
An improved despeckling filter for OCT was recently proposed [1] aiming to differentiate between background (vitreous) and tissue (retina) while keeping the objective of preserving the interfaces. In the work herewith presented, we resort to synthetic OCT data, based on real OCT data scans, to quantify the filter performance in speckle noise removal and to determine the optimum diffusion time according to quantitative metrics. Additionally, real OCT data scans were filtered using a set of diffusion times (from 0.5 to 3.0 in 0.25 seconds steps) and assessed in a blind way by technicians (3) and clinicians (3). The selected cases represent a large spectrum of retinas status conditions including healthy references cases and retinas with diabetic retinopathy, age-related macular degeneration, choroidal neo-vascularization and cystoid macular edema. In addition, users were given the opportunity to select the exact same images (filtered and unfiltered) with or without gamma-correction (gamma=1.25).
Results: :
Despite the fact that quantitatively the best filter performance requires a diffusion time of 0.75 seconds, technicians elected as the best images the ones without any diffusion time (the original scans) in 91% of cases, while clinicians split themselves between the two options (47% originals and 53% filtered images). In the later case, the best diffusion time range between 0.5 and 1.0 seconds [(m±SD): 0.79 ± 0.17 seconds]. With respect of gamma-correction, technicians and clinicians opted for the gamma-corrected images in 58% and 82% of the cases, respectively.
Conclusions: :
The efficacy of the filter in removing speckle noise is clearly demonstrated based on quantitative metrics. When assessed qualitatively, technicians follow their training by electing the unfiltered images in 91% of the cases. On the other hand, clinicians found the filter useful in over half of the cases. Finally, both technicians and clinicians found the gamma-correction useful.[1] Bernardes R, Maduro C, Serranho P, Araújo A, Barbeiro S, Cunha-Vaz J. Improved adaptive complex diffusion despeckling filter. OPTICS EXPRESS 18(23):24048-24059, 2010.
Clinical Trial: :
http://www.clinicaltrials.gov NTC00797524
Keywords: imaging/image analysis: clinical • image processing • retina