Adaptive complex diffusion noise despeckling for 3D OCT Data

Bernardes, Rui; Maduro, Cristina; Serranho, Pedro; Dinis, João; Cunha-Vaz, José

http://hdl.handle.net/10400.2/2806

Use this identifier to reference this record.

Name:	Description:	Size:	Format:
AbstractRB.pdf		103.18 KB	Adobe PDF	Download

Send Feedback

Authors

Abstract(s)

Purpose: To demonstrate the performance of a recently proposed despeckling filter when extended to 3D in OCT data. Methods: A new formulation for a complex diffusion filter was recently proposed [1] being adaptive in time and adjusting parameters to data, facilitating diffusion in the vitreous and reducing it in the retina to preserve tissue information. This new formulation outperformed, both quantitatively and qualitatively, currently existing filters, while at the same time was computationally more efficient attaining the same despeckling level in 34% of the computing time. We have now extended it to perform 3D OCT despeckling achieving a significant improvement in noise removal. We have resorted to a mathematical based synthetic OCT scan in order to assess quantitative results in 3D. In addition, we have applied this 3D filter to a set of 72 eye scans, from healthy volunteers (20), diabetic retinopathy (20), cystoid macular edema (2), age-related macular degeneration (20) and choroidal neo-vascularization (10), who underwent high-definition Cirrus OCT (Carl Zeiss Meditec, Dublin, CA, USA) using macular cube protocols (200x200x1024 and/or 512x128x1024). Results: The extension from 2D to 3D of this adaptive complex diffusion filter proved to be beneficial by achieving an increased level of noise reduction while simultaneously better preserving fundamental information. These facts are supported from well known metrics (e.g. MSE-mean squared error, ENL-effective number of looks and CNR-contrast-to-noise ratio) for the synthetic OCT scan and from the assessment made by 3 retina specialist who graded qualitatively the output of the 2D and 3D filters. Conclusions: These findings demonstrate this new formulation and its extension to 3D is beneficial for human analysis of OCT data and suggest it might be an important tool for automated data processing as in segmentation of retinal structures. [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.