Optical coherence tomography retinal image reconstruction via nonlocal weighted sparse representation

Abstract

We present a nonlocal weighted sparse representation (NWSR) method for reconstruction of retinal optical coherence tomography (OCT) images. To reconstruct a high signal-to-noise ratio and high-resolution OCT images, utilization of efficient denoising and interpolation algorithms are necessary, especially when the original data were subsampled during acquisition. However, the OCT images suffer from the presence of a high level of noise, which makes the estimation of sparse representations a difficult task. Thus, the proposed NWSR method merges sparse representations of multiple similar noisy and denoised patches to better estimate a sparse representation for each patch. First, the sparse representation of each patch is independently computed over an overcomplete dictionary, and then a nonlocal weighted sparse coefficient is computed by averaging representations of similar patches. Since the sparsity can reveal relevant information from noisy patches, combining noisy and denoised patches' representations is beneficial to obtain a more robust estimate of the unknown sparse representation. The denoised patches are obtained by applying an off-the-shelf image denoising method and our method provides an efficient way to exploit information from noisy and denoised patches' representations. The experimental results on denoising and interpolation of spectral domain OCT images demonstrated the effectiveness of the proposed NWSR method over existing state-of-the-art methods.

Axial Super-Resolution Study for Optical Coherence Tomography Images Via Deep Learning

“…5. As a comparison, the SR results of nonlocal weighted sparse representation (NWSR) [15], CNN-based method SRCNN [20], and the ordinary differential equation (ODE) Runge-Kutta (RK) methodinspired design network OISR-RK3 [39] are also shown in Fig. 5.…”

“…In 2013, L. Fang et al studied reconstruction of OCT images by sparse representation [13] and they further added a segmentation step in their methods when reconstructing images [14]. The nonlocal weighted sparse representation (NWSR) method [15] was presented to exploit information from noisy and denoised patches' representations to reconstruct images. Unfortunately, their improvement of resolution was not obvious and processing speed was not fast enough to be applied in real time.…”

Three-dimensional optical coherence tomography image denoising through multi-input fully-convolutional networks

“…In the future, we would like to incorporate segmentation information into the proposed MIFCN method [14]. Also, we would like to extend the proposed MIFCN method for OCT image interpolation [14,15]. In addition, although we only considered the task of retinal OCT image denoising, the proposed MIFCN method might also be applied to denoising of other medical images.…”

“…Some of the most successful modeling approaches consist of the Markov random field (MRF) [9], sparse representation [10], and Gaussian mixture models (GMM) [11]. Some of the recent studies where the patch-based sparse representation is applied to OCT image reconstruction consist of [12][13][14][15][16]. Recently, a variant of GMM [17] is applied to OCT image denoising with promising results [18].…”

“…In this paper, following [12][13][14][15] where they learned mappings from noisy images to high signal-to-noise (SNR) images using sparse representations, a specifically designed CNN is proposed for this purpose. To the best of the authors' knowledge, the proposed method is the first CNN-based OCT image denoising method.…”