Abstract-In

Abstract-Hyperspectral imagery typically provides a wealth of information captured in a wide range of the electromagnetic spectrum for each pixel in the image; however, when used in statistical pattern-classification tasks, the resulting high-dimensional feature spaces often tend to result in ill-conditioned formulations. Popular dimensionality-reduction techniques such as principal component analysis, linear discriminant analysis, and their variants typically assume a Gaussian distribution. The quadratic maximumlikelihood classifier commonly employed for hyperspectral analysis also assumes single-Gaussian class-conditional distributions. Departing from this single-Gaussian assumption, a classification paradigm designed to exploit the rich statistical structure of the data is proposed. The proposed framework employs local Fisher's discriminant analysis to reduce the dimensionality of the data while preserving its multimodal structure, while a subsequent Gaussian mixture model or support vector machine provides effective classification of the reduced-dimension multimodal data. Experimental results on several different multiple-class hyperspectral-classification tasks demonstrate that the proposed approach significantly outperforms several traditional alternatives.Index Terms-Dimensionality reduction, Gaussian-mixturemodel (GMM), hyperspectral data, local discriminant analysis, support vector machine.

No abstract

Objective To evaluate the association of subretinal hyper-reflective material (SHRM) with visual acuity (VA), geographic atrophy (GA) and scar in the Comparison of Age related Macular Degeneration Treatments Trials (CATT) Design Prospective cohort study within a randomized clinical trial. Participants The 1185 participants in CATT. Methods Participants were randomly assigned to ranibizumab or bevacizumab treatment monthly or as-needed. Masked readers graded scar and GA on fundus photography and fluorescein angiography images, SHRM on time domain (TD) and spectral domain (SD) optical coherence tomography (OCT) throughout 104 weeks. Measurements of SHRM height and width in the fovea, within the center 1mm2, or outside the center 1mm2 were obtained on SD-OCT images at 56 (n=76) and 104 (n=66) weeks. VA was measured by certified examiners. Main Outcome Measures SHRM presence, location and size, and associations with VA, scar, and GA. Results Among all CATT participants, the percentage with SHRM at enrollment was 77%, decreasing to 68% at 4 weeks after treatment and 54% at 104 weeks. At 104 weeks, scar was present more often in eyes with persistent SHRM than eyes with SHRM that resolved (64% vs. 31%; p<0.0001). Among eyes with detailed evaluation of SHRM at weeks 56 (n=76) and 104 (n=66), mean [SE] VA letter score was 73.5 [2.8], 73.1 [3.4], 65.3 [3.5], and 63.9 [3.7] when SHRM was absent, present outside the central 1mm2, present within the central 1mm2 but not the foveal center, or present at the foveal center (p=0.02). SHRM was present at the foveal center in 43 (30%), within the central 1mm2 in 21 (15%) and outside the central 1mm2 in 19 (13%). When SHRM was present, the median maximum height in microns under the fovea, within the central 1 mm2 including the fovea and anywhere within the scan was 86; 120; and 122, respectively. VA was decreased with greater SHRM height and width (p<0.05). Conclusions SHRM is common in eyes with NVAMD and often persists after anti-VEGF treatment. At 2 years, eyes with scar were more likely to have SHRM than other eyes. Greater SHRM height and width were associated with worse VA. SHRM is an important morphological biomarker in eyes with NVAMD.

No abstract