Texture classification of retinal layers in optical coherence tomography

The purpose of this paper was to propose a new computer method for quantitative evaluation of representative features of the retina using optical coherence tomography (OCT). A multi-step approach was devised and positively tested for segmentation of the three main retinal layers: the vitreo-retinal interface and the inner and outer retina. Following a preprocessing step, three regions of interest were delimited. Significant peaks corresponding to high and low intensity strips were located along the OCT A-scan lines and accurate boundaries between different layers were obtained by maximizing an edge likelihood function. For a quantitative description, thickness measurement, densitometry, texture and curvature analyses were performed. As a first application, the effect of intravitreal injection of triamcinolone acetonide (IVTA) for the treatment of vitreo-retinal interface syndrome was evaluated. Almost all the parameters, measured on a set of 16 pathologic OCT images, were statistically different before and after IVTA injection (p < 0.05). Shape analysis of the internal limiting membrane confirmed the reduction of the pathological traction state. Other significant parameters, such as reflectivity and texture contrast, exhibited relevant changes both at the vitreo-retinal interface and in the inner retinal layers. Texture parameters in the inner and outer retinal layers significantly correlated with the visual acuity restoration. According to these findings an IVTA injection might be considered a possible alternative to surgery for selected patients. In conclusion, the proposed approach appeared to be a promising tool for the investigation of tissue changes produced by pathology and/or therapy.