| Abstract [eng] |
This master„s work explains algorithm for optic nerve detection and its structural analysis. Most common eye visualisation methods are shortly reviewed. These methods include ultrasonography, computed tomography (CT), magnetic resonance imaging (MRI) and optical coherence tomography (OCT). Ultrasound imaging was chosen for being cheap, short and portable method to visualise optic nerve and its surrounding tissues. 19 in total, real images expressing drusen, edema and drusen-edema cases were examined. Images came from Eye Clinics of Lithuania University of Health Sciences Hospital, acquired using OTI-Scan 1000 ultrasound scanner using 10 MHz frequency. Developed algorithm detected optic nerve zone, retinal surface finally measuring optic nerve diameter and few other parameters. All measurements are summarized in discriminant analysis. 26 total parameters (21 normalised histogram counts, mean, variance, skewness, kurtosis and entropy) for every image are analysed. 3 main pathologies were investigated – drusen, edema and combined drusen and edema cases. Discriminant analysis selected only certain histogram counts and variance as most significant for canonical function. In the end, territorial map was found, showing no overlapping zones, to determine which of 3 investigated cases was found in image. |
