VALIDATION OF LOCAL AND GLOBAL TORTUOSITY QUANTIFICATIONS IN RETINAL BLOOD VESSELS

• Main Author: DWI NAFIA - NIM: 23215319 , TESI
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/24506
• Institution: Institut Teknologi Bandung

Tortuosity is a parameter that describes the increase of twists, dilations or tangles in a blood vessel segment. Tortuosity in retinal blood vessel indicate the beginning of some vascular diseases. Digital images processing can detect tortuosity automatically and deliver a quantitative value termed tortuosity index. Several possible quantification index of tortuosity have been proposed, but none of them has gained universal acceptance. This study aims to investigate the response of various tortuosity quantifying methods on perceptual changes in tortuosity properties and get the best quantifying method. The steps of tortuosity quantification are data acquisition, preprocessing to obtain vessels centerline coordinate, and calculation of various quantifying method to get local and global tortuosity indexes from retinal image. From this study, Tortuosity Density (TD) method is consistently proportional to frequency, amplitude and length of the turn curve, invariant to vessel orientation and not too dependent on scale in a blood vessels segment. Tortuosity Density (TD) is the best method for quantifying local tortuosity, with high positive correlation with the clinical order as shown by spearman’s rank correlation coefficient of 0.95 (p-value< 0.05) in arterial segments and 0.85 (p-value < 0.05) in venous segments. Tortuosity Density Global (TDG) weighting scheme combinated with Tortuosity Density (TD) method and Weighted Additivity Chord (WCA) weighting scheme combinated with Derivative Curvature Integral (DCI) are the best methods for quantifying global tortuosity, with high positive correlation with the clinical order as shown by spearman’s rank correlation coefficient of 0.97 (p-value< 0.05).