University of Illinois Chicago
Browse

Differential Artery-Vein Analysis in OCT Angiography of Diabetic Retinopathy

Download (2.74 MB)
thesis
posted on 2021-05-01, 00:00 authored by Mattia Castelnuovo
Given the rising prevalence of Diabetes and thus Diabetic Retinopathy in the world, developing better techniques for early detection is paramount. This study aims to build upon previous studies that used geometric quantitative features in OCTA images to classify Non-Proliferative Diabetic Retinopathy (NPDR). Six geometric features, three angle-related (VBA, CBA1 and CBA2) and three width-related (VBC, VWR1 and VWR2), were automatically calculated in four distinct types of analyses made possible by the classification of OCTA vessel structures in arteries and veins using a clustering technique built around four features derived from the same vessels in the relative OCT images. Arteries’ and veins’ features were calculated separately, their ratios were computed and then their separate measurements were put together in a “Weighted Average” method. Comparative analysis of healthy patients, patients with diabetes but no retinopathy (NoDR) and patients with various stages of NPDR was conducted. This study found that sensitivity increased in width-related measurements and decreased in angle-related ones compared with the previous study. Some explanations are proposed as causes of this difference. Also, it was observed that the ratio measurements between artery values and vein values do not seem to contribute to the classification of patients and data seem to suggest that arteries and veins may be affected in circa the same way by NPDR with respect to the analyzed six features. VBC and VWR1 showed very good ability to separate between Control and NoDR patients from the NPDR patients even with the most conservative Bonferroni Correction.

History

Advisor

Yao, Xincheng

Chair

Yao, Xincheng

Department

Bioengineering

Degree Grantor

University of Illinois at Chicago

Degree Level

  • Masters

Degree name

MS, Master of Science

Committee Member

Lee, James Caiani, Enrico Gianluca

Submitted date

May 2021

Thesis type

application/pdf

Language

  • en

Usage metrics

    Categories

    No categories selected

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC