May 2006
Volume 47, Issue 13
Free
ARVO Annual Meeting Abstract  |   May 2006
Identification of Different Phenotypes of Mild Non Proliferative Retinopathy of Type 2 Diabetes Using Cluster and Discriminant Mathematical Analysis
Author Affiliations & Notes
  • S. Nunes
    AIBILI – Assoc. for Innovation and Biomedical Research on Light anb Image, Coimbra, Portugal
  • R.C. Bernardes
    AIBILI – Assoc. for Innovation and Biomedical Research on Light anb Image, Coimbra, Portugal
  • L. Duarte
    AIBILI – Assoc. for Innovation and Biomedical Research on Light anb Image, Coimbra, Portugal
  • J. Cunha–Vaz
    AIBILI – Assoc. for Innovation and Biomedical Research on Light anb Image, Coimbra, Portugal
    Department of Ophthalmology, University Hospital
  • Footnotes
    Commercial Relationships  S. Nunes, None; R.C. Bernardes, None; L. Duarte, None; J. Cunha–Vaz, None.
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science May 2006, Vol.47, 1018. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      S. Nunes, R.C. Bernardes, L. Duarte, J. Cunha–Vaz; Identification of Different Phenotypes of Mild Non Proliferative Retinopathy of Type 2 Diabetes Using Cluster and Discriminant Mathematical Analysis . Invest. Ophthalmol. Vis. Sci. 2006;47(13):1018.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose: : To test the existence of different phenotypes of mild NPDR of type 2 diabetic patients using mathematical segmentation techniques (cluster and discriminant analysis) on clinical data from a two–year longitudinal study.

Methods: : Thirty–five eyes from 35 patients with diabetes type 2, under stabilized metabolic control, with mild NPDR, were followed during a 2–year time period performing systemic and ophthalmic examinations at 6 months intervals. The following measurements were obtained from field 2 in these eyes: retinal leakage (% of area with increased leakage – HRA, Heidelberg Engineering, Germany), accumulated number of red dots in field 2 (RD – in fundus photographs), retinal thickness (% of area of increased thickness – RTA, Talia, Israel) and retinal blood flow (mean retinal blood flow – HRF, Heidelberg Engineering, Germany). Each repeated measurement was considered as an independent one. First, Ward’s hierarchical clustering method was used to explore the number of clusters underlying the sample. Secondly, k–means clustering allowed to characterize the N selected number of clusters by means of reproducibility and stability of the clustering solutions. The reproducibility was assessed by discriminant analysis and the stability (over the 5 visits) by the changes in cluster membership for each patient.

Results: : Three potential best solutions were provided by Ward’s hierarchical clustering method (3, 4 and 5 clusters) with the 3 clusters solution achieving the best results (stability=68.6% and reproducibility=93.1%). A stability of 77.1% and reproducibility of 97.1% was achieved for the N=3 solution by the k–means clustering, thus supporting the existence of 3 distinct groups in the sample, i.e., 3 different phenotypes. Phenotype 1 is characterized by low values on 3 ophthalmic parameters, i.e., low leakage, low retinal thickness and low accumulated number of red dots. Phenotype 2 characterized predominantly by high leakage and phenotype 3 by a high accumulated number of red dots.

Conclusions: : Cluster analysis allowed to identify 3 different phenotypes in 35 eyes of 35 type 2 diabetic patients, under stabilized metabolic control, with mild NPDR for a 2–years period, with a reproducibility of 97.1% and a stability of 77.1%.

Keywords: diabetic retinopathy • retina • grouping and segmentation 
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×