Àá½Ã¸¸ ±â´Ù·Á ÁÖ¼¼¿ä. ·ÎµùÁßÀÔ´Ï´Ù.
KMID : 0213520190330020103
Korean Journal of Ophthalmology
2019 Volume.33 No. 2 p.103 ~ p.112
Clinical Efficacy of Custom-built Software for the Early Detection of Glaucoma: A Comparison of Axial-length and Major Retinal Artery Location Data
Jang Hyo-Ju

Lee Sang-Min
Ahn Jae-Hong
Rho Seung-Soo
Abstract
Purpose: To assess the clinical efficacy for early detection of glaucoma using custom-built image software visualizing translucent retinal nerve fiber layer thickness (RNFLT) that is graphed based on a normative database.

Methods: This prospective study was conducted using a normative database constructed with RNFLT data of 151 healthy Korean eyes. The reference lines of the mean, the lower 5%, and the lower 1% limit were visualized as a translucent RNFLT graph produced by our software after inputting each subject's major retinal artery position and overlaying the results onto the RNFLT measurements. Fifty-eight additional healthy control and 79 early-glaucoma eyes were collected for the validation group. If a subject's RNFLT graph was outside the reference line of the lower 1% limit, the graph was defined as abnormal. The lower 1% limit, which was generated by three criteria (criterion 1, built-in software; criterion 2, axial-length data; criterion 3, major retinal artery data), was used to address the difference of agreement with a standard answer.

Results: For criteria 1, 2, and 3, the accuracy of our custom-built software was significantly higher than that of the manufacturer's database (kappa of 0.475 vs. 0.852 vs. 0.940; sensitivity of 62.0% vs. 91.1% vs. 97.5%, respectively) maintaining high specificity (87.9% vs. 94.8% vs. 96.6%, respectively).

Conclusions: The custom-built imaging software with the constructed RNFLT normative database showed high clinical efficiency for early detection of glaucoma with negligible user-related variability.
KEYWORD
Axial length, Database, Glaucoma, Major retinal artery position, Optical coherence tomography
FullTexts / Linksout information
  
Listed journal information
MEDLINE ÇмúÁøÈïÀç´Ü(KCI) KoreaMed ´ëÇÑÀÇÇÐȸ ȸ¿ø