Comparison of semi-automated center-dot and fully automated endothelial cell analyses from specular microscopy images
Purpose To evaluate two specular microscopy analysis methods across different endothelial cell densities (ECDs). Methods Endothelial images of one eye from each of 45 patients were taken by using three different specular microscopes (three replicates each). To determine the consistency of the center...
Gespeichert in:
Veröffentlicht in: | International ophthalmology 2018-12, Vol.38 (6), p.2495-2507 |
---|---|
Hauptverfasser: | , , , , , , , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | Purpose
To evaluate two specular microscopy analysis methods across different endothelial cell densities (ECDs).
Methods
Endothelial images of one eye from each of 45 patients were taken by using three different specular microscopes (three replicates each). To determine the consistency of the center-dot method, we compared SP-6000 and SP-2000P images. CME-530 and SP-6000 images were compared to assess the consistency of the fully automated method. The SP-6000 images from the two methods were compared. Intraclass correlation coefficients (ICCs) for the three measurements were calculated, and parametric multiple comparisons tests and Bland–Altman analysis were performed.
Results
The ECD mean value was 2425 ± 883 (range 516–3707) cells/mm
2
. ICC values were > 0.9 for all three microscopes for ECD, but the coefficients of variation (CVs) were 0.3–0.6. For ECD measurements, Bland–Altman analysis revealed that the mean difference was 42 cells/mm
2
between the SP-2000P and SP-6000 for the center-dot method; 57 cells/mm
2
between the SP-6000 measurements from both methods; and −5 cells/mm
2
between the SP-6000 and CME-530 for the fully automated method (95% limits of agreement: − 201 to 284 cell/mm
2
, − 410 to 522 cells/mm
2
, and − 327 to 318 cells/mm
2
, respectively). For CV measurements, the mean differences were − 3, − 12, and 13% (95% limits of agreement − 18 to 11, − 26 to 2, and − 5 to 32%, respectively).
Conclusions
Despite using three replicate measurements, the precision of the center-dot method with the SP-2000P and SP-6000 software was only ± 10% for ECD data and was even worse for the fully automated method.
Clinical trial registration
Japan Clinical Trials Register (
http://www.umin.ac.jp/ctr/index/htm9
) number UMIN 000015236. |
---|---|
ISSN: | 0165-5701 1573-2630 |
DOI: | 10.1007/s10792-017-0760-7 |