Errors in Retinal Thickness Measurements Obtained by Optical Coherence Tomography
Received 14 June 2005; accepted 15 September 2005. published online 09 January 2006.
Objective
To report the frequency and severity of optical coherence tomography (OCT) retinal thickness measurement errors and to describe parameters that predict these errors.
Design
Observational case series.
Participants
Two hundred consecutive patients undergoing OCT imaging.
Methods
One eye (primary) from each of 200 consecutive patients undergoing Stratus OCT imaging (Carl Zeiss Meditec, Dublin, CA) with radial lines or fast macular thickness-based acquisition protocols was selected for review by 2 graders. On each of the line scans, graders evaluated the position of the automated retinal boundary lines (inner retinal surface and retinal pigment epithelium band) used by the OCT machine for thickness calculations and graded the positioning on a 6-point subjective, categorical error scale to generate an error score. The presence of thickness errors was correlated with various parameters, including the analysis confidence assessment reported by the OCT software, disease diagnosis, retinal morphologic features, the foveal center thickness standard deviation (FCTSD), and the FCTSD-to-foveal center thickness (FCT) ratio.
Main Outcome Measure
Average OCT retinal thickness error score.
Results
Errors of retinal boundary detection and thickness measurement were observed in 92% of eyes, but were severe in only 13.5% of eyes. The identification of an error or low analysis confidence by the OCT software was strongly associated with the severity of the retinal thickness errors. A higher FCTSD-to-FCT ratio and presence of subretinal fluid also were associated with more severe errors. Retinal cysts and a diagnosis of retinal vascular disease such as diabetic macular edema were less likely to be associated with significant errors.
Conclusions
Retinal thickness measurement errors occur frequently with current OCT segmentation and analysis algorithms. Severe errors are more frequent in eyes with subretinal pathologic features, but generally are detected by the OCT software. A high FCTSD-to-FCT ratio (>0.1) also may alert the clinician to the possibility of thickness errors. Clinical studies, particularly those pertaining to subretinal diseases, should consider these errors when incorporating OCT imaging in the study design.
1Doheny Image Reading Center, Doheny Eye Institute, Los Angeles, California
2Doheny Ophthalmic Imaging Unit, Doheny Eye Institute, Los Angeles, California
3Department of Preventive Medicine, Keck School of Medicine, University of Southern California, Los Angeles, California
Correspondence to Srinivas R. Sadda, MD, Doheny Eye Institute, DEI 3623, 1450 San Pablo Street, Los Angeles, CA 90033
Manuscript no. 2005-529.
Supported in part by the National Eye Institute, Bethesda, Maryland (grant no.: EY03040).
The authors have no proprietary interest in any of the topics discussed in the article.
ABSTRACT