Evaluation of the accuracy of grading indirect ophthalmoscopy video images for retinopathy of prematurity screening.

Abstract

To determine whether digital retinal images obtained using a video indirect ophthalmoscopy system can be accurately graded for zone, stage of retinopathy of prematurity (ROP), and plus or pre-plus disease, and used to screen for type 1 ROP. Charts of 114 infants who had retinal video images acquired using the video indirect ophthalmoscopy system during routine ROP examinations were retrospectively reviewed. Two masked ophthalmologists (1 expert and 1 non-expert in ROP screening) graded the video images for image quality, zone, stage of ROP, and pre-plus or plus disease. The ophthalmologists' grades of the videos were compared to the clinical examination results, which served as the reference standard. The sensitivity and specificity of two predefined criteria for referral in detecting disease requiring treatment (ie, type 1 ROP) were then determined. Of images the expert considered fair or good quality (n = 68), the expert and non-expert correctly identified zone (75% vs 74% of images, respectively), stage of ROP (75% vs 40% of images, respectively), and the presence of pre-plus or plus disease (79% of images). Expert and non-expert judgment of pre-threshold disease, pre-plus disease, or plus disease had 100% sensitivity and 75% versus 79% specificity, respectively, for detecting type 1 ROP. Expert and non-expert judgment of pre-plus or plus disease had 92% versus 100% sensitivity and 77% versus 82% specificity, respectively, for detecting type 1 ROP. High-quality retinal video images can be read with high sensitivity and acceptable specificity to screen for type 1 ROP. Grading for pre-plus or plus disease alone may be sufficient for the purpose of ROP screening.