Purpose:
Commercially available automated mosaic software can be used to stitch together fundus photographs for the evaluation of retinal diseases. While these programs usually perform well when combining photos of healthy retinas, their performance in eyes with severe retinal disease such as CMV retinitis is unclear. In this study, we tested the relative quality of 3 different commercially available software programs using fundus images acquired from a larger study of telemedicine screening for CMV retinitis.
Methods:
Photographs were taken from 99 eyes affected by CMV retinitis using a Topcon TRC-NW 6S digital fundus camera at Maharaj Nakorn Chiang Mai Hospital in Chiang Mai, Thailand. The individual images from each eye were loaded onto three different programs: Topcon IMAGENet, I2k Retina, and Ophthalmic Imaging System AutoMontage (OIS). Two masked graders assessed the overall image quality, using a 5-point scoring system, with 5 being the best score. Graders also documented image artifacts in each mosaic, including vessel breakage, ghost vessels (duplicated vessels), misplaced images and blurriness.
Results:
The mean score for overall quality of the montaged images obtained from IMAGENet, i2k, and OIS were 2.7 (CI 2.4-2.8), 4.6 (CI 4.5-4.8), and 4.5 (CI 4.3-4.6), respectively. There was a significant difference in score among the three programs tested (p<0.0001). In pairwise comparisons, IMAGEnet had a lower quality score than i2k (p<0.0001) or OIS (p<0.0001). IMAGENet had a higher percentage of ghost vessels (49% vs. 5.1% and 3.5%), misplaced images (55.6% vs. 4.5% and 4.0%), vessel breakage (31.8% vs. 7.6% and 28.8%) and increased image blurriness (15.2% vs. 4.5% and 2.5%) compared to i2k and OIS, respectively. I2k had a significantly lower frequency of vessel breakage than OIS or IMAGEnet (p<.0001).
Conclusions:
Both i2k and OIS software packages appear to create similar high quality photographic mosaics from digital images of eyes with CMV retinitis. Furthermore, both programs create mosaics that are of higher quality than those created by IMAGENet. Images auto-montaged by i2K and OIS software may prove valuable in both diagnosing CMV retinitis and following disease progression.
Keywords: imaging/image analysis: non-clinical