Abstract
Purpose: :
To evaluate the performance of RTVue-100 retinal imaging in normal and pathological eyes.
Methods: :
The performance evaluation was conducted on the RTVue-100 MM5/EMM5 scans from three datasets: 1) DB1: 126 EMM5 scans from 42 eyes, in which 3 eyes have hard drusen, 8 eyes have soft drusen, 5 eyes have PED, 3 eyes have GA, 7 eyes have CNV, 8 eyes have CSME, 1 eye has DME and 7 eyes are normal. One EMM5 scan was chosen for each eye to be graded by a well-trained examiner who manually drew the retinal layer boundaries (manual segmentation) of ILM and RPE, in order to determine the segmentation algorithm accuracy; 2) DB2: 354 MM5 scans from 118 normal eyes, 3 scans each; 3) DB3: 144 EMM5 scans from 48 eyes from 3 different RTVue-100 machines by different operators. For each machine, 4 normal subjects and 4 patients with retinal diseases were enrolled and both eyes had 3 EMM5 scans each. The ETDRS 9-sector FRT measurements were investigated. The accuracy was computed from DB1 by comparing the automated segmentation results with the hand-drawn segmentation. The repeatability was computed for each individual dataset, DB1, DB2 and DB3. The reproducibility was computed from DB3.
Results: :
In DB1, the averaged accuracy (mean error ± standard deviation) of the FRT for the 9-sector regions was 1.24±3.62 µm; and the averaged repeatability over the 9-sector regions for 7 normal eyes was 3.07 µm, for the 35 pathological eyes was 4.17 µm. In DB2, the averaged repeatability over the 9-sector regions for the 118 normal eyes was 3.65 µm. In DB3, the averaged repeatability and reproducibility over the 9-sector regions for the 24 normal eyes were 3.03 µm and 6.84 µm, and for the 24 pathological eyes were 3.97 µm and 3.97 µm respectively.
Conclusions: :
The performance of RTVue-100 retinal imaging was found to be excellent for both normal and pathological eyes. Segmentation accuracy was shown to be within 2 µm for normals and retina patients. Repeatability was also excellent and ranged between 3.07 µm and 4.17 µm from three large datasets. Reproducibility was also very good suggesting the RTVue can be used to monitor changes over time.
Keywords: age-related macular degeneration • imaging/image analysis: clinical • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound)