Purpose
Off-axis Spectral Domain Optical Coherence Tomography (SDOCT) acquisition can result in tilted images and erroneous thickness measurements. Using scans that were purposefully acquired off-axis, we developed an automated method for the assessment and correction of thickness errors based on angular deviation.
Methods
At least five volumetric SDOCT scans of the macula were obtained from one eye of 20 normal subjects. One scan was acquired through a pupil position such that the scan appeared “flat” in both the horizontal and vertical dimensions. Subsequent scans were acquired through multiple pupil positions, resulting in an apparent axial “tilt” along either the horizontal or vertical dimension. The inner limiting membrane (ILM) and the retinal pigment epithelium (RPE) were then segmented using an automated approach. The angular deviation from the “flat” optical axis was assessed by fitting a plane to the RPE surface. The multiple volumes were registered to each other and the average total retinal thickness was evaluated in 9 sectors (Fig. 1), where the diameters of the annuli are 1mm, 3mm and 5mm. The increase in thickness seen in the tilted volumes was modeled using a regression model that incorporated the thickness measurements as well as the angular deviation. The model was built using the measurements obtained from 10 subjects (5 OD, 5 OS) and subsequently tested using the volumes obtained from the remaining 10 subjects.
Results
The mean thickness error (Fig. 1A) noted between the off-axis scans and the correctly acquired “flat” scans was 7.91 +/- 4.31 µm. The linear regression model developed showed an adjusted R-square value of 0.986, and reduced the error (Fig. 1B) to 0.49 +/- 4.07 µm (p < 0.001). The model indicated a strong relationship between the cosine of the angular deviation and the increased thickness (p < 0.001).
Conclusions
The regression model developed was able to effectively reduce errors in thickness measurements derived from scans that were acquired off-axis. In large clinical trials using volumetric SDOCT, the improved accuracy possible using this method may lead to more reliable clinical end points.
Keywords: 549 image processing •
550 imaging/image analysis: clinical •
552 imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound)