Retinal shape assessment in young phakic eyes by a combined partial coherence interferometry - magnetic resonance imaging approach

Abstract

Purpose: The current study proposes a comparative partial coherence interferometry (PCI) - magnetic resonance imaging (MRI) quantitative assessment of the retinal shape enabling an individualized description of the image plane usable for constructing ray-tracing model eyes valid for both on and off-axis evaluations.

Methods: The on and off-axis axial lengths of 16 left eyes of healthy volunteers were measured using a Lenstar LS900 (Haag-Streit AG) PCI. The on-axis axial length varied between 22.59 and 25.99 mm with an average of 23.74 mm. The measurements were performed along the 0 degrees horizontal meridian and within a -30 to +30 degrees field angle range with an angular step of 10 degrees. To enable the off-axis measurements, a wide-angle LED gazing system was fabricated and fixed to the interferometer head. With the addition of the measured anterior corneal geometry on-axis, an algorithm was created to geometrically reconstruct the retinal shape. Subsequently, the retinal shape was quantified by a parameterized elliptical fit algorithm using a linear least square method with a trace constraint. Concurrently, the volunteers left eyes were scanned with a custom-made dedicated 3-channel receive eye-coil combined with a volume transmit coil (Nova Medical) in a Philips Achieva 7 T whole body MR scanner. The retinal shape was segmented from the acquired images by the means of an automatic algorithm.

Results: For the entire data set, the PCI fitted vertex radius varied between 10.31 and 12.26 mm. When compared on-axis, there is a strong correlation (R² = 0.9923) between the PCI reconstructed and MRI segmented centre of the lens-retina distances with a systematic error of 0.083 mm and a reproducibility of 0.132 mm. The 95% confidence interval limits are between -0.176 and 0.341 mm. These values are inferior to the spatial resolution of the MRI (0.5 x 0.5 x 1.0 mm³). For 30 degrees field of view, the correlation coefficient is R²= 0.9424, the systematic error is 0.160 mm and the reproducibility 0.244 mm. The 95% confidence interval is between -0.319 and 0.639 mm.

Conclusions: The comparative analysis between the PCI and MRI reconstructed retinal shapes demonstrated a strong correlation for both on and off-axis values thus enabling the construction of individual eye models with an anatomical image plane geometry.