Acquisition of high resolution three-dimensional ocular images at 7 Tesla to generate patient-specific eye-models for clinical ray-tracing
Jan-Willem Beenakker1, Lucia Hervella2, Juan Tabarnero2, Dennis Shamonin1, Andrew Webb1, Gregorius Luyten1, and Pablo Artal2

1Leiden University Medical Centre, Leiden, Netherlands, 2University of Murcia, Murcia, Spain

Synopsis

Patient-specific three-dimensional eye models obtained using very high resolution scans on a human 7T MRI system have been shown to form a much more accurate input for ray tracing algorithms than the current state-of-the-art generalized eye models used for clinical ophthalmology. Using a cued-blink protocol, custom-built phased array coil and segmentation software, accuracy of less than one-half dioptre can be achieved using the MRI data. These patient-specific models should provide much improved input for therapeutic procedures such as intra-ocular lens replacement for post-cataract surgery.

Purpose

Ray-tracing is a powerful technique for assessing the optical characteristics of the eye, and predicts the exact path of light rays as they pass through a geometric model of the eye. In the last few years, this method has seen an increased use in ophthalmology as it allows for a high degree of personalisation for optical treatments. With ray-tracing, for example, the actual shape of the cornea can be used to determine the optimal intra-ocular lens for patients post cataract surgery.1,2 However, current ray-tracing methods are only accurate along the optical axis, as there is no accurate method to determine the entire retinal shape. Along the optical axis the length of the eye can be measured by many different optical techniques. Off-axis distances, however, cannot be measured by these techniques, because refraction induces potentially significant systematic errors.3,4 Since MRI is not affected by refraction, it has the potential to measure the three-dimensional retinal shape, and therefore to improve the quality of ray-tracing.5,6 In this context, we have constructed personalized MRI-derived eye-models and compared their accuracy with the general eye-model of Escudero et al.7

Methods

Ocular MRI was performed on a Philips Achieva 7 Tesla whole body magnet. Volunteers were scanned with a custom-made dedicated receive eye-coil, in combination with a volume transmit coil (Nova Medical Inc., Wilmington, MA).5 Eye-motion artefacts were minimized by the use of a cued-blinking protocol and the MR-images were acquired using a 3D inversion recovery turbo gradient echo technique (TI=1280 ms). The total MRI examination takes less than 15 minutes. The resulting MR-images were processed using an automatic segmentation algorithm, which detects the retinal contour with sub-pixel accuracy.4 For each subject a personalized eye-model was built in OpticStudio (Zemax LCC, Kirkland, WA), similar to the approach of Canovas et al.1 The curvature of the cornea and the position of the lens were personalized with values from biometry measurements (Lenstar LS900, Haag-Streit AG). The shape of the retina was described either by the segmented MRI-data (MR-eye model) or as a half-sphere with a radius of 12 mm (general eye-model), figure 1. Finally, the back curvature of the eye-lens was adapted to reproduce the subjects’ central refraction and astigmatism. For each of the eyes, the peripheral aberrations, such as refraction and astigmatism, were measured using a custom-built peripheral wavefront scanner using a Harmann-Shack sensor.8

Results

The ray-tracing results, figure 2, show that the MRI based model is able to predict the peripheral aberrations within approximately half a Dioptre, a significant improvement compared to the general eye-model which has errors of up to 2 Dioptres.

Discussion

The MRI-based subject-specific eye-models allow for a more accurate description of the optical properties of the eye compared to conventional general eye-models. The MRI-based eye-model still uses an approximate description of the cornea, based on its local curvature. The accuracy of this eye-model can therefore be improved even further by incorporating the actual shape of the cornea, measured with Scheimpflug photography. A further area of improvement involves better assessment ofthe properties of the eye-lens, as the current model does not take the spatially varying index of refraction into account, but instead relies on an averaged index of refraction. The general eye-model fails to accurately describe many of the side-effects in current refractive surgery, as this model does not allow for a sufficient level of personalisation.9 The improved personalisation offered by the MRI has, therefore, great clinical potential for ophthalmology, not only to gain a better understanding of the patients’ condition, but also to test the efficacy of different treatments.

Conclusion

MRI-based eye-models provide an accurate description of the optical characteristics in patients’ vision over the entire eye and associated peripheral vision. This opens up many new applications in ophthalmology, which is currently hindered by the lack of an accurate tool to understand and treat conditions which occur outside the central field of view.

Acknowledgements

No acknowledgement found.

References

1.Canovas, C. & Artal, P. Customized eye models for determining optimized intraocular lenses power. Biomed. Opt. Express, BOE 2, 1649–1662 (2011).

2.Tabernero, J., Piers, P., Benito, A., Redondo, M. & Artal, P. Predicting the Optical Performance of Eyes Implanted with IOLs to Correct Spherical Aberration. Invest. Ophthalmol. Vis. Sci. 47, 4651–4658 (2006).

3.Beenakker, J. W. M. et al. Improved retinal shape detection using high-resolution MRI compared to partial coherence interferometry. Proc Intl Soc Mag Reson Img 23, 3146 (2015).

4.Atchison, D. A. & Charman, W. N. Can Partial Coherence Interferometry be Used to Determine Retinal Shape? Optometry Vision Sci 88, E601 (2011).

5.Beenakker, J.-W. M., Shamonin, D. P., Webb, A. G., Luyten, G. P. M. & Stoel, B. C. Automated retinal topographic maps measured with magnetic resonance imaging. Invest. Ophthalmol. Vis. Sci. 56, 1033–1039 (2015).

6.Atchison, D. A. et al. Eye Shape in Emmetropia and Myopia. Invest. Ophthalmol. Vis. Sci. 45, 3380–3386 (2004).

7. Escudero-Sanz, I. & Navarro, R. Off-axis aberrations of a wide-angle schematic eye model. J Opt Soc Am A Opt Image Sci Vis 16, 1881–1891 (1999).

8. Jaeken, B., Lundström, L. & Artal, P. Fast scanning peripheral wave-front sensor for the human eye. Opt. Express, OE 19, 7903–7913 (2011).

9.Holladay, J. T., Zhao, H. & Reisin, C. R. Negative dysphotopsia: the enigmatic penumbra. Journal of Cataract & Refractive Surgery 38, 1251–1265 (2012).

Figures

Figure 1.By combining shape of the cornea, the location and thickness of the lens, and the MRI-derived retinal profile, a patient-specific eye-model was constructed for ray-tracing simulations.

Figure 2. (left) One slice of a high resolution MR image of the eye acquired at 7 Tesla using a cued blink protocol and three-element phased array. (centre) Automatic detection of the retina. (right) Full three-dimensional segmentation of the retinal surface

Figure 3. (a) Schematic cross-section and (b) 3D-view of the ray-tracing setup, in which the retinal shape (orange) is taken from the MRI data

Figure 4. (a) Ray-traced peripheral defocus and astigmatism from one volunteer. (d) A comparison between the peripheral wavefront measurements and the standard Navarro eye-model and the eye-model with the MRI-based retina. The MRI-model better predicts the peripheral aberrations compared to the standard eye-model.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
0097