Optical eye imaging techniques are widely available for examining anterior and retinal structures but are limited in 3D assessments of the whole eyeball. MRI is the preferred modality for these regions but fine eye structures are difficult to resolve on clinical systems. Ultra high field magnets offer superior signal-to-noise, providing increased resolution, but there have been only a limited number of studies1,2,3. We performed a preliminary assessment of achievable resolution, anatomy visible on differing image weightings and MR parameter measurements, in eyes of healthy subjects at 7 Tesla.
7 volunteers were imaged using a 7 Tesla Siemens scanner and 6-channel eye coil3. Subjects performed a visual attention task to maintain eye fixation. Select subjects underwent higher resolution scanning of one eye taped comfortably to reduce susceptibility artefact2,4,5. 2D T2-w fast spin echo sequence (TEeff/TR/ETL/FA/NSA/TAcq: 11ms/5.4sec/18/170°/1/2mins) 0.26mm in-plane, 0.7mm slices and 3D gradient-echo (GRE) images were acquired (TE/TR/FA/GRAPPA/NSA/TAcq:10ms/4ms/12°/2/1/2or4mins) 0.2x0.2x0.4 or 0.9mm. One subject had GRE imaging at multiple flip angles (5-30°). Estimated T1 maps were calculated.
The T2-w protocol produced detailed images, particularly revealing the subarachnoid space cerebrospinal fluid surrounding the optic nerve. Acquisition of multiple flip-angle GREs allowed an assessment of SNR, CNR and T1 estimates for a number of structures (typical values: mean lens nucleus T1=1775ms, vitreous humour 3225ms, choroid-retina 1567ms). Imaging a taped eye and effective fixation, allowed high-resolution T1-w images (voxel volume 0.016mm3) with few artefacts and high SNR (whole-eye SNR=37). Structures were readily segmented.
This work demonstrates high resolution eye images on both T2 and T1-w sequences, with excellent anatomical detail including: lens, ciliary body, optic nerve, ophthalmic vessels and arachnoid space. For the GRE sequence, we achieved a voxel volume of 0.016mm3in vivo – lower than previous studies2,3 – which readily allowed structural segmentation. Further optimisation may allow higher resolution, potentially increasing sensitivity to pathology.