High-resolution MR imaging of the human eye 2005

We studied preparatory strategies for high-resolution human eye in vivo imaging with commercially available magnets and coils.

We imaged normal volunteers on 1.5T systems by Philips, GE, and Siemens, using commercial approximately 9 cm temporomandibular joint receive coils. Subjects fixated the nonimaged eye on a target. We compared signal (S) to tissue noise (Nt) and system noise (Ns) between images acquired with the imaged eye: 1) open, 2) held closed, 3) taped closed, and 4) taped closed with overlying water-soaked gauze. We standardized technique 4 to compare studies between manufacturers using T1-weighted parameters (repetition time/echo time/flip angle/number of signal averages = 400 ms/10-17 ms/90 degrees /4-6, in-plane resolution approximately 250 x 250 microm2; 2-3 mm slices, image time = 4.3-5.2 min). We obtained similar images of an enucleated human eye to estimate in vivo effects of microsaccades and ocular pulsations.

Measurements of S/Nt and S/Ns gave surprising results of Nt < Ns in some instances. Ns/Nt was congruent with 1, varying approximately 20%, when the eye was taped shut and covered with water-soaked gauze. T1-weighted spin echo sequences, using technique 4, produced high-quality images with good S/Nt on all systems. Images from the three manufacturers were comparable when parameters were normalized for pulse repetition time, echo time, number of signal averages, bandwidth in Hz/pixel, and reconstruction matrix. Images of the enucleated eye produced S/Nt ratios that were comparable to in vivo images in some structures.

Our best preparatory technique produced images with in-plane resolution of approximately 250 mum in 4-5 minutes with three commercial 1.5 T systems. The in vivo S/Nt was comparable to in vitro values in most solid tissues but decreased in most fluid compartments.