No abstract

No abstract

Magnetic resonance imaging (MRI) was used to measure in vivo local strains in the optic nerve (ON) associated with horizontal duction in humans. Axial and coronal MRI were collected in target-controlled gazes in 24 eyes of 12 normal adults (six males and six females, 59 ± 16 years) during large (~28°) and moderate (~24°) ductions. The ON, globe, and extraocular muscles were manually identified, and the pixels were converted to point-sets that were registered across different imaging planes and eye positions. Shape of the ON was parameterized based on point-sets. Displacements and strains were computed by comparing deformed with initial ON configurations. Displacements were the largest in the most anterior region. However, strains from adduction were uniform along the length of the ON, while those during abduction increased with distance from the globe and were maximal near the orbital apex. For large gaze angles, ON strain during abduction was primarily due to bending near the orbital apex that is less transmitted to the eye, but during adduction the ON undergoes uniform stretching that transmits much greater loading to the posterior eye, implied by greater strain on the ON.

In this study, we investigated the performance of four deep learning frameworks of U-Net, U-NeXt, DeepLabV3+, and ConResNet in multi-class pixel-based segmentation of the extraocular muscles (EOMs) from coronal MRI. Performances of the four models were evaluated and compared with the standard F-measure-based metrics of intersection over union (IoU) and Dice, where the U-Net achieved the highest overall IoU and Dice scores of 0.77 and 0.85, respectively. Centroid distance offset between identified and ground truth EOM centroids was measured where U-Net and DeepLabV3+ achieved low offsets (p > 0.05) of 0.33 mm and 0.35 mm, respectively. Our results also demonstrated that segmentation accuracy varies in spatially different image planes. This study systematically compared factors that impact the variability of segmentation and morphometric accuracy of the deep learning models when applied to segmenting EOMs from MRI.

No abstract