Prediction of 3D Strain Field in Posterior Sclera Based on Attention 3D U-Net

To analyze the biomechanical properties of the posterior sclera, we denoised OCT images and segmented them using the nnU-Net model. After obtaining the segmented tissues, we performed registration and calculated the deformation of the posterior eye tissues using mathematical methods. Finally, we combined the baseline images and the force-loaded images and trained an Attention 3D U-Net model to predict the deformation of the posterior sclera, i.e., to predict its 3D strain field. The model performance was evaluated using the mean absolute error (MAE), which was below 0.2.