| Title |
|---|
| Automated Posterior Scleral Tomography Assessment for Staphyloma Visualization with Improved Maculopathy Correlation |
| Authors |
|---|
| Xi Chen, Xiao Fei Wang |
| Presenting |
|---|
| Xi Chen |
| PURPOSE: |
|---|
| Subjective morphological assessment of posterior staphyloma is limited in reproducibility and scalability, and its relationship with myopic macular degeneration (MMD) remains unclear. We developed an automated, quantitative approach to characterize posterior scleral shape from MRI. |
| METHODS: |
|---|
| Using deep learning, we automatically segmented the eyeball and extracted surface points from MRI scans. In a 3D Cartesian coordinate system, we computed the distance (D) from each posterior scleral point to a hypothetical pre-elongation eye center and the local curvature (C). We derived indices (Dmean, Cmean, Dmax, Cmax, Dvar, and C·D), evaluated concordance with conventional staphyloma classification, and tested associations with MMD grade across three types: type-0 (nasal–temporal symmetry with elongated posterior globe without apparent protrusion), type-1 (nasal–temporal symmetry with elongated posterior globe and conical protrusion), and type-2 (nasal or temporal protrusion with elongated posterior globe). |
| RESULTS: |
|---|
| We analyzed 102 eyes from 52 participants. Automated segmentation achieved a Dice coefficient of 0.935. For pairwise discrimination, C·Dmax best separated type-0 vs type-1 (AUC 0.923) and Dvar best separated type-1 vs type-2 (AUC 0.981); Dmax, C·Dmean, and C·Dmax also differentiated types. For three-class discrimination, C·Dmax showed the best overall performance, and all indices achieved AUCs >0.900. After adjustment for age and axial length, Dvar (β=0.75 mm−2, p=0.02) and C·Dmax (β=2.44, p=0.02) were significantly associated with higher MMD grades. |
| CONCLUSIONS: |
|---|
| Automated MRI-based posterior scleral tomography enables scalable visualization and quantitative grading of staphyloma location and severity. |