Automatic segmentation of the posterior vitreous boundary in retinal optical coherence tomography

Abstract

 

Purpose

 

Disorders of the vitreomacular interface (VMI) such as vitreomacular traction and macular hole formation have recently been made accessible to pharmacologic treatment by the introduction of enzymatic vitreolysis. However, this therapeutic option is only efficacious in a subset of patients with strictly defined patterns of vitreous adhesions. Moreover, posterior vitreous detachment has been demonstrated to impact the efficacy of intravitreally administered antiangiogenic agents. Therefore, precise and reproducible quantification and classification of the posterior vitreous boundary and its adhesions at the macula is of major importance.<br /> The aim of this study was to develop a method to fully automatically segment the vitreous boundary in Spectral Domain - Optical Coherence Tomography (SD-OCT) scans.

 

Methods

 

A set of 61 macula-centered SD-OCT volume scans from patients available at the Vienna Reading Center was included. The posterior vitreous boundary was manually annotated in 333 B-scans (159/174 for training/testing).<br /> A segmentation method was developed based on the concept of a random forest trained on multiscale, rotation invariant eigenfeatures.<br /> The local image orientation is estimated using the second order central image moments and used to extract rotated windows around each voxel. A principal component analysis was performed on these windows and the resulting eigenvectors were used as features for a random forest classifier.<br /> The position of the vitreomacular interface was then detected by a modified A* algorithm using the predicted probability map given by the random forest as a cost function.

 

Results

 

Fully automated segmentation of the posterior vitreous boundary was feasible in all included cases. The mean signed distance between the expert annotation and the calculated vitreomacular interface was -6.38 pixels (95% interval [-106, 77]). Excluding areas where no expert annotation was present we find a mean signed error of -11.327798 [-109.0, 7.0].

 

Conclusions

 

A fully automated segmentation method for the detection of the posterior vitreous boundary in 3D SD-OCT was developed and validated. This method may be used to automatically determine patient eligibility for enzymatic vitreolysis and may facilitate further studies of the vitreomacular interface in large-scale antiangiogenic treatment trials.  

 

Fig. 1: top: expert annotation, middle: automated segmentation, bottom: raw B-scan