Three-Dimensional Bruch’s Membrane Shape Model in Cases of Papilledema

Abstract

Purpose : Our previous work (Wang et al., ARVO 2015; SPIE 2015) demonstrated that a 2D statistical shape model of Bruch’s membrane (BM) can quantify the anterior displacement of BM towards the vitreous in papilledema, providing a potential non-invasive measure of raised intracranial pressure. However, these studies were limited to a 2D analysis because the input images (i.e. HD five-line raster scans) lack 3D contextual information. In this work, a semi-automated method was developed to extend the previous 2D shape measure to 3D and to correlate the 3D shape measure with the 2D shape measure.

Methods : The proposed algorithm automatically transferred two manual landmarks at BM opening (BMO) in the HD raster central B-scan into the corresponding volumetric SD-OCT image domain. The BMO contour was first segmented in the retinal pigment epithelium (RPE) en-face image using a graph-based algorithm. Then, along the segmented BMO contour from 0^o to 360^o, eight equidistant landmarks were automatically placed covering 1.5 mm in the radial direction for every 10^o (Fig. 1). Next, principal component analysis was used to generate the 3D BM shape model, and the first principal component coefficient, which reflects the magnitude of anterior/posterior displacement of the BM layer, was extracted as the “3D shape measure.”

Results : We used 116 baseline right eyes with both SD-OCT image protocols from the IIHTT OCT sub-study. The 2D and 3D BM shape measures were both computed. Fig. 2 shows the 2D and 3D BM model shape changes with the parameter (c) varying from -3 to 3. The correlation coefficient between 2D and 3D BM shape measures was 0.80 (p-value < 0.01).

Conclusions : BM shape measures in 2D and 3D are potentially new parameters to monitor changes in the ONH due to raised intracranial pressure. Because the 3D shape model uses significantly more landmarks than the 2D shape model (2D: 20 landmarks; 3D: 288 landmarks) and is able to incorporate 3D contextual information, it has promising advantages, which will be validated in ongoing studies. Future work includes developing a fully automated method and establishing relationships between 2D/3D BM shape measures and other clinical measures over time, including intracranial pressure (ICP), and visual field sensitivity.

This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.

 

Fig.1 Example of 3D landmark placement

 

Fig.2 Comparison between 2D and 3D BM shape measures