Abstract
Purpose :
To evaluate the performance of a novel AI algorithm for precise segmentation of the choroidal thickness in Optical Coherence Tomography (OCT) macular scans in eyes with pathologic myopia.
Methods :
91 pathological myopia eyes were acquired using a commercial swept-source OCT (SS-OCT) system, (PLEX Elite 9000, Carl Zeiss Meditec Inc., Dublin, CA, USA) at a 1050 nm wavelength, scanning speed of 100,000 A-scans/sec and 3 mm × 3 mm scanning protocol, centred at the macula. Manual segmentation of the retina and choroid on OCT images were used as the ground truth. We implemented a novel multi-task deep convolutional neural network architecture, Spatial Aggregated Networks (SA-Net), that reconstructs and segments a target B-scan with the incorporation of spatial context from neighbouring B-scans. Intersection over Union (IoU) of the volumetric segmentation, Dice coefficient and Structural Similarity Index Measure (SSIM) were used to assess performance.
Results :
A total of 91 eyes with pathologic myopia were analysed with spherical equivalent of -7.00 ± 3.96 and axial length of 27.54 ± 1.30. Subjects were aged 51.07 ±13.02 with 64 (70.32%) females. 94.51% were Chinese, 4.40% were Malay and the remaining were Others. SA-Net was able to replicate segmentation of the anatomical layers of the retina and choroid on OCT images that was comparable to that of the manually segmented ground truths, with an IoU of 0.87 ± 0.09, Dice coefficient of 0.93 ± 0.05 and SSIM of 0.60 ± 0.18.
Conclusions :
Our study demonstrated that the novel SA-Net approach showed a high accuracy in segmentation and delineation of choroid from volumetric OCT cube scans in eyes with pathologic myopia. The results are promising for the automated detection of the choroid and could be beneficial in further studies pertaining to the choroid in eyes with pathologic myopia.
This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.