ABSTRACT
Given that the neural and connective tissues of the optic nerve head (ONH) exhibit complex morphological changes with the development and progression of glaucoma, their simultaneous isolation from optical coherence tomography (OCT) images may be of great interest for the clinical diagnosis and management of this pathology. A deep learning algorithm (custom U-NET) was designed and trained to segment 6 ONH tissue layers by capturing both the local (tissue texture) and contextual information (spatial arrangement of tissues). The overall Dice coefficient (mean of all tissues) was 0.91 ± 0.05 when assessed against manual segmentations performed by an expert observer. Further, we automatically extracted six clinically relevant neural and connective tissue structural parameters from the segmented tissues. We offer here a robust segmentation framework that could also be extended to the 3D segmentation of the ONH tissues.
ABSTRACT
Purpose: To develop a deep learning approach to digitally stain optical coherence tomography (OCT) images of the optic nerve head (ONH). Methods: A horizontal B-scan was acquired through the center of the ONH using OCT (Spectralis) for one eye of each of 100 subjects (40 healthy and 60 glaucoma). All images were enhanced using adaptive compensation. A custom deep learning network was then designed and trained with the compensated images to digitally stain (i.e., highlight) six tissue layers of the ONH. The accuracy of our algorithm was assessed (against manual segmentations) using the dice coefficient, sensitivity, specificity, intersection over union (IU), and accuracy. We studied the effect of compensation, number of training images, and performance comparison between glaucoma and healthy subjects. Results: For images it had not yet assessed, our algorithm was able to digitally stain the retinal nerve fiber layer + prelamina, the RPE, all other retinal layers, the choroid, and the peripapillary sclera and lamina cribrosa. For all tissues, the dice coefficient, sensitivity, specificity, IU, and accuracy (mean) were 0.84 ± 0.03, 0.92 ± 0.03, 0.99 ± 0.00, 0.89 ± 0.03, and 0.94 ± 0.02, respectively. Our algorithm performed significantly better when compensated images were used for training (P < 0.001). Besides offering a good reliability, digital staining also performed well on OCT images of both glaucoma and healthy individuals. Conclusions: Our deep learning algorithm can simultaneously stain the neural and connective tissues of the ONH, offering a framework to automatically measure multiple key structural parameters of the ONH that may be critical to improve glaucoma management.
Subject(s)
Algorithms , Glaucoma/diagnosis , Machine Learning , Nerve Fibers/pathology , Optic Disk/pathology , Adult , Female , Humans , Male , Middle Aged , Reproducibility of Results , Retinal Ganglion Cells/pathology , Tomography, Optical Coherence/methods , Visual FieldsABSTRACT
PURPOSE: To map the 3-dimensional (3D) strain of the optic nerve head (ONH) in vivo after intraocular pressure (IOP) lowering by trabeculectomy (TE) and to establish associations between ONH strain and retinal sensitivity. DESIGN: Observational case series. PARTICIPANTS: Nine patients with primary open-angle glaucoma (POAG) and 3 normal controls. METHODS: The ONHs of 9 subjects with POAG (pre-TE IOP: 25.3±13.9 mmHg; post-TE IOP: 11.8±8.6 mmHg) were imaged (1 eye per subject) using optical coherence tomography (OCT) (Heidelberg Spectralis, Heidelberg Engineering GmbH, Heidelberg, Germany) before (<21 days) and after (<50 days) TE. The imaging protocol was repeated for 3 controls in whom IOP was not altered. In each post-TE OCT volume, 4 tissues were manually segmented (prelamina, choroid, sclera, and lamina cribrosa [LC]). For each ONH, a 3D tracking algorithm was applied to both post- and pre-TE OCT volumes to extract IOP-induced 3D displacements at segmented nodes. Displacements were filtered, smoothed, and processed to extract 3D strain relief (the amount of tissue deformation relieved after TE). Strain relief was compared with measures of retinal sensitivity from visual field testing. MAIN OUTCOME MEASURES: Three-dimensional ONH displacements and strain relief. RESULTS: On average, strain relief (averaged or effective component) in the glaucoma ONHs (8.6%) due to TE was higher than that measured in the normal controls (1.07%). We found no associations between the magnitude of IOP decrease and the LC strain relief (P > 0.05), suggesting biomechanical variability across subjects. The LC displaced posteriorly, anteriorly, or not at all. Furthermore, we found linear associations between retinal sensitivity and LC effective strain relief (P < 0.001; high strain relief associated with low retinal sensitivity). CONCLUSIONS: We demonstrate that ONH displacements and strains can be measured in vivo and that TE can relieve ONH strains. Our data suggest a wide variability in ONH biomechanics in the subjects examined in this study. We further demonstrate associations between LC effective strain relief and retinal sensitivity.