Discriminating Between Papilledema and Optic Disc Drusen Using 3D Structural Analysis of the Optic Nerve Head

Abstract

Background and Objectives:The distinction of papilledema from other optic nerve head (ONH) lesions mimicking papilledema, such as optic disc drusen (ODD), can be difficult in clinical practice. We aimed: (1) To develop a deep learning algorithm to automatically identify major structures of the optic nerve head (ONH) in three dimensional (3D) optical coherence tomography (OCT) scans; (2) To exploit such information to robustly differentiate among ODD, papilledema, and healthy ONHs.Methods:This was a cross-sectional comparative study of patients from three sites (Singapore, Denmark, Australia) with confirmed ODD, papilledema due to raised intracranial pressure, and healthy controls. Raster scans of the ONH were acquired using OCT imaging, then processed to improve deep-tissue visibility. First, a deep learning algorithm was developed to identify major ONH tissues, and ODD regions. The performance of our algorithm was assessed using the Dice coefficient. Second, a classification algorithm (random forest) was designed to perform 3-class classifications (1: ODD, 2: papilledema, 3: healthy) strictly from their drusen and prelamina swelling scores (calculated from the segmentations). To assess performance, we reported the area under the receiver operating characteristic curve (AUC) for each class.Results:A total of 241 patients (256 imaged ONHs, including 105 ODD, 51 papilledema, and 100 healthy ONHs) were retrospectively included in this study. Using OCT images of the ONH, our segmentation algorithm was able to isolate neural and connective tissues, and ODD regions/conglomerates whenever present. This was confirmed by an averaged Dice coefficient of 0.93±0.03 on the test set, corresponding to good segmentation performance. Classification was achieved with high AUCs, i.e. 0.99±0.001 for the detection of ODD, 0.99±0.005 for the detection of papilledema, and 0.98±0.01 for the detection of healthy ONHs.Discussion:Our AI approach can discriminate ODD from papilledema, strictly using a single OCT scan of the ONH. Our classification performance was very good in the studied population, with the caveat that validation in a much larger population is warranted. Our approach may have the potential to establish OCT imaging as one of the mainstays of diagnostic imaging for ONH disorders in neuro-ophthalmology, in addition to fundus photography.

Early Detection of Optic Nerve Changes on Optical Coherence Tomography Using Deep Learning for Risk-Stratification of Papilledema and Glaucoma

“…For papilledema, a larger scale study has been conducted to show effective use of deep learning in differentiating normal disc, papilledema, and other disc abnormalities. [5][6][7] In glaucoma, multiple successful attempts have been made to differentiate normal nerves from glaucomatous nerves using parameters such as mean RNFL and fundus photos. [8][9][10] However, none of the above attempts worked with prediagnostic or early disease information, assessing the ability of AI to predict the eventual development of papilledema or glaucoma before the diagnosis was confidently made by the physicians.…”

Peripapillary Vessel Density in Relation to Optic Disc Drusen: A Multimodal Optical Coherence Tomography Study

“…These assumptions make the method susceptible to low inter- and intrarater reliability. Deep learning looks promising as a tool to overcome these limitations and replace manual segmentation in the future (20).…”

Recent advances in the use of optical coherence tomography in neuro‐ophthalmology: A review

“…Within ophthalmology, deep learning has shown robust performance such as in the detection of diabetic retinopathy, 110 glaucoma 111 and cataract 112 . In neuro‐ophthalmology, deep learning models have shown excellent performance in: (1) differentiating optic neuropathies from pseudopapilledema, 113,114 (2) differentiating demyelinating optic neuritis and non‐arteritic anterior ischemic optic neuropathy, 115 (3) differentiating glaucomatous optic neuropathy from non‐glaucomatous optic neuropathy, 116 all from optical coherence tomography or clinical fundal photographs. Future research can explore deep learning models on other investigation modalities such as OCT 114 and OCTA as well as multimodal imaging with clinical parameters to better diagnose and prognosticate neuro‐ophthalmological conditions.…”