Quality factor based correction for SD-OCT measurements in cataract patients

AI-based clinical assessment of optic nerve head robustness superseding biomechanical testing

BACKGROUND/AIMS

To use artificial intelligence (AI) to: (1) exploit biomechanical knowledge of the optic nerve head (ONH) from a relatively large population; (2) assess ONH robustness (ie, sensitivity of the ONH to changes in intraocular pressure (IOP)) from a single optical coherence tomography (OCT) volume scan of the ONH without the need for biomechanical testing and (3) identify what critical three-dimensional (3D) structural features dictate ONH robustness.

METHODS

316 subjects had their ONHs imaged with OCT before and after acute IOP elevation through ophthalmo-dynamometry. IOP-induced lamina cribrosa (LC) deformations were then mapped in 3D and used to classify ONHs. Those with an average effective LC strain superior to 4% were considered fragile, while those with a strain inferior to 4% robust. Learning from these data, we compared three AI algorithms to predict ONH robustness strictly from a baseline (undeformed) OCT volume: (1) a random forest classifier; (2) an autoencoder and (3) a dynamic graph convolutional neural network (DGCNN). The latter algorithm also allowed us to identify what critical 3D structural features make a given ONH robust.

RESULTS

All three methods were able to predict ONH robustness from a single OCT volume scan alone and without the need to perform biomechanical testing. The DGCNN (area under the curve (AUC): 0.76±0.08) outperformed the autoencoder (AUC: 0.72±0.09) and the random forest classifier (AUC: 0.69±0.05). Interestingly, to assess ONH robustness, the DGCNN mainly used information from the scleral canal and the LC insertion sites.

CONCLUSIONS

We propose an AI-driven approach that can assess the robustness of a given ONH solely from a single OCT volume scan of the ONH, and without the need to perform biomechanical testing. Longitudinal studies should establish whether ONH robustness could help us identify fast visual field loss progressors.

PRECIS

Using geometric deep learning, we can assess optic nerve head robustness (ie, sensitivity to a change in IOP) from a standard OCT scan that might help to identify fast visual field loss progressors.

Parapapillary gamma zone associated with increased peripapillary scleral bowing: the Beijing Eye Study 2011

AIMS

To investigate the association between the backward configuration of the peripapillary sclera (PPS), measured as PPS angle (PPSA), and presence and extent of parapapillary gamma zone.

METHODS

Out of the population-based Beijing Eye Study 2011, we randomly selected individuals free of optic nerve and retinal diseases. With Spectralis optical coherence tomography, we measured gamma zone (zone free of Bruch's membrane (BM)) and determined the PPSA, defined as the angle between the anterior scleral surface lines from both sides of the optic nerve head (ONH).

RESULTS

The study included 678 individuals with age of 59.5±7.6 years (range: 50-90) and axial length of 23.5±1.3 mm (20.9-29.2). Gamma zone was more prevalent in eyes with larger PPSA (p=0.006) after adjustment for axial length (p<0.001) and BM opening area (p<0.001). Gamma zone width was positively associated with PPSA, axial length and BM opening area (all p<0.001) in multivariable analysis. Circular gamma zone was accompanied with larger PPSA as compared with focal gamma zone (19.9°±7.2° vs 6.3°±5.3°, p<0.001). Focal temporal gamma and focal inferior gamma had similar mean PPSA (p=0.69). However, the horizontal PPSA was significantly larger than the vertical PPSA in inferior gamma (6.9°±6.3° vs 4.7°±6.6°; p=0.005), while they were comparable in temporal gamma (6.1°±5.8° vs 6.3°±6.4°; p=0.073).

CONCLUSIONS

A more backward bowing of the PPS was linearly and spatially associated with the presence, size and extent of gamma zone. It suggested that the BM and the sclera were closely related in participating the biomechanical behaviour of the ONH.

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

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 the following: (1) to develop a deep learning algorithm to automatically identify major structures of the ONH in 3-dimensional (3D) optical coherence tomography (OCT) scans and (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 3 sites (Singapore, Denmark, and Australia) with confirmed ODD, those with papilledema due to raised intracranial pressure, and healthy controls. Raster scans of the ONH were acquired using OCT imaging and 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, and 3: healthy ONHs) 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 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, that is, 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 artificial intelligence 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.

Use of multimodal dataset in AI for detecting glaucoma based on fundus photographs assessed with OCT: focus group study on high prevalence of myopia

BACKGROUND

Glaucoma is one of the major causes of blindness; it is estimated that over 110 million people will be affected by glaucoma worldwide by 2040. Research on glaucoma detection using deep learning technology has been increasing, but the diagnosis of glaucoma in a large population with high incidence of myopia remains a challenge. This study aimed to provide a decision support system for the automatic detection of glaucoma using fundus images, which can be applied for general screening, especially in areas of high incidence of myopia.

METHODS

A total of 1,155 fundus images were acquired from 667 individuals with a mean axial length of 25.60 ± 2.0 mm at the National Taiwan University Hospital, Hsinchu Br. These images were graded based on the findings of complete ophthalmology examinations, visual field test, and optical coherence tomography into three groups: normal (N, n = 596), pre-perimetric glaucoma (PPG, n = 66), and glaucoma (G, n = 493), and divided into a training-validation (N: 476, PPG: 55, G: 373) and test (N: 120, PPG: 11, G: 120) sets. A multimodal model with the Xception model as image feature extraction and machine learning algorithms [random forest (RF), support vector machine (SVM), dense neural network (DNN), and others] was applied.

RESULTS

The Xception model classified the N, PPG, and G groups with 93.9% of the micro-average area under the receiver operating characteristic curve (AUROC) with tenfold cross-validation. Although normal and glaucoma sensitivity can reach 93.51% and 86.13% respectively, the PPG sensitivity was only 30.27%. The AUROC increased to 96.4% in the N + PPG and G groups. The multimodal model with the N + PPG and G groups showed that the AUROCs of RF, SVM, and DNN were 99.56%, 99.59%, and 99.10%, respectively; The N and PPG + G groups had less than 1% difference. The test set showed an overall 3%-5% less AUROC than the validation results.

CONCLUSION

The multimodal model had good AUROC while detecting glaucoma in a population with high incidence of myopia. The model shows the potential for general automatic screening and telemedicine, especially in Asia.

TRIAL REGISTRATION

The study was approved by the Institutional Review Board of the National Taiwan University Hospital, Hsinchu Branch (no. NTUHHCB 108-025-E).

Histology of neovascular myopic macular degeneration

To assess the histological correlate of neovascular or exudative myopic macular degeneration (nMMD) in highly myopic human eyes, we examined histomorphometrically histologic sections of enucleated eyes of Caucasian patients. The study included 284 eyes (age: 61.9 ± 13.7 years; range: 24–89 years; axial length: 25.5 ± 3.1 mm; range: 20–37 mm). An nMMD was detected in 5 eyes (axial length: 29.6 ± 2.6 mm; range: 26.0–31.0 mm). All these eyes showed within or close to the nMMD a macular Bruch’s membrane (BM) defect, fibrous tissue with erythrocyte-filled blood vessels, and proliferations of irregularly pigmented and irregularly piled-up retinal pigment epithelium (RPE) cells each of which was connected with a curled-up, PAS (Periodic-Acid-Shiff)-positive membrane. The nMMD lesions were covered by proliferated RPE cells. RPE cells were not detected within the retina. In binary regression analysis, a higher nMMD prevalence was associated with a higher prevalence of macular BM defects (odds ratio: > 1000; P < 0.001), while the association with axial length was not significant (P = 0.43) in that model. After adjustment for the presence of macular BM defects, the nMMD prevalence was not associated with BM thickness (measured at the posterior pole, equator-posterior pole midpoint, equator and shortly posterior to the ora serrata) (P = 0.10; P = 0.87; P = 0.40; and P = 0.36, respectively), RPE cell layer thickness (P = 0.83; P = 0.79; P = 0.31; P = 0.38, resp.), RPE cell density (P = 0.56; P = 0.91; P = 0.47; P = 0.87, resp.), choriocapillaris thickness (P = 0.47; P = 0.93; P = 0.41; P = 0.75, resp.), and choriocapillaris density (P = 0.99; P = 0.94; P = 0.17; P = 0.97, resp.). The results suggest that nMMD is characterized by a fibrous pseudo-metaplasia of the RPE and is strongly associated with macular BM defects, without other detected histomorphometric differences in thickness or density of the RPE, BM, and choriocapillaris.

Choroidal shift in myopic eyes in the 10-year follow-up Beijing eye study

The aim of the study was to assess longitudinal changes in the spatial relationship of the choroidal vasculature to retinal vasculature in myopic eyes. In the population-based longitudinal Beijing Eye Study in 2001/2011, we examined all highly myopic eyes with assessable fundus photographs and a randomized group of non-highly myopic. Using fundus photographs, we qualitatively assessed changes in the location of major choroidal vessels in relationship to retinal vessels. The study consisted of 85 highly myopic eyes (58 participants;age:64.8 ± 9.4 years) and 85 randomly selected non-highly myopic eyes. A choroidal shift in relationship to the retinal vessels was detected more often in the highly myopic group than the non-highly myopic group (47/85 (55%) vs 6/85 (7%); P < 0.001). In the highly myopic group, the choroidal vessel shift occurring on the disc-fovea line in 39 (44%) eyes, was similar to, or smaller than, the enlargement in gamma zone width in 26 (67%) eyes and in 11 (28%) eyes respectively. The choroidal vessel shift was larger (P = 0.002) in eyes without choroidal vessels in gamma zone than in eyes with large choroidal vessels in gamma zone. In 14 (17%) eyes, a localized centrifugal choroidal shift was observed in association with an increase in the stage of myopic maculopathy. The results suggest that highly myopic eyes show a change in the position of large choroidal vessels in relationship to retinal vessels, in association with development or enlargement of gamma zone and an increase in the stage of myopic maculopathy.

Utilisation of poor-quality optical coherence tomography scans: adjustment algorithm from the Singapore Epidemiology of Eye Diseases (SEED) study

PURPOSE

To evaluate the effect of signal strength (SS) on optical coherence tomography (OCT) parameters, and devise an algorithm to adjust the effect, when acceptable SS cannot be obtained.

METHODS

5085 individuals (9582 eyes), aged ≥40 years from the Singapore Epidemiology of Eye Diseases population-based study were included. Everyone underwent a standardised ocular examination and imaging with Cirrus HD-OCT. Effect of SS was evaluated using multiple structural breaks linear mixed-effect models. Expected change for increment in SS between 4 and 10 for individual parameter was calculated. Subsequently we devised and evaluated an algorithm to adjust OCT parameters to higher SS.

RESULTS

Average retinal nerve fibre layer (RNFL) thickness showed shift of 4.11 µm from SS of 5 to 6. Above 6, it increased by 1.72 and 3.35 µm to 7 and 8; and by 1.09 µm (per unit increase) above 8 SS. Average ganglion cell-inner plexiform layer (GCIPL) thickness shifted 5.15 µm from SS of 5 to 6. Above 6, increased by 0.94 µm from 7 to 8; and by 0.16 µm (per unit increase) above 8 SS. When compared with reference in an independent test set, the algorithm produced less systemic bias. Algorithm-adjusted average RNFL was 0.549 µm thinner than the reference, while the unadjusted one was 2.841 µm thinner (p<0.001). Algorithm-adjusted and unadjusted average GCIPL was 1.102 µm and 2.228 µm thinner (p<0.001).

CONCLUSIONS

OCT parameters can be adjusted for poor SS using an algorithm. This can potentially assist in diagnosis and monitoring of glaucoma when scans with acceptable SS cannot be acquired from patients in clinics.