Automatic geographic atrophy segmentation using optical attenuation in OCT scans with deep learning

Comparison between Spectral-Domain and Swept-Source OCT Angiography for the Measurement of Persistent Hypertransmission Defects in Age-Related Macular Degeneration

Purpose

Spectral-domain OCT angiography (SD-OCTA) scans were tested in an algorithm developed for use with swept-source OCT angiography (SS-OCTA) scans to determine if SD-OCTA scans yielded similar results for the detection and measurement of persistent choroidal hypertransmission defects (hyperTDs).

Design

Retrospective study.

Participants

Forty pairs of scans from 32 patients with late-stage nonexudative age-related macular degeneration (AMD).

Methods

Patients underwent both SD-OCTA and SS-OCTA imaging at the same visit using the 6 × 6 mm OCTA scan patterns. Using a semiautomatic algorithm that helped with outlining the hyperTDs, 2 graders independently validated persistent hyperTDs, which are defined as having a greatest linear dimension ≥250 μm on the en face images generated using a slab extending from 64 to 400 μm beneath Bruch's membrane. The number of lesions and square root (sqrt) total area of the hyperTDs were obtained from the algorithm using each imaging method.

Main Outcome Measures

The mean sqrt area measurements and the number of hyperTDs were compared.

Results

The number of lesions and sqrt total area of the hyperTDs were highly concordant between the 2 instruments (rc = 0.969 and rc = 0.999, respectively). The mean number of hyperTDs was 4.3 ± 3.1 for SD-OCTA scans and 4.5 ± 3.3 for SS-OCTA scans (P = 0.06). The mean sqrt total area measurements were 1.16 ± 0.64 mm for the SD-OCTA scans and 1.17 ± 0.65 mm for the SS-OCTA scans (P < 0.001). Because of the small standard error of the differences, the mean difference between the scans was statistically significant but not clinically significant.

Conclusions

Spectral-domain OCTA scans provide similar results to SS-OCTA scans when used to obtain the number and area measurements of persistent hyperTDs through a semiautomated algorithm previously developed for SS-OCTA. This facilitates the detection of atrophy with a more widely available scan pattern and the longitudinal study of early to late-stage AMD.

Financial Disclosures

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Evaluating the persistence of large choroidal hypertransmission defects using SS-OCT imaging

In age-related macular degeneration (AMD), large choroidal hypertransmission defects (hyperTDs) are identified on en face optical coherence tomography (OCT) images as bright lesions measuring at least 250 μm in greatest linear dimension (GLD). These choroidal hyperTDs arise from focal attenuation or loss of the retinal pigment epithelium (RPE). We previously reported that once large hyperTDs formed, they were likely to persist compared with smaller lesions that were more likely to be transient. Due to their relative persistence, these large persistent choroidal hyperTDs are a point-of-no-return in the progression of intermediate AMD to the late stage of atrophic AMD. Moreover, the onset of these large choroidal hyperTDs can serve as a clinical trial endpoint when studying therapies that might slow disease progression from intermediate AMD to late atrophic AMD. To confirm the persistence of these large choroidal hyperTDs, we studied an independent dataset of AMD eyes enrolled in an ongoing prospective swept-source OCT (SS-OCT) natural history study to determine their overall persistence. We identified a total of 202 eyes with large choroidal hyperTDs containing 1725 hyperTDs followed for an average of 46.6 months. Of the 1725 large hyperTDs, we found that 1718 (99.6%) persisted while only 7 hyperTDs (0.4%) were non-persistent. Of the 7 non-persistent large hyperTDs in 6 eyes, their average GLD at baseline was 385 μm. Of the large hyperTDs ranging in size between 250 and 300 μm when first detected, only one was not persistent with a baseline GLD of 283 μm. In 6 of the non-persistent hyperTDs, the loss of a detectable large hyperTD was due to the accumulation of hyperreflective material along the retinal pigment epithelium (RPE) and in the retina over the area where the hyperTD was located. This hyperreflective material is thought to represent the migration and aggregation of RPE cells into this focal region where the choroidal hyperTD arose due to attenuated or lost RPE.

Artificial intelligence for geographic atrophy: pearls and pitfalls

PURPOSE OF REVIEW

This review aims to address the recent advances of artificial intelligence (AI) in the context of clinical management of geographic atrophy (GA), a vision-impairing late-stage manifestation of age-related macular degeneration (AMD).

RECENT FINDINGS

Recent literature shows substantial advancements in the development of AI systems to segment GA lesions on multimodal retinal images, including color fundus photography (CFP), fundus autofluorescence (FAF) and optical coherence tomography (OCT), providing innovative solutions to screening and early diagnosis. Especially, the high resolution and 3D-nature of OCT has provided an optimal source of data for the training and validation of novel algorithms. The use of AI to measure progression in the context of newly approved GA therapies, has shown that AI methods may soon be indispensable for patient management. To date, while many AI models have been reported on, their implementation in the real-world has only just started. The aim is to make the benefits of AI-based personalized treatment accessible and far-reaching.

SUMMARY

The most recent advances (pearls) and challenges (pitfalls) associated with AI methods and their clinical implementation in the context of GA will be discussed.

AI in the clinical management of GA: A novel therapeutic universe requires novel tools

Regulatory approval of the first two therapeutic substances for the management of geographic atrophy (GA) secondary to age-related macular degeneration (AMD) is a major breakthrough following failure of numerous previous trials. However, in the absence of therapeutic standards, diagnostic tools are a key challenge as functional parameters in GA are hard to provide. The majority of anatomical biomarkers are subclinical, necessitating advanced and sensitive image analyses. In contrast to fundus autofluorescence (FAF), optical coherence tomography (OCT) provides high-resolution visualization of neurosensory layers, including photoreceptors, and other features that are beyond the scope of human expert assessment. Artificial intelligence (AI)-based methodology strongly enhances identification and quantification of clinically relevant GA-related sub-phenotypes. Introduction of OCT-based biomarker analysis provides novel insight into the pathomechanisms of disease progression and therapeutic, moving beyond the limitations of conventional descriptive assessment. Accordingly, the Food and Drug Administration (FDA) has provided a paradigm-shift in recognizing ellipsoid zone (EZ) attenuation as a primary outcome measure in GA clinical trials. In this review, the transition from previous to future GA classification and management is described. With the advent of AI tools, diagnostic and therapeutic concepts have changed substantially in monitoring and screening of GA disease. Novel technology combined with pathophysiological knowledge and understanding of the therapeutic response to GA treatments, is currently opening the path for an automated, efficient and individualized patient care with great potential to improve access to timely treatment and reduce health disparities.

Comparison between OCT B-scan and En Face Imaging for the Diagnosis of Early Macular Atrophy in Age-Related Macular Degeneration

PURPOSE

The gradings of complete retinal pigment epithelium and outer retinal atrophy (cRORA) and incomplete retinal pigment epithelium and outer retinal atrophy (iRORA) on spectral domain optical coherence tomography (SD-OCT) B-scans were compared with the grading of persistent choroidal hypertransmission defects (hyperTDs) on swept-source OCT angiography (SS-OCTA) en face images.

DESIGN

Comparative diagnostic analysis of prospective study data METHODS: Patients with late nonexudative AMD underwent same day 6 × 6 mm macular scans using both SD-OCT (Spectralis® Heidelberg, 512 × 97, ART:9) and SS-OCTA (PLEX® Elite 9000, Carl Zeiss Meditec, 500 × 500 angio pattern) instruments. SS-OCTA and SD-OCT en face images were generated from a sub-retinal pigment epithelium slab positioned 64-400 μm below Bruch's membrane. SD-OCT B-scan gradings, which included an inspection of neighboring B-scans for the diagnosis of cRORA and iRORA, were performed at the Moran Eye Center, while gradings of en face images to identify persistent choroidal hyperTDs were performed at the Bascom Palmer Eye Institute and Tel Aviv Medical Center.

RESULTS

There was a high degree of agreement (99.6%) between the gradings of cRORA lesions and persistent hyperTDs. However, 27.4% of iRORA lesions were found to be contained within persistent hyperTDs. This discrepancy was due to the finding that 27.5% of iRORA lesions were diagnosed as having a greatest linear horizontal dimension of < 250 µm on B-scans, but on en face images, these B-scan defined iRORA lesions were found to have a greatest linear dimensions in the non-horizontal dimension that were ≥ 250 µm.

CONCLUSION

This report demonstrates the benefits of using en face OCT imaging to identify cRORA lesions and highlights the need to acquire dense raster B-scans with the grading neighboring B-scans when identifying iRORA lesions to assess the full extent of the iRORA lesions in the non-horizontal dimension. Even though neighboring B-scans were inspected, 27.5% of iRORA lesions were actually part of larger cRORA lesions when graded using an en face strategy.

Strong versus Weak Data Labeling for Artificial Intelligence Algorithms in the Measurement of Geographic Atrophy

Purpose

To gain an understanding of data labeling requirements to train deep learning models for measurement of geographic atrophy (GA) with fundus autofluorescence (FAF) images.

Design

Evaluation of artificial intelligence (AI) algorithms.

Subjects

The Age-Related Eye Disease Study 2 (AREDS2) images were used for training and cross-validation, and GA clinical trial images were used for testing.

Methods

Training data consisted of 2 sets of FAF images; 1 with area measurements only and no indication of GA location (Weakly labeled) and the second with GA segmentation masks (Strongly labeled).

Main Outcome Measures

Bland-Altman plots and scatter plots were used to compare GA area measurement between ground truth and AI measurements. The Dice coefficient was used to compare accuracy of segmentation of the Strong model.

Results

In the cross-validation AREDS2 data set (n = 601), the mean (standard deviation [SD]) area of GA measured by human grader, Weakly labeled AI model, and Strongly labeled AI model was 6.65 (6.3) mm2, 6.83 (6.29) mm2, and 6.58 (6.24) mm2, respectively. The mean difference between ground truth and AI was 0.18 mm2 (95% confidence interval, [CI], -7.57 to 7.92) for the Weakly labeled model and -0.07 mm2 (95% CI, -1.61 to 1.47) for the Strongly labeled model. With GlaxoSmithKline testing data (n = 156), the mean (SD) GA area was 9.79 (5.6) mm2, 8.82 (4.61) mm2, and 9.55 (5.66) mm2 for human grader, Strongly labeled AI model, and Weakly labeled AI model, respectively. The mean difference between ground truth and AI for the 2 models was -0.97 mm2 (95% CI, -4.36 to 2.41) and -0.24 mm2 (95% CI, -4.98 to 4.49), respectively. The Dice coefficient was 0.99 for intergrader agreement, 0.89 for the cross-validation data, and 0.92 for the testing data.

Conclusions

Deep learning models can achieve reasonable accuracy even with Weakly labeled data. Training methods that integrate large volumes of Weakly labeled images with small number of Strongly labeled images offer a promising solution to overcome the burden of cost and time for data labeling.

Financial Disclosures

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Rediscovering Age-Related Macular Degeneration with Swept-Source OCT Imaging: The 2022 Charles L. Schepens, MD, Lecture

PURPOSE

Swept-source OCT angiography (SS-OCTA) scans of eyes with age-related macular degeneration (AMD) were used to replace color, autofluorescence, infrared reflectance, and dye-based fundus angiographic imaging for the diagnosis and staging of AMD. Through the use of different algorithms with the SS-OCTA scans, both structural and angiographic information can be viewed and assessed using both cross sectional and en face imaging strategies.

DESIGN

Presented at the 2022 Charles L. Schepens, MD, Lecture at the American Academy of Ophthalmology Retina Subspecialty Day, Chicago, Illinois, on September 30, 2022.

PARTICIPANTS

Patients with AMD.

METHODS

Review of published literature and ongoing clinical research using SS-OCTA imaging in AMD.

MAIN OUTCOME MEASURES

Swept-source OCT angiography imaging of AMD at different stages of disease progression.

RESULTS

Volumetric SS-OCTA dense raster scans were used to diagnose and stage both exudative and nonexudative AMD. In eyes with nonexudative AMD, a single SS-OCTA scan was used to detect and measure structural features in the macula such as the area and volume of both typical soft drusen and calcified drusen, the presence and location of hyperreflective foci, the presence of reticular pseudodrusen, also known as subretinal drusenoid deposits, the thickness of the outer retinal layer, the presence and thickness of basal laminar deposits, the presence and area of persistent choroidal hypertransmission defects, and the presence of treatment-naïve nonexudative macular neovascularization. In eyes with exudative AMD, the same SS-OCTA scan pattern was used to detect and measure the presence of macular fluid, the presence and type of macular neovascularization, and the response of exudation to treatment with vascular endothelial growth factor inhibitors. In addition, the same scan pattern was used to quantitate choriocapillaris (CC) perfusion, CC thickness, choroidal thickness, and the vascularity of the choroid.

CONCLUSIONS

Compared with using several different instruments to perform multimodal imaging, a single SS-OCTA scan provides a convenient, comfortable, and comprehensive approach for obtaining qualitative and quantitative anatomic and angiographic information to monitor the onset, progression, and response to therapies in both nonexudative and exudative AMD.

FINANCIAL DISCLOSURE(S)

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

iStent insertion orientation and impact on trabecular meshwork motion resolved by optical coherence tomography imaging

Significance

The iStent is a popular device designed for glaucoma treatment, functioning by creating an artificial fluid pathway in the trabecular meshwork (TM) to drain aqueous humor. The assessment of iStent implantation surgery is clinically important. However, current tools offer limited information.

Aim

We aim to develop innovative assessment strategies for iStent implantation using optical coherence tomography (OCT) to evaluate the position and orientation of the iStent and its biomechanical impact on outflow system dynamics.

Approach

We examined four iStents in the two eyes of a glaucoma patient. Three-dimensional (3D) OCT structural imaging was conducted for each iStent, and a semi-automated algorithm was developed for iStent segmentation and visualization, allowing precise measurement of position and orientation. In addition, phase-sensitive OCT (PhS-OCT) imaging was introduced to measure the biomechanical impact of the iStent on the outflow system quantified by cumulative displacement (CDisp) of pulse-dependent trabecular TM motion.

Results

The 3D structural image processed by our algorithm definitively resolved the position and orientation of the iStent in the anterior segment, revealing substantial variations in relevant parameters. PhS-OCT imaging demonstrated significantly higher CDisp in the regions between two iStents compared to locations distant from the iStents in both OD ( p = 0.0075 ) and OS ( p = 0.0437 ).

Conclusions

Our proposed structural imaging technique improved the characterization of the iStent's placement. The imaging results revealed inherent challenges in achieving precise control of iStent insertion. Furthermore, PhS-OCT imaging unveiled potential biomechanical alterations induced by the iStent. This unique methodology shows potential as a valuable clinical tool for evaluating iStent implantation.

COVLIAS 3.0: cloud-based quantized hybrid UNet3+ deep learning for COVID-19 lesion detection in lung computed tomography

Background and novelty

When RT-PCR is ineffective in early diagnosis and understanding of COVID-19 severity, Computed Tomography (CT) scans are needed for COVID diagnosis, especially in patients having high ground-glass opacities, consolidations, and crazy paving. Radiologists find the manual method for lesion detection in CT very challenging and tedious. Previously solo deep learning (SDL) was tried but they had low to moderate-level performance. This study presents two new cloud-based quantized deep learning UNet3+ hybrid (HDL) models, which incorporated full-scale skip connections to enhance and improve the detections.

Methodology

Annotations from expert radiologists were used to train one SDL (UNet3+), and two HDL models, namely, VGG-UNet3+ and ResNet-UNet3+. For accuracy, 5-fold cross-validation protocols, training on 3,500 CT scans, and testing on unseen 500 CT scans were adopted in the cloud framework. Two kinds of loss functions were used: Dice Similarity (DS) and binary cross-entropy (BCE). Performance was evaluated using (i) Area error, (ii) DS, (iii) Jaccard Index, (iii) Bland-Altman, and (iv) Correlation plots.

Results

Among the two HDL models, ResNet-UNet3+ was superior to UNet3+ by 17 and 10% for Dice and BCE loss. The models were further compressed using quantization showing a percentage size reduction of 66.76, 36.64, and 46.23%, respectively, for UNet3+, VGG-UNet3+, and ResNet-UNet3+. Its stability and reliability were proved by statistical tests such as the Mann-Whitney, Paired t-Test, Wilcoxon test, and Friedman test all of which had a p < 0.001.

Conclusion

Full-scale skip connections of UNet3+ with VGG and ResNet in HDL framework proved the hypothesis showing powerful results improving the detection accuracy of COVID-19.

Calcified Drusen Prevent the Detection of Underlying Choriocapillaris Using Swept-Source Optical Coherence Tomography Angiography

Purpose

In age-related macular degeneration (AMD), choriocapillaris flow deficits (CCFDs) under soft drusen can be measured using established compensation strategies. This study investigated whether CCFDs can be quantified under calcified drusen (CaD).

Methods

CCFDs were measured in normal eyes (n = 30) and AMD eyes with soft drusen (n = 30) or CaD (n = 30). CCFD density masks were generated to highlight regions with higher CCFDs. Masks were also generated for soft drusen and CaD based on both structural en face OCT images and corresponding B-scans. Dice similarity coefficients were calculated between the CCFD density masks and both the soft drusen and CaD masks. A phantom experiment was conducted to simulate the impact of light scattering that arises from CaD.

Results

Area measurements of CCFDs were highly correlated with those of CaD but not soft drusen, suggesting an association between CaD and underlying CCFDs. However, unlike soft drusen, the detected optical coherence tomography (OCT) signals underlying CaD did not arise from the defined CC layer but were artifacts caused by the multiple scattering property of CaD. Phantom experiments showed that the presence of highly scattering material similar to the contents of CaD caused an artifactual scattering tail that falsely generated a signal in the CC structural layer but the underlying flow could not be detected. Similarly, CaD also caused an artifactual scattering tail and prevented the penetration of light into the choroid, resulting in en face hypotransmission defects and an inability to detect blood flow within the choriocapillaris. Upon resolution of the CaD, the CC perfusion became detectable.

Conclusions

The high scattering property of CaD leads to a scattering tail under these drusen that gives the illusion of a quantifiable optical coherence tomography angiography signal, but this signal does not contain the angiographic information required to assess CCFDs. For this reason, CCFDs cannot be reliably measured under CaD, and CaD must be identified and excluded from macular CCFD measurements.

Spectral-Domain and Swept-Source OCT Angiographic Scans Yield Similar Drusen Measurements When Processed with the Same Algorithm

Purpose

An algorithm developed to obtain drusen area and volume measurements using swept-source OCT angiography (SS-OCTA) scans was tested on spectral-domain OCT angiography (SD-OCTA) scans.

Design

Retrospective study.

Participants

Forty pairs of scans from 27 eyes with intermediate age-related macular degeneration and drusen.

Methods

Patients underwent both SD-OCTA and SS-OCTA imaging at the same visit using the 6 mm × 6 mm OCTA scan patterns. Using the same algorithm, we obtained drusen area and volume measurements within both 3 mm and 5 mm fovea-centered circles. Paired 2-sample t-tests were performed along with Pearson's correlation tests.

Main Outcome Measures

Mean square root (sqrt) drusen area and cube root (cbrt) drusen volume within the 3 mm and 5 mm fovea-centered circles.

Results

Mean sqrt drusen area values from SD-OCTA and SS-OCTA scans were 1.57 (standard deviation [SD] 0.57) mm and 1.49 (SD 0.58) mm in the 3 mm circle and 1.88 (SD 0.59) mm and 1.76 (SD 0.58) mm in the 5 mm circle, respectively. Mean cbrt drusen volume measurements were 0.54 (SD 0.19) mm and 0.51 (SD 0.20) mm in the 3 mm circle, and 0.60 (SD 0.17) mm and 0.57 (SD 0.17) mm in the 5 mm circle. Small differences in area and volume measurements were found (all P < 0.001); however, the correlations between the instruments were strong (all coefficients > 0.97; all P < 0.001).

Conclusions

An algorithm originally developed for SS-OCTA scans performs well when used to obtain drusen volume and area measurements from SD-OCTA scans; thus, a separate SD-OCT structural scan is unnecessary to obtain measurements of drusen.

Financial Disclosures

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Recurrent and Concurrent Prediction of Longitudinal Progression of Stargardt Atrophy and Geographic Atrophy

Stargardt disease and age-related macular degeneration are the leading causes of blindness in the juvenile and geriatric populations, respectively. The formation of atrophic regions of the macula is a hallmark of the end-stages of both diseases. The progression of these diseases is tracked using various imaging modalities, two of the most common being fundus autofluorescence (FAF) imaging and spectral-domain optical coherence tomography (SD-OCT). This study seeks to investigate the use of longitudinal FAF and SD-OCT imaging (month 0, month 6, month 12, and month 18) data for the predictive modelling of future atrophy in Stargardt and geographic atrophy. To achieve such an objective, we develop a set of novel deep convolutional neural networks enhanced with recurrent network units for longitudinal prediction and concurrent learning of ensemble network units (termed ReConNet) which take advantage of improved retinal layer features beyond the mean intensity features. Using FAF images, the neural network presented in this paper achieved mean (± standard deviation, SD) and median Dice coefficients of 0.895 (± 0.086) and 0.922 for Stargardt atrophy, and 0.864 (± 0.113) and 0.893 for geographic atrophy. Using SD-OCT images for Stargardt atrophy, the neural network achieved mean and median Dice coefficients of 0.882 (± 0.101) and 0.906, respectively. When predicting only the interval growth of the atrophic lesions with FAF images, mean (± SD) and median Dice coefficients of 0.557 (± 0.094) and 0.559 were achieved for Stargardt atrophy, and 0.612 (± 0.089) and 0.601 for geographic atrophy. The prediction performance in OCT images is comparably good to that using FAF which opens a new, more efficient, and practical door in the assessment of atrophy progression for clinical trials and retina clinics, beyond widely used FAF. These results are highly encouraging for a high-performance interval growth prediction when more frequent or longer-term longitudinal data are available in our clinics. This is a pressing task for our next step in ongoing research.

Deep-learning-based automated measurement of outer retinal layer thickness for use in the assessment of age-related macular degeneration, applicable to both swept-source and spectral-domain OCT imaging

Effective biomarkers are required for assessing the progression of age-related macular degeneration (AMD), a prevalent and progressive eye disease. This paper presents a deep learning-based automated algorithm, applicable to both swept-source OCT (SS-OCT) and spectral-domain OCT (SD-OCT) scans, for measuring outer retinal layer (ORL) thickness as a surrogate biomarker for outer retinal degeneration, e.g., photoreceptor disruption, to assess AMD progression. The algorithm was developed based on a modified TransUNet model with clinically annotated retinal features manifested in the progression of AMD. The algorithm demonstrates a high accuracy with an intersection of union (IoU) of 0.9698 in the testing dataset for segmenting ORL using both SS-OCT and SD-OCT datasets. The robustness and applicability of the algorithm are indicated by strong correlation (r = 0.9551, P < 0.0001 in the central-fovea 3 mm-circle, and r = 0.9442, P < 0.0001 in the 5 mm-circle) and agreement (the mean bias = 0.5440 um in the 3-mm circle, and 1.392 um in the 5-mm circle) of the ORL thickness measurements between SS-OCT and SD-OCT scans. Comparative analysis reveals significant differences (P < 0.0001) in ORL thickness among 80 normal eyes, 30 intermediate AMD eyes with reticular pseudodrusen, 49 intermediate AMD eyes with drusen, and 40 late AMD eyes with geographic atrophy, highlighting its potential as an independent biomarker for predicting AMD progression. The findings provide valuable insights into the ORL alterations associated with different stages of AMD and emphasize the potential of ORL thickness as a sensitive indicator of AMD severity and progression.

Unleashing the power of optical attenuation coefficients to facilitate segmentation strategies in OCT imaging of age-related macular degeneration: perspective

The use of optical attenuation coefficients (OAC) in optical coherence tomography (OCT) imaging of the retina has improved the segmentation of anatomic layers compared with traditional intensity-based algorithms. Optical attenuation correction has improved our ability to measure the choroidal thickness and choroidal vascularity index using dense volume scans. Algorithms that combine conventional intensity-based segmentation with depth-resolved OAC OCT imaging have been used to detect elevations of the retinal pigment epithelium (RPE) due to drusen and basal laminar deposits, the location of hyperpigmentation within the retina and along the RPE, the identification of macular atrophy, the thickness of the outer retinal (photoreceptor) layer, and the presence of calcified drusen. OAC OCT algorithms can identify the risk-factors that predict disease progression in age-related macular degeneration.

Geographic Atrophy Segmentation Using Multimodal Deep Learning

Purpose

To examine deep learning (DL)-based methods for accurate segmentation of geographic atrophy (GA) lesions using fundus autofluorescence (FAF) and near-infrared (NIR) images.

Methods

This retrospective analysis utilized imaging data from study eyes of patients enrolled in Proxima A and B (NCT02479386; NCT02399072) natural history studies of GA. Two multimodal DL networks (UNet and YNet) were used to automatically segment GA lesions on FAF; segmentation accuracy was compared with annotations by experienced graders. The training data set comprised 940 image pairs (FAF and NIR) from 183 patients in Proxima B; the test data set comprised 497 image pairs from 154 patients in Proxima A. Dice coefficient scores, Bland-Altman plots, and Pearson correlation coefficient (r) were used to assess performance.

Results

On the test set, Dice scores for the DL network to grader comparison ranged from 0.89 to 0.92 for screening visit; Dice score between graders was 0.94. GA lesion area correlations (r) for YNet versus grader, UNet versus grader, and between graders were 0.981, 0.959, and 0.995, respectively. Longitudinal GA lesion area enlargement correlations (r) for screening to 12 months (n = 53) were lower (0.741, 0.622, and 0.890, respectively) compared with the cross-sectional results at screening. Longitudinal correlations (r) from screening to 6 months (n = 77) were even lower (0.294, 0.248, and 0.686, respectively).

Conclusions

Multimodal DL networks to segment GA lesions can produce accurate results comparable with expert graders.

Translational Relevance

DL-based tools may support efficient and individualized assessment of patients with GA in clinical research and practice.

Detection of macular atrophy in age-related macular degeneration aided by artificial intelligence

INTRODUCTION

Age-related macular degeneration (AMD) is a leading cause of irreversible visual impairment worldwide. The endpoint of AMD, both in its dry or wet form, is macular atrophy (MA) which is characterized by the permanent loss of the RPE and overlying photoreceptors either in dry AMD or in wet AMD. A recognized unmet need in AMD is the early detection of MA development.

AREAS COVERED

Artificial Intelligence (AI) has demonstrated great impact in detection of retinal diseases, especially with its robust ability to analyze big data afforded by ophthalmic imaging modalities, such as color fundus photography (CFP), fundus autofluorescence (FAF), near-infrared reflectance (NIR), and optical coherence tomography (OCT). Among these, OCT has been shown to have great promise in identifying early MA using the new criteria in 2018.

EXPERT OPINION

There are few studies in which AI-OCT methods have been used to identify MA; however, results are very promising when compared to other imaging modalities. In this paper, we review the development and advances of ophthalmic imaging modalities and their combination with AI technology to detect MA in AMD. In addition, we emphasize the application of AI-OCT as an objective, cost-effective tool for the early detection and monitoring of the progression of MA in AMD.

Deep learning segmentation of the tear fluid reservoir under the sclera lens in optical coherence tomography images

The tear fluid reservoir (TFR) under the sclera lens is a unique characteristic providing optical neutralization of any aberrations from corneal irregularities. Anterior segment optical coherence tomography (AS-OCT) has become an important imaging modality for sclera lens fitting and visual rehabilitation therapy in both optometry and ophthalmology. Herein, we aimed to investigate whether deep learning can be used to segment the TFR from healthy and keratoconus eyes, with irregular corneal surfaces, in OCT images. Using AS-OCT, a dataset of 31850 images from 52 healthy and 46 keratoconus eyes, during sclera lens wear, was obtained and labeled with our previously developed algorithm of semi-automatic segmentation. A custom-improved U-shape network architecture with a full-range multi-scale feature-enhanced module (FMFE-Unet) was designed and trained. A hybrid loss function was designed to focus training on the TFR, to tackle the class imbalance problem. The experiments on our database showed an IoU, precision, specificity, and recall of 0.9426, 0.9678, 0.9965, and 0.9731, respectively. Furthermore, FMFE-Unet was found to outperform the other two state-of-the-art methods and ablation models, suggesting its strength in segmenting the TFR under the sclera lens depicted on OCT images. The application of deep learning for TFR segmentation in OCT images provides a powerful tool to assess changes in the dynamic tear film under the sclera lens, improving the efficiency and accuracy of lens fitting, and thus supporting the promotion of sclera lenses in clinical practice.

Automated segmentation and quantification of calcified drusen in 3D swept source OCT imaging

Qualitative and quantitative assessments of calcified drusen are clinically important for determining the risk of disease progression in age-related macular degeneration (AMD). This paper reports the development of an automated algorithm to segment and quantify calcified drusen on swept-source optical coherence tomography (SS-OCT) images. The algorithm leverages the higher scattering property of calcified drusen compared with soft drusen. Calcified drusen have a higher optical attenuation coefficient (OAC), which results in a choroidal hypotransmission defect (hypoTD) below the calcified drusen. We show that it is possible to automatically segment calcified drusen from 3D SS-OCT scans by combining the OAC within drusen and the hypoTDs under drusen. We also propose a correction method for the segmentation of the retina pigment epithelium (RPE) overlying calcified drusen by automatically correcting the RPE by an amount of the OAC peak width along each A-line, leading to more accurate segmentation and quantification of drusen in general, and the calcified drusen in particular. A total of 29 eyes with nonexudative AMD and calcified drusen imaged with SS-OCT using the 6 × 6 mm² scanning pattern were used in this study to test the performance of the proposed automated method. We demonstrated that the method achieved good agreement with the human expert graders in identifying the area of calcified drusen (Dice similarity coefficient: 68.27 ± 11.09%, correlation coefficient of the area measurements: r = 0.9422, the mean bias of the area measurements = 0.04781 mm²).

A Deep Learning Model for Automated Segmentation of Geographic Atrophy Imaged Using Swept-Source OCT

PURPOSE

To present a deep learning algorithm for segmentation of geographic atrophy (GA) using en face swept-source OCT (SS-OCT) images that is accurate and reproducible for the assessment of GA growth over time.

DESIGN

Retrospective review of images obtained as part of a prospective natural history study.

SUBJECTS

Patients with GA (n = 90), patients with early or intermediate age-related macular degeneration (n = 32), and healthy controls (n = 16).

METHODS

An automated algorithm using scan volume data to generate 3 image inputs characterizing the main OCT features of GA-hypertransmission in subretinal pigment epithelium (sub-RPE) slab, regions of RPE loss, and loss of retinal thickness-was trained using 126 images (93 with GA and 33 without GA, from the same number of eyes) using a fivefold cross-validation method and data augmentation techniques. It was tested in an independent set of one hundred eighty 6 × 6-mm² macular SS-OCT scans consisting of 3 repeated scans of 30 eyes with GA at baseline and follow-up as well as 45 images obtained from 42 eyes without GA.

MAIN OUTCOME MEASURES

The GA area, enlargement rate of GA area, square root of GA area, and square root of the enlargement rate of GA area measurements were calculated using the automated algorithm and compared with ground truth calculations performed by 2 manual graders. The repeatability of these measurements was determined using intraclass coefficients (ICCs).

RESULTS

There were no significant differences in the GA areas, enlargement rates of GA area, square roots of GA area, and square roots of the enlargement rates of GA area between the graders and the automated algorithm. The algorithm showed high repeatability, with ICCs of 0.99 and 0.94 for the GA area measurements and the enlargement rates of GA area, respectively. The repeatability limit for the GA area measurements made by grader 1, grader 2, and the automated algorithm was 0.28, 0.33, and 0.92 mm², respectively.

CONCLUSIONS

When compared with manual methods, this proposed deep learning-based automated algorithm for GA segmentation using en face SS-OCT images was able to accurately delineate GA and produce reproducible measurements of the enlargement rates of GA.

Machine Learning-Based Automated Detection and Quantification of Geographic Atrophy and Hypertransmission Defects Using Spectral Domain Optical Coherence Tomography

The current study describes the development and assessment of innovative, machine learning (ML)-based approaches for automated detection and pixel-accurate measurements of regions with geographic atrophy (GA) in late-stage age-related macular degeneration (AMD) using optical coherence tomography systems. 900 OCT volumes, 100266 B-scans, and en face OCT images from 341 non-exudative AMD patients with or without GA were included in this study from both Cirrus (Zeiss) and Spectralis (Heidelberg) OCT systems. B-scan and en face level ground truth GA masks were created on OCT B-scan where the segmented ellipsoid zone (EZ) line, retinal pigment epithelium (RPE) line, and bruchs membrane (BM) line overlapped. Two deep learning-based approaches, B-scan level and en face level, were trained. The OCT B-scan model had detection accuracy of 91% and GA area measurement accuracy of 94%. The en face OCT model had detection accuracy of 82% and GA area measurement accuracy of 96% with primary target of hypertransmission on en face OCT. Accuracy was good for both devices tested (92-97%). Automated lesion size stratification for CAM cRORA definition of 250um minimum lesion size was feasible. High-performance models for automatic detection and segmentation of GA area were achieved using OCT systems and deep learning. The automatic measurements showed high correlation with the ground truth. The en face model excelled at identification of hypertransmission defects. The models performance generalized well across device types tested. Future development will include integration of both models to enhance feature detection across GA lesions as well as isolating hypertransmission defects without GA for pre-GA biomarker extraction.

Does the Outer Retinal Thickness Around Geographic Atrophy Represent Another Clinical Biomarker for Predicting Growth?

PURPOSE

To determine whether the outer retinal layer (ORL) thickness around geographic atrophy (GA) could serve as a clinical biomarker to predict the annual enlargement rate (ER) of GA.

DESIGN

Retrospective analysis of a prospective, observational case series.

METHODS

Eyes with GA were imaged with a swept-source OCT 6 × 6 mm scan pattern. GA lesions were measured from customized en face OCT images and the annual ERs were calculated. The ORL was defined and segmented from the inner boundary of outer plexiform layer (OPL) to the inner boundary of retinal pigment epithelium (RPE) layer. The ORL thickness was measured at different subregions around GA.

RESULTS

A total of 38 eyes from 27 participants were included. The same eyes were used for the choriocapillaris (CC) flow deficit (FD) analysis and the RPE to the Bruch membrane (RPE-BM) distance measurements. A negative correlation was observed between the ORL thickness and the GA growth. The ORL thickness in a 300-μm rim around GA showed the strongest correlation with the GA growth (r = -0.457, P = .004). No correlations were found between the ORL thickness and the CC FDs; however, a significant correlation was found between the ORL thickness and the RPE-BM distances around GA (r = -0.398, P = .013).

CONCLUSIONS

ORL thickness showed a significant negative correlation with annual GA growth, but also showed a significant correlation with the RPE-BM distances, suggesting that they were dependently correlated with GA growth. This finding suggests that the loss of photoreceptors was associated with the formation of basal laminar deposits around GA.

Automated segmentation and feature discovery of age-related macular degeneration and Stargardt disease via self-attended neural networks

Age-related macular degeneration (AMD) and Stargardt disease are the leading causes of blindness for the elderly and young adults respectively. Geographic atrophy (GA) of AMD and Stargardt atrophy are their end-stage outcomes. Efficient methods for segmentation and quantification of these atrophic lesions are critical for clinical research. In this study, we developed a deep convolutional neural network (CNN) with a trainable self-attended mechanism for accurate GA and Stargardt atrophy segmentation. Compared with traditional post-hoc attention mechanisms which can only visualize CNN features, our self-attended mechanism is embedded in a fully convolutional network and directly involved in training the CNN to actively attend key features for enhanced algorithm performance. We applied the self-attended CNN on the segmentation of AMD and Stargardt atrophic lesions on fundus autofluorescence (FAF) images. Compared with a preexisting regular fully convolutional network (the U-Net), our self-attended CNN achieved 10.6% higher Dice coefficient and 17% higher IoU (intersection over union) for AMD GA segmentation, and a 22% higher Dice coefficient and a 32% higher IoU for Stargardt atrophy segmentation. With longitudinal image data having over a longer time, the developed self-attended mechanism can also be applied on the visual discovery of early AMD and Stargardt features.

Depth-resolved visualization and automated quantification of hyperreflective foci on OCT scans using optical attenuation coefficients

An automated depth-resolved algorithm using optical attenuation coefficients (OACs) was developed to visualize, localize, and quantify hyperreflective foci (HRF) seen on OCT imaging that are associated with macular hyperpigmentation and represent an increased risk of disease progression in age related macular degeneration. To achieve this, we first transformed the OCT scans to linear representation, which were then contrasted by OACs. HRF were visualized and localized within the entire scan by differentiating HRF within the retina from HRF along the retinal pigment epithelium (RPE). The total pigment burden was quantified using the en face sum projection of an OAC slab between the inner limiting membrane (ILM) to Bruch's membrane (BM). The manual total pigment burden measurements were also obtained by combining manual outlines of HRF in the B-scans with the total area of hypotransmission defects outlined on sub-RPE slabs, which was used as the reference to compare with those obtained from the automated algorithm. 6×6 mm swept-source OCT scans were collected from a total of 49 eyes from 42 patients with macular HRF. We demonstrate that the algorithm was able to automatically distinguish between HRF within the retina and HRF along the RPE. In 24 test eyes, the total pigment burden measurements by the automated algorithm were compared with measurements obtained from manual segmentations. A significant correlation was found between the total pigment area measurements from the automated and manual segmentations (P < 0.001). The proposed automated algorithm based on OACs should be useful in studying eye diseases involving HRF.

Intraoral optical coherence tomography and angiography combined with autofluorescence for dental assessment

There remains a clinical need for an accurate and non-invasive imaging tool for intraoral evaluation of dental conditions. Optical coherence tomography (OCT) is a potential candidate to meet this need, but the design of current OCT systems limits their utility in the intraoral examinations. The inclusion of light-induced autofluorescence (LIAF) can expedite the image collection process and provides a large field of view for viewing the condition of oral tissues. This study describes a novel LIAF-OCT system equipped with a handheld probe designed for intraoral examination of microstructural (via OCT) and microvascular information (via OCT angiography, OCTA). The handheld probe is optimized for use in clinical studies, maintaining the ability to detect and image changes in the condition of oral tissue (e.g., hard tissue damage, presence of dental restorations, plaque, and tooth stains). The real-time LIAF provides guidance for OCT imaging to achieve a field of view of approximately 6.9 mm × 7.8 mm, and a penetration depth of 1.5 mm to 3 mm depending on the scattering property of the target oral tissue. We demonstrate that the proposed system is successful in capturing reliable depth-resolved images from occlusal and palatal surfaces and offers added design features that can enhance its usability in clinical settings.

COVLIAS 1.0Lesion vs. MedSeg: An Artificial Intelligence Framework for Automated Lesion Segmentation in COVID-19 Lung Computed Tomography Scans

Background: COVID-19 is a disease with multiple variants, and is quickly spreading throughout the world. It is crucial to identify patients who are suspected of having COVID-19 early, because the vaccine is not readily available in certain parts of the world. Methodology: Lung computed tomography (CT) imaging can be used to diagnose COVID-19 as an alternative to the RT-PCR test in some cases. The occurrence of ground-glass opacities in the lung region is a characteristic of COVID-19 in chest CT scans, and these are daunting to locate and segment manually. The proposed study consists of a combination of solo deep learning (DL) and hybrid DL (HDL) models to tackle the lesion location and segmentation more quickly. One DL and four HDL models—namely, PSPNet, VGG-SegNet, ResNet-SegNet, VGG-UNet, and ResNet-UNet—were trained by an expert radiologist. The training scheme adopted a fivefold cross-validation strategy on a cohort of 3000 images selected from a set of 40 COVID-19-positive individuals. Results: The proposed variability study uses tracings from two trained radiologists as part of the validation. Five artificial intelligence (AI) models were benchmarked against MedSeg. The best AI model, ResNet-UNet, was superior to MedSeg by 9% and 15% for Dice and Jaccard, respectively, when compared against MD 1, and by 4% and 8%, respectively, when compared against MD 2. Statistical tests—namely, the Mann−Whitney test, paired t-test, and Wilcoxon test—demonstrated its stability and reliability, with p < 0.0001. The online system for each slice was <1 s. Conclusions: The AI models reliably located and segmented COVID-19 lesions in CT scans. The COVLIAS 1.0Lesion lesion locator passed the intervariability test.