Deep learning to distinguish Best vitelliform macular dystrophy (BVMD) from adult-onset vitelliform macular degeneration (AVMD)

Artificial intelligence in retinal screening using OCT images: A review of the last decade (2013-2023)

BACKGROUND AND OBJECTIVES

Optical coherence tomography (OCT) has ushered in a transformative era in the domain of ophthalmology, offering non-invasive imaging with high resolution for ocular disease detection. OCT, which is frequently used in diagnosing fundamental ocular pathologies, such as glaucoma and age-related macular degeneration (AMD), plays an important role in the widespread adoption of this technology. Apart from glaucoma and AMD, we will also investigate pertinent pathologies, such as epiretinal membrane (ERM), macular hole (MH), macular dystrophy (MD), vitreomacular traction (VMT), diabetic maculopathy (DMP), cystoid macular edema (CME), central serous chorioretinopathy (CSC), diabetic macular edema (DME), diabetic retinopathy (DR), drusen, glaucomatous optic neuropathy (GON), neovascular AMD (nAMD), myopia macular degeneration (MMD) and choroidal neovascularization (CNV) diseases. This comprehensive review examines the role that OCT-derived images play in detecting, characterizing, and monitoring eye diseases.

METHOD

The 2020 PRISMA guideline was used to structure a systematic review of research on various eye conditions using machine learning (ML) or deep learning (DL) techniques. A thorough search across IEEE, PubMed, Web of Science, and Scopus databases yielded 1787 publications, of which 1136 remained after removing duplicates. Subsequent exclusion of conference papers, review papers, and non-open-access articles reduced the selection to 511 articles. Further scrutiny led to the exclusion of 435 more articles due to lower-quality indexing or irrelevance, resulting in 76 journal articles for the review.

RESULTS

During our investigation, we found that a major challenge for ML-based decision support is the abundance of features and the determination of their significance. In contrast, DL-based decision support is characterized by a plug-and-play nature rather than relying on a trial-and-error approach. Furthermore, we observed that pre-trained networks are practical and especially useful when working on complex images such as OCT. Consequently, pre-trained deep networks were frequently utilized for classification tasks. Currently, medical decision support aims to reduce the workload of ophthalmologists and retina specialists during routine tasks. In the future, it might be possible to create continuous learning systems that can predict ocular pathologies by identifying subtle changes in OCT images.

DEEP LEARNING FOR AUTOMATIC PREDICTION OF EARLY ACTIVATION OF TREATMENT-NAIVE NONEXUDATIVE MACULAR NEOVASCULARIZATIONS IN AGE-RELATED MACULAR DEGENERATION

BACKGROUND

Around 30% of nonexudative macular neovascularizations exudate within 2 years from diagnosis in patients with age-related macular degeneration. The aim of this study is to develop a deep learning classifier based on optical coherence tomography (OCT) and OCT angiography (OCTA) to identify nonexudative macular neovascularizations at risk of exudation.

METHODS

Patients with age-related macular degeneration showing OCTA and fluorescein angiography-documented nonexudative macular neovascularization with a 2-year minimum imaging follow-up were retrospectively selected. Patients showing OCT B-scan-documented macular neovascularization exudation within the first 2 years formed the EX GROUP while the others formed the QU GROUP. ResNet-101, Inception-ResNet-v2, and DenseNet-201 were independently trained on OCTA and OCT B-scan images. Combinations of the six models were evaluated with major and soft voting techniques.

RESULTS

Eighty-nine eyes of 89 patients with a follow-up of 5.7 ± 1.5 years were recruited (35 EX GROUP and 54 QU GROUP). Inception-ResNet-v2 was the best performing among the three single convolutional neural networks. The major voting model resulting from the association of the three different convolutional neural networks resulted in an improvement of performance both for OCTA and OCT B-scan (both significantly higher than human graders' performance). The soft voting model resulting from the combination of OCTA and OCT B-scan-based major voting models showed a testing accuracy of 94.4%. Peripheral arcades and large vessels on OCTA en face imaging were more prevalent in the QU GROUP.

CONCLUSION

Artificial intelligence shows high performances in identifications of nonexudative macular neovascularizations at risk for exudation within the first 2 years of follow-up, allowing better customization of follow-up timing and avoiding treatment delay. Better results are obtained with the combination of OCTA and OCT B-scan image analysis.

Artificial Intelligence (AI) for Early Diagnosis of Retinal Diseases

Artificial intelligence (AI) has emerged as a transformative tool in the field of ophthalmology, revolutionizing disease diagnosis and management. This paper provides a comprehensive overview of AI applications in various retinal diseases, highlighting its potential to enhance screening efficiency, facilitate early diagnosis, and improve patient outcomes. Herein, we elucidate the fundamental concepts of AI, including machine learning (ML) and deep learning (DL), and their application in ophthalmology, underscoring the significance of AI-driven solutions in addressing the complexity and variability of retinal diseases. Furthermore, we delve into the specific applications of AI in retinal diseases such as diabetic retinopathy (DR), age-related macular degeneration (AMD), Macular Neovascularization, retinopathy of prematurity (ROP), retinal vein occlusion (RVO), hypertensive retinopathy (HR), Retinitis Pigmentosa, Stargardt disease, best vitelliform macular dystrophy, and sickle cell retinopathy. We focus on the current landscape of AI technologies, including various AI models, their performance metrics, and clinical implications. Furthermore, we aim to address challenges and pitfalls associated with the integration of AI in clinical practice, including the "black box phenomenon", biases in data representation, and limitations in comprehensive patient assessment. In conclusion, this review emphasizes the collaborative role of AI alongside healthcare professionals, advocating for a synergistic approach to healthcare delivery. It highlights the importance of leveraging AI to augment, rather than replace, human expertise, thereby maximizing its potential to revolutionize healthcare delivery, mitigate healthcare disparities, and improve patient outcomes in the evolving landscape of medicine.

RobOCTNet: Robotics and Deep Learning for Referable Posterior Segment Pathology Detection in an Emergency Department Population

Purpose

To evaluate the diagnostic performance of a robotically aligned optical coherence tomography (RAOCT) system coupled with a deep learning model in detecting referable posterior segment pathology in OCT images of emergency department patients.

Methods

A deep learning model, RobOCTNet, was trained and internally tested to classify OCT images as referable versus non-referable for ophthalmology consultation. For external testing, emergency department patients with signs or symptoms warranting evaluation of the posterior segment were imaged with RAOCT. RobOCTNet was used to classify the images. Model performance was evaluated against a reference standard based on clinical diagnosis and retina specialist OCT review.

Results

We included 90,250 OCT images for training and 1489 images for internal testing. RobOCTNet achieved an area under the curve (AUC) of 1.00 (95% confidence interval [CI], 0.99-1.00) for detection of referable posterior segment pathology in the internal test set. For external testing, RAOCT was used to image 72 eyes of 38 emergency department patients. In this set, RobOCTNet had an AUC of 0.91 (95% CI, 0.82-0.97), a sensitivity of 95% (95% CI, 87%-100%), and a specificity of 76% (95% CI, 62%-91%). The model's performance was comparable to two human experts' performance.

Conclusions

A robotically aligned OCT coupled with a deep learning model demonstrated high diagnostic performance in detecting referable posterior segment pathology in a cohort of emergency department patients.

Translational Relevance

Robotically aligned OCT coupled with a deep learning model may have the potential to improve emergency department patient triage for ophthalmology referral.

Deep learning model for automatic differentiation of EMAP from AMD in macular atrophy

To create a deep learning (DL) classifier pre-trained on fundus autofluorescence (FAF) images that can assist the clinician in distinguishing age-related geographic atrophy from extensive macular atrophy and pseudodrusen-like appearance (EMAP). Patients with complete outer retinal and retinal pigment epithelium atrophy secondary to either EMAP (EMAP Group) or to dry age related macular degeneration (AMD group) were retrospectively selected. Fovea-centered posterior pole (30° × 30°) and 55° × 55° degree-field-of-view FAF images of sufficiently high quality were collected and used to train two different deep learning (DL) classifiers based on ResNet-101 design. Testing was performed on a set of images coming from a different center. A total of 300 patients were recruited, 135 belonging to EMAP group and 165 belonging to AMD group. The 30° × 30° FAF based DL classifier showed a sensitivity of 84.6% and a specificity of 85.3% for the diagnosis of EMAP. The 55° × 55° FAF based DL classifier showed a sensitivity of 90% and a specificity of 84.6%, a performance that was significantly higher than that of the 30° × 30° classifer (p = 0.037). Artificial intelligence can accurately distinguish between atrophy caused by AMD or by EMAP on FAF images. Its performance are improved using wide field acquisitions.

Artificial intelligence in retinal disease: clinical application, challenges, and future directions

Retinal diseases are a leading cause of blindness in developed countries, accounting for the largest share of visually impaired children, working-age adults (inherited retinal disease), and elderly individuals (age-related macular degeneration). These conditions need specialised clinicians to interpret multimodal retinal imaging, with diagnosis and intervention potentially delayed. With an increasing and ageing population, this is becoming a global health priority. One solution is the development of artificial intelligence (AI) software to facilitate rapid data processing. Herein, we review research offering decision support for the diagnosis, classification, monitoring, and treatment of retinal disease using AI. We have prioritised diabetic retinopathy, age-related macular degeneration, inherited retinal disease, and retinopathy of prematurity. There is cautious optimism that these algorithms will be integrated into routine clinical practice to facilitate access to vision-saving treatments, improve efficiency of healthcare systems, and assist clinicians in processing the ever-increasing volume of multimodal data, thereby also liberating time for doctor-patient interaction and co-development of personalised management plans.

ARTIFICIAL INTELLIGENCE'S ROLE IN DIFFERENTIATING THE ORIGIN FOR SUBRETINAL BLEEDING IN PATHOLOGIC MYOPIA

PURPOSE

To identify salient imaging features to support human-based differential diagnosis between subretinal hemorrhage (SH) due to choroidal neovascularization (CNV) onset and SH without CNV (simple bleeding [SB]) in pathologic myopia eyes using a machine learning (ML)-based step-wise approach.

METHODS

Four different methods for feature extraction were applied: GradCAM visualization, reverse engineering, image processing, and human graders' measurements. GradCAM was performed on a deep learning model derived from Inception-ResNet-v2 trained with OCT B-scan images. Reverse engineering consisted of merging U-Net architecture with a deconvolutional network. Image processing consisted of the application of a local adaptive threshold. Available OCT B-scan images were divided in two groups: the first group was classified by graders before knowing the results of feature extraction and the second (different images) was classified after familiarization with the results of feature extraction.

RESULTS

Forty-seven and 37 eyes were included in the CNV group and the simple bleeding group, respectively. Choroidal neovascularization eyes showed higher baseline central macular thickness ( P = 0.036). Image processing evidenced in CNV eyes an inhomogeneity of the subretinal material and an interruption of the Bruch membrane at the margins of the SH area. Graders' classification performance improved from an accuracy of 76.9% without guidance to 83.3% with the guidance of the three methods ( P  = 0.02). Deep learning accuracy in the task was 86.0%.

CONCLUSION

Artificial intelligence helps identifying imaging biomarkers suggestive of CNV in the context of SH in myopia, improving human ability to perform differential diagnosis on unprocessed baseline OCT B-scan images. Deep learning can accurately distinguish between the two causes of SH.

Correlation between AI-measured lacquer cracks extension and development of myopic choroidal neovascularization

OBJECTIVES

To investigate the correlation between the AI-measured area of the lacquer cracks (LC) at their first detection and the occurrence of a choroidal neovascularization (CNV) during the follow-up in patients affected by pathologic myopia. Secondary outcome was the detection of a correlation between the time to onset of CNV with both baseline LC area and LC area increase during follow-up.

METHODS

Optical coherence tomography (OCT) acquisitions of patients diagnosed with LC were retrospectively analysed. The study population was divided in a CNV group (showing the documented onset of a CNV) and a n-CNV group (no CNV development during follow-up). LC area was measured using MatLab software after the application of a customized method for LC segmentation on infrared (IR) enface images.

RESULTS

Forty-five (45) patients with a mean follow-up of 4.9 ± 1.5 years were included. LC area at baseline was 2.82 ± 0.54 mm2 and 1.70 ± 0.49 mm2 in CNV (20 patients) and n-CNV group (25 patients) group respectively (p < 0.001). LC area increase was significantly higher in CNV group (p < 0.001). Time to onset of CNV was linearly correlated with both LC area at baseline (p = 0.006) and LC area increase (p < 0.001).

CONCLUSIONS

Myopic CNV development is associated with lager LC areas and higher LC area increase during time. Earlier CNV onset is inversely correlated with LC area and LC area increase.

Adult-onset foveomacular vitelliform dystrophy: epidemiology, pathophysiology, imaging, and prognosis

Adult-onset foveomacular dystrophy (AOFVD) is a retinal pattern dystrophy that may affect up to 1 in 7,400 individuals. There is much that is unknown regarding this disease's epidemiology, risk factors for development, and rate of progression through its four stages. Advancements in retinal imaging over the past 15 years have enabled improved characterization of the different stages of AOFVD. These imaging advancements also offer new ways of differentiating AOFVD from phenotypically similar retinal diseases like age-related macular degeneration and Best disease. This review synthesizes the most recent discoveries regarding imaging correlates within AOFVD as well as risk factors for the development of AOFVD, complications of AOFVD, and treatment options. Our aim is to provide ophthalmologists a succinct resource so that they may offer clarity, guidance, and appropriate monitoring and treatments for their patients with suspected AOFVD.

Advanced ImageJ Analysis in Degenerative Acquired Vitelliform Lesions Using Techniques Based on Optical Coherence Tomography

Acquired vitelliform lesions (AVLs) are associated with a large spectrum of retinal diseases, among which is age-related macular degeneration (AMD). The purpose of this study was to characterize AVLs' evolution in AMD patients using optical coherence tomography (OCT) technology and ImageJ software. We measured AVLs' size and density and followed their impacts over surrounding retinal layers. Average retinal pigment epithelium (RPE) thickness in the central 1 mm quadrant (45.89 ± 27.84 µm vs. 15.57 ± 1.40 µm) was significantly increased, as opposed to the outer nuclear layer (ONL) thickness, which was decreased (77.94 ± 18.30 µm vs. 88.64 ± 7.65 µm) in the vitelliform group compared to the control group. We found a continuous external limiting membrane (ELM) in 55.5% of the eyes compared to a continuous ellipsoid zone (EZ) in 22.2% of the eyes in the vitelliform group. The difference between the mean AVLs' volume at baseline compared to the last visit for the nine eyes with ophthalmologic follow-up was not statistically significant (p = 0.725). The median follow-up duration was 11 months (range 5-56 months). Seven eyes (43.75%) were treated with intravitreal anti-vascular endothelium growth factor (anti-VEGF) agent injections, in which we noted a 6.43 ± 9 letter decrease in the best-corrected visual acuity (BCVA). The increased RPE thickness could suggest hyperplasia contrary to the decreased ONL, which could mirror the impact of the vitelliform lesion on photoreceptors (PR). Eyes that received anti-VEGF injections did not show signs of improvement regarding BCVA.

Glutamate Permeability of Chicken Best1

Bestrophin-1 (Best1) is a calcium (Ca²⁺)-activated chloride (Cl^-) channel which has a phylogenetically conserved channel structure with an aperture and neck in the ion-conducting pathway. Mammalian mouse Best1 (mBest1) has been known to have a permeability for large organic anions including gluconate, glutamate, and D-serine, in addition to several small monovalent anions, such as Cl^‑, bromine (Br^-), iodine (I^-), and thiocyanate (SCN^-). However, it is still unclear whether non-mammalian Best1 has a glutamate permeability through the ion-conducting pathway. Here, we report that chicken Best1 (cBest1) is permeable to glutamate in a Ca²⁺-dependent manner. The molecular docking and molecular dynamics simulation showed a glutamate binding at the aperture and neck of cBest1 and a glutamate permeation through the ion-conducting pore, respectively. Moreover, through electrophysiological recordings, we calculated the permeability ratio of glutamate to Cl^- (P_Glutamate/P_Cl) as 0.28 based on the reversal potential shift by ion substitution from Cl^- to glutamate in the internal solution. Finally, we directly detected the Ca²⁺-dependent glutamate release through cBest1 using the ultrasensitive two-cell sniffer patch technique. Our results propose that Best1 homologs from non-mammalian (cBest1) to mammalian (mBest1) have a conserved permeability for glutamate.