Scanning Laser Ophthalmoscope Images Research Articles

Systemic Sodium Iodate Injection as a Model for Expanding Geographic Atrophy.

Geographic atrophy (GA), an advanced form of dry age-related macular degeneration (AMD), has limited treatment options. This study introduces a novel mouse model featuring an expanding GA patch that can be used to test mechanisms and therapeutics. C57Bl/6J male mice (n = 96) aged 9-10 weeks received an intraperitoneal (IP) injection of 20mg/kg sodium iodate (NaIO3). In vivo confocal scanning laser ophthalmoscope (cSLO) and optical coherence tomography imaging were done at one, four, eight, and 16 weeks after injection, with GA area measurements taken at weeks 8 and 16. Mice were euthanized on weeks 8 and 16 for histological analysis. Administration of 20mg/kg intraperitoneal NaIO3 caused variable damage levels. Approximately 22% of cases showed damage (speckled autofluorescence) covering 35% to 90% of the 102° field of view cSLO image at one week after injection. These mice developed an expanding patch of GA by week 8, with a mean 1.45-fold increase in area by week 16. This region showed complete photoreceptor and retinal pigment epithelium loss and complement activation at the atrophy edge, whereas the inner retina remained undamaged. Mice with less damage (48% of cases) only developed incomplete outer retinal degeneration, and mice with more damage (30% of cases) had too much GA for measurable expansion. Although expanding GA formed in only 22% of mice, the model's simplicity and predictability for GA development via one-week post-injection imaging make it suitable for GA therapeutic experimentation. This murine model provides a valuable tool for testing GA therapies, mirroring clinical endpoints relevant to human trials.

Optic nerve assessment with stereo photographs and ultra-widefield scanning laser ophthalmoscope images

ABSTRACT Clinical relevance Clinical evaluation of the optic nerve using 3-D stereo disc photographs is considered the gold standard for estimating vertical cup-to-disc ratios. Ultra-widefield retinal imaging has gained increasing popularity to document and screen the health of the retina and optic nerve. Background Glaucoma is often first identified or suspected based on initial optic nerve assessment. Despite technological advancements in imaging, stereo disc photography remains the gold standard of optic nerve assessment. This study compared vertical cup-to-disc ratio (VCDR) estimations using ultra-widefield scanning laser ophthalmoscopy images with estimations using paired stereo disc images. Methods Thirty-five healthy and 35 glaucomatous eyes were imaged using Zeiss VISUCAM® PRO NM and Optos Monaco® (optomap®) [Registered in U.S. Patent and Trademark Office] devices. Four experienced clinicians assessed each image in a masked fashion and graded the VCDR. Stereo viewers and review software tools were used to determine VCDRs. Bland-Altman plots and regression/correlation analyses were used to check VCDR agreement. Results Bland-Altman analyses showed low mean differences between devices with VCDR estimation for each grader (−0.03, +0.01, −0.02, and +0.03). Difference vs. means plots suggest stable differences across the range of measurement. Conclusions VCDR estimations were comparable using optomap® images and VISUCAM® PRO NM stereo optic disc images.

SLOctolyzer: Fully Automatic Analysis Toolkit for Segmentation and Feature Extracting in Scanning Laser Ophthalmoscopy Images.

The purpose of this study was to introduce SLOctolyzer: an open-source analysis toolkit for en face retinal vessels in infrared reflectance scanning laser ophthalmoscopy (SLO) images. SLOctolyzer includes two main modules: segmentation and measurement. The segmentation module uses deep learning methods to delineate retinal anatomy, and detects the fovea and optic disc, whereas the measurement module quantifies the complexity, density, tortuosity, and caliber of the segmented retinal vessels. We evaluated the segmentation module using unseen data and measured its reproducibility. SLOctolyzer's segmentation module performed well against unseen internal test data (Dice for all-vessels = 0.91; arteries = 0.84; veins = 0.85; optic disc = 0.94; and fovea = 0.88). External validation against severe retinal pathology showed decreased performance (Dice for arteries = 0.72; veins = 0.75; and optic disc = 0.90). SLOctolyzer had good reproducibility (mean difference for fractal dimension = -0.001; density = -0.0003; caliber = -0.32 microns; and tortuosity density = 0.001). SLOctolyzer can process a 768 × 768 pixel macula-centered SLO image in under 20 seconds and a disc-centered SLO image in under 30 seconds using a laptop CPU. To our knowledge, SLOctolyzer is the first open-source tool to convert raw SLO images into reproducible and clinically meaningful retinal vascular parameters. It requires no specialist knowledge or proprietary software, and allows manual correction of segmentations and re-computing of vascular metrics. SLOctolyzer is freely available at https://github.com/jaburke166/SLOctolyzer. SLO images are captured simultaneous to optical coherence tomography (OCT), and we believe SLOctolyzer will be useful for extracting retinal vascular measurements from large OCT image sets and linking them to ocular or systemic diseases.

Large animal model species in pluripotent stem cell therapy research and development for retinal diseases: a systematic review.

Retinal cell therapy modalities, in the category of advanced therapy medicinal products (ATMPs), are being developed to target several retinal diseases. Testing in large animal models (LAMs) is a crucial step in translating retinal ATMPs into clinical practice. However, challenges including budgetary and infrastructure constraints can hinder LAM research design and execution. Here, to facilitate the comparison of the various LAMs in pluripotent retinal cell therapy research, we aimed to systematically evaluate the species distribution, reported scientific utility, and methodology of a range of LAMs. A systematic search using the words retina, stem cell, transplantation, large animal, pig, rabbit, dog, and nonhuman primate was conducted in the PubMed, Embase, Science Direct and GoogleScholar databases in February 2023. We included 22 studies involving pluripotent stem cells (induced pluripotent stem cells or human embryonic stem cells) in LAMs, including non-human primates (NHP), pigs, dogs, and rabbits. Nearly half of the studies utilized wild-type animal models. In other studies, retinal degeneration features were simulated via laser, chemical, or genetic insult. Transplants were delivered subretinally, either as cell suspensions or pre-formed monolayers (with or without biodegradable scaffolding). The transplanted cells dose per eye varied widely (40,000 - 4,000,000 per dose). Cells were delivered via vitrectomy surgery in 15 studies and by an "ab externo" approach in one study. Structural outcomes were assessed using confocal scanning laser ophthalmoscopy imaging. Functional outcomes included multifocal electroretinogram and, in one case, a measure of visual acuity. Generally, cell suspension transplants exhibited low intraretinal incorporation, while monolayer transplants incorporated more efficiently. Immune responses posed challenges for allogeneic transplants, suggesting that autologous iPSC-derived transplants may be required to decrease the likelihood of rejection. The use of appropriate LAMs helps to advance the development of retinal ATMPs. The anatomical similarity of LAM and human eyes allows the implementation of clinically-relevant surgical techniques. While the FDA Modernization Act 2.0 has provided a framework to consider alternative methods including tissue-on-a-chip and human cell culture models for pharmacologic studies, LAM testing remains useful for cell and tissue replacement studies to inform the development of clinical trial protocols.

Imaging the Area of Internal Limiting Membrane Peeling after Macular Hole Surgery.

Background: The aim of this study was to compare en-face optical coherence tomography (OCT) imaging and confocal scanning laser ophthalmoscopy (cSLO) imaging at different wavelengths to identify the internal limiting membrane (ILM) peeling area after primary surgery with vitrectomy and ILM peeling for macular hole (MH). Methods: In total, 50 eyes of 50 consecutive patients who underwent primary surgery with vitrectomy and ILM peeling for MH were studied. The true ILM rhexis based on intraoperative color fundus photography was compared to the presumed ILM rhexis identified by a blinded examiner using en-face OCT imaging and cSLO images at various wavelengths. To calculate the fraction of overlap (FoO), the common intersecting area and the total of both areas were measured. Results: The FoO for the measured areas was 0.93 ± 0.03 for en-face OCT, 0.76 ± 0.06 for blue reflectance (BR; 488 nm), 0.71 ± 0.09 for green reflectance (GR; 514 nm), 0.56 ± 0.07 for infrared reflectance (IR; 815 nm) and 0.73 ± 0.06 for multispectral (MS). The FoO in the en-face OCT group was significantly higher than in all other groups, whereas the FoO in the IR group was significantly lower compared to all other groups. No significant differences were observed in FoO among the MS, BR, and GR groups. In en-face OCT, there was no significant change in the ILM peeled area measured intraoperatively and postoperatively (8.37 ± 3.01 vs. 8.24 ± 2.81 mm2; p = 0.8145). Nasal-inferior foveal displacement was observed in 38 eyes (76%). Conclusions: En-face OCT imaging demonstrates reliable postoperative visualization of the ILM peeled area. Although the size of the ILM peeling remains stable after one month, our findings indicate a notable inferior-nasal shift of the overall ILM peeling area towards the optic disc.

Blue Wavelength of Scanning Laser Ophthalmoscope Potentially Detects Arteriosclerotic Lesions in Diabetic Retinopathy.

(1) Background: The fundus examination is one of the best and popular methods in the assessment of vascular status in the human body. Direct viewing of retinal vessels by ophthalmoscopy has been utilized in judging hypertensive change or arteriosclerosis. Recently, fundus imaging with the non-mydriatic scanning laser ophthalmoscope (SLO) has been widely used in ophthalmological clinics since it has multimodal functions for optical coherence tomography or angiography with contrast agent dye. The purpose of this study was to examine the utility in detecting arteriosclerosis of retinal vessels in SLO images; (2) Methods: Both color and blue standard field SLO images of eyes with diabetic retinopathy (DR) were examined retrospectively. Retinal arteriosclerosis in color SLO images was graded according to the Scheie classification. Additionally, characteristics of retinal arterioles in blue SLO images were identified and examined for their relevance to arteriosclerosis grades, stages of DR or general complications; (3) Results: Relative to color fundus images, blue SLO images showed distinct hyper-reflective retinal arterioles against a monotone background. Irregularities of retinal arterioles identified in blue SLO images were frequently observed in the eyes of patients with severe arteriosclerosis (Grade 3: 79.0% and Grade 4: 81.8%). Furthermore, the findings on arterioles were more frequently associated with the eyes of DR patients with renal dysfunction (p < 0.05); (4) Conclusions: While color SLO images are equally as useful in assessing retinal arteriosclerosis as photography or ophthalmoscopy, the corresponding blue SLO images show arteriosclerotic lesions with high contrast in a monotone background. Retinal arteriosclerosis in eyes of advanced grades or advanced DR frequently show irregularities of retinal arterioles in the blue images. The findings of low, uneven, or discontinuous attenuation were easier to find in blue than in color SLO images. Consequently, blue SLO images can show pathological micro-sclerosis in retinal arterioles and are potentially one of the safe and practical methods for the vascular assessment of diabetic patients.

Discrimination of multiple sclerosis using scanning laser ophthalmoscopy images with autoencoder-based feature extraction

ObjectiveOptical coherence tomography (OCT) investigations have revealed that the thickness of inner retinal layers becomes decreased in multiple sclerosis (MS) patients, compared to healthy control (HC) individuals. To date, a number of studies have applied machine learning to OCT thickness measurements, aiming to enable accurate and automated diagnosis of the disease. However, there have much less emphasis on other less common retinal imaging modalities, like infrared scanning laser ophthalmoscopy (IR-SLO), for classifying MS. IR-SLO uses laser light to capture high-resolution fundus images, often performed in conjunction with OCT to lock B-scans at a fixed position. MethodsWe incorporated two independent datasets of IR-SLO images from the Isfahan and Johns Hopkins centers, consisting of 164 MS and 150 HC images. A subject-wise data splitting approach was employed to ensure that there was no leakage between training and test datasets. Several state-of-the-art convolutional neural networks (CNNs), including VGG-16, VGG-19, ResNet-50, and InceptionV3, and a CNN with a custom architecture were employed. In the next step, we designed a convolutional autoencoder (CAE) to extract semantic features subsequently given as inputs to four conventional ML classifiers, including support vector machine (SVM), k-nearest neighbor (K-NN), random forest (RF), and multi-layer perceptron (MLP). ResultsThe custom CNN (85 % accuracy, 85 % sensitivity, 87 % specificity, 93 % area under the receiver operating characteristics [AUROC], and 94 % area under the precision-recall curve [AUPRC]) outperformed state-of-the-art models (84 % accuracy, 83 % sensitivity, 87 % specificity, 92 % AUROC, and 94 % AUPRC); however, utilizing a combination of the CAE and MLP yields even superior results (88 % accuracy, 86 % sensitivity, 91 % specificity, 94 % AUROC, and 95 % AUPRC). ConclusionsWe utilized IR-SLO images to differentiate between MS and HC eyes, with promising results achieved using a combination of CAE and MLP. Future multi-center studies involving more heterogenous data are necessary to assess the feasibility of integrating IR-SLO images into routine clinical practice.

Algorithm of automatic identification of diabetic retinopathy foci based on ultra-widefield scanning laser ophthalmoscopy.

To propose an algorithm for automatic detection of diabetic retinopathy (DR) lesions based on ultra-widefield scanning laser ophthalmoscopy (SLO). The algorithm utilized the FasterRCNN (Faster Regions with CNN features)+ResNet50 (Residua Network 50)+FPN (Feature Pyramid Networks) method for detecting hemorrhagic spots, cotton wool spots, exudates, and microaneurysms in DR ultra-widefield SLO. Subimage segmentation combined with a deeper residual network FasterRCNN+ResNet50 was employed for feature extraction to enhance intelligent learning rate. Feature fusion was carried out by the feature pyramid network FPN, which significantly improved lesion detection rates in SLO fundus images. By analyzing 1076 ultra-widefield SLO images provided by our hospital, with a resolution of 2600×2048 dpi, the accuracy rates for hemorrhagic spots, cotton wool spots, exudates, and microaneurysms were found to be 87.23%, 83.57%, 86.75%, and 54.94%, respectively. The proposed algorithm demonstrates intelligent detection of DR lesions in ultra-widefield SLO, providing significant advantages over traditional fundus color imaging intelligent diagnosis algorithms.

Multispectral Fundus Photography of Choroidal Nevi With Trans-Palpebral Illumination.

The purpose of this study was to investigate the spectral characteristics of choroidal nevi and assess the feasibility of quantifying the basal diameter of choroidal nevi using multispectral fundus images captured with trans-palpebral illumination. The study used a widefield fundus camera with multispectral (625 nm, 780 nm, 850 nm, and 970 nm) trans-palpebral illumination to examine eight subjects diagnosed with choroidal nevi. Geometric features of nevi, including border clarity, overlying drusen, and lesion basal diameter, were characterized. Clinical imagers, including scanning laser ophthalmoscopy (SLO), autofluorescence (AF), and optical coherence tomography (OCT), were utilized for comparative assessment. Fundus images depicted nevi as dark regions with high contrast against the background. Near-infrared (NIR) fundus images provided enhanced visibility of lesion borders compared to visible fundus images and SLO images. Lesion-background contrast measurements revealed 635 nm SLO at 11% and 625 nm fundus at 42%. Significantly enhanced contrasts were observed in NIR fundus images at 780 nm (73%), 850 nm (63%), and 970 nm (67%). For quantifying the diameter of nevi, NIR fundus images at 780 nm and 850 nm yielded a deviation of less than 10% when compared to OCT measurements. NIR fundus photography with trans-palpebral illumination enhances nevi visibility and boundary definition compared to SLO. Agreement in diameter measurements with OCT validates the accuracy and reliability of this method for choroidal nevi assessment. Multispectral fundus imaging with trans-palpebral illumination improves choroidal nevi visibility and accurately measures basal diameter, promising to enhance clinical practices in screening, diagnosis, and monitoring of choroidal nevi.

LONGITUDINAL ADAPTIVE OPTICS SCANNING LASER OPHTHALMOSCOPY REVEALS REGIONAL VARIATION IN CONE AND ROD PHOTORECEPTOR LOSS IN STARGARDT DISEASE.

To investigate the temporal sequence of changes in the photoreceptor cell mosaic in patients with Stargardt disease type 1, using adaptive optics scanning laser ophthalmoscopy. Two brothers with genetically confirmed Stargardt disease type 1 underwent comprehensive eye exams, spectral-domain optical coherence tomography, fundus autofluorescence, and adaptive optics scanning laser ophthalmoscopy imaging 3 times over the course of 28 months. Confocal images of the cones and rods were obtained from the central fovea to 10° inferiorly. Photoreceptors were counted in sampling windows at 100- µ m intervals of 200 µ m × 200 µ m for cones and 50 µ m × 50 µ m for rods, using custom cell marking software with manual correction. Photoreceptor density and spacing were measured and compared across imaging sessions using one-way analysis of variance. Adaptive optics scanning laser ophthalmoscopy revealed the younger brother had a 30% decline in foveal cone density after 8 months, followed by complete loss of foveal cones at 28 months; the older brother had no detectable foveal cones at baseline. In the peripheral macula, cone and rod spacings were greater than normal in both patients. The ratio of the cone spacing to rod spacing was greater than normal across all eccentricities, with a greater divergence closer to the foveal center. Cone cell loss may be an early pathogenetic step in Stargardt disease. Adaptive optics scanning laser ophthalmoscopy provides the capability to track individual photoreceptor changes longitudinally in Stargardt disease.

Quantitative Analysis of the Vasculature and Cone Photoreceptors in Subjects With Diabetes Without Diabetic Retinopathy

Purpose To characterize any differences in the vasculature and cone photoreceptor packing geometry (CPG) between subjects with diabetes without/no diabetic retinopathy (NDR) and healthy controls. Methods Eight NDR and five controls were enrolled. Optical coherence tomography angiography (OCTA) taken at the macula was used to measure vessel density, vessel length density, and vessel density index (VDI) in three vascular plexuses, namely, the superficial vascular plexus, intermediate capillary plexus, and deep capillary plexus (DCP). The choriocapillaris (CC) flow deficit (FD) was also measured. OCTA images were binarized and processed to extrapolate the parafovea and parafoveal quadrants and the OCTA indices mentioned above. The CC was processed with six different radii to quantify FD. Adaptive optics - scanning laser ophthalmoscopy images were acquired and processed to extract CPG indices, i.e., cone density (CD), cone-to-cone spacing (CS), linear dispersion index, heterogeneity packing index and percent of cells with six neighbors at 3.6° in the temporal retina. Results In all eyes, statistically significant differences were found (i) in parafoveal FD across the six radii (p < 0.001) and (ii) in the correlation between the parafoveal temporal quadrant (PTQ) DCP VDI and CS (r = 0.606, p = 0.048). No other significant correlations were found. For OCTA or CPG indices, no significant differences were found between the cohorts in the parafovea or parafoveal quadrants. Conclusions CS is the most sensitive CPG index for detecting alterations in the cone mosaic. The DCP and the cone photoreceptors are significantly correlated, indicating that alterations in the DCP can affect the cones. Future work elucidating the vascular alterations and neurodegeneration present in diabetic eyes should focus on the DCP and multiple CPG indices, not solely CD. Moreover, such alterations are highly localized, hence using larger regions e.g. parafovea versus smaller areas, such as the PTQ, will potentially mask significant correlations.

What artificial intelligence can bring us in the diagnosis and progression of glaucoma

Aims/Purpose: Our goal is to focus on the contribution of artificial intelligence in glaucoma from diagnosis to progression.Methods: By a literature review, we will review work using AI in the field of glaucoma, whether for screening, diagnosis or detection of progression.Results: For the diagnosis: Recently, two papers were published reporting results of DL algorithms to diagnose glaucoma from Visual Field data. Several authors have evaluated AI‐based fundus photograph analysis for its utility for detecting glaucoma such as the study of Liu et al which grouped 241 032 images of 68 013 patients with a sensitivity of 96.2% and specificity of 87.2% also for analysing OCT imaging data from peripapillary RNFL thickness maps and the macular ganglion cell complex for discriminating between glaucomatous and normal eyes with AROC values ranging from 0.69 to 0.99. Prediction of IOP trends from previous data and medications would be a useful and plausible use of AI. For the progression of glaucoma: Multiclass support vector machines (SVMs), have been used to simultaneously discriminate between normal, non progressing, and progressing eyes through the analysis of confocal scanning laser ophthalmoscopy (CSLO) images with a correct classification rate of 88% .Conclusions: Many AI strategies have shown promising results for glaucoma detection using fundus photography, optical coherence tomography, or automated perimetry. The combination of these imaging modalities increases the performance of AI algorithms, with results comparable to those of humans. Research in the coming years will need to address unavoidable questions regarding the clinical significance of such results and the explicability of the predictions.

Application of artificial intelligence in glaucoma

Glaucoma is the mainly cause of irreversible blindness characterized with structural changes of optic nerve head, retinal ganglion cell death and loss of visual acuity, with estimated 64.3 million patients aged from 40 to 80 worldwide. When patients seek medical advice because of visual acuity, glaucoma often has advanced to late stage. One major challenge is to identify large number of undiagnosed patients. The reason why screening programs are not employed regularly is that misdiagnoses can be made due to large amounts of false positives in the procedure.Artificial intelligence (AI) has brought new breakthroughs for automated screening for glaucoma，which can perform a task by classifying and identifying data without human intervention, including supervised and unsupervised machine learning two categories. Due to the limitation of manually designed features and disability in detection for high dimensionality of optic disc images, advances in techniques for evaluation of glaucoma came with the use of segmentation‐free approaches. We reported a deep learning system and assessed its generalizability in various data sets, with similar patterns in sensitivity (82.2%–96.1%) and specificity (70.4%–97.1%).Visual field data presents low dimensionality and high noise compared to optic disc images, whose hidden patterns can be detected by unsupervised learning better. Two most reported unsupervised algorithms were clustering and component analysis. Such technologies adopted in glaucoma screening have significant potential to increase accessibility in remote areas and decrease the cost for widely screening. Further studies similarly demonstrated that machine learning can perform comparable to or over human experts in mean deviation, pattern standard deviation and glaucoma hemifield test. A back‐propagation neural network was even reported to successfully detect visual field progression with Area Under Curve (AUC) of 0.92. Further advances were also made through algorithms based on CNN showing both higher sensitivity and specificity than traditional machine learning method.In terms of optical coherence tomography (OCT), early studies based on MLC using TD‐OCT showed an accuracy not worse than non‐artificial intelligence analysis. The latest OCT technologies, including SD‐OCT and SS‐OCT, combined with deep learning are reported more sensitive for early glaucoma detection, with a high AUC up to 0.937. Recently a study reported the ability of their model to detect RNFL thickness from OCT with the AUC of 0.944. Recent years, more parameters have been put in including OCT RNFL thickness, standard automated perimetry and confocal scanning laser ophthalmoscopy imaging, making an improvement in glaucoma detection.However, there remains several limitations and further studies is required. First, the exact mechanism how DL algorithms evaluate features and make predictions remains unknown, which is called ‘black boxes’. Another limitation is that it can be difficult for DL to classify glaucoma in eyes with less severe disease manifestations or multiple comorbid eye conditions, especially high myopia, which asks for a large images database. Furthermore, in order to develop a more dependent screening method, other clinical parameters including intraocular pressure and central corneal thickness should be integrated, and application and validation of such in a real‐world screening need additional research to support.

Foveal cone loss in tamoxifen maculopathy: a case report

BackgroundTamoxifen is used in low dose concentrations (20–40 mg per day) as a therapy for breast cancer but is known to have ocular side effects. In this case report, the foveal cone integrity in a tamoxifen-treated patient who complained of a small central scotoma in the left eye while reading was examined using high resolution adaptive optics imaging.Case presentationBoth eyes of a 54-year-old Caucasian, non-hispanic female who had been treated with tamoxifen for 1.5 years were examined using various imaging modalities including fundus photography, fundus autofluorescence, fluorescein angiography, spectral-domain optical coherence tomography, and adaptive optics scanning laser ophthalmoscopy. Clinical spectral-domain optical coherence tomography showed a very small disruption to the photoreceptor layer at the fovea in the left eye only. However, adaptive optics scanning laser ophthalmoscopy imaging revealed foveal cone loss in both eyes, but to a lesser extent in the right eye. Inner retinal changes were not observed in either eye.ConclusionThe area of cone loss was similar in size to a single newsprint letter when projected onto the retina, matching the patient’s description of a scotoma in the left eye. Given the isolated loss of foveal cone photoreceptors with the absence of previously reported inner retinal and vascular changes, our results may indicate the earliest retinal changes associated with tamoxifen retinopathy.

Multi-scale Bottleneck Residual Network for Retinal Vessel Segmentation.

Precise segmentation of retinal vessels is crucial for the prevention and diagnosis of ophthalmic diseases. In recent years, deep learning has shown outstanding performance in retinal vessel segmentation. Many scholars are dedicated to studying retinal vessel segmentation methods based on color fundus images, but the amount of research works on Scanning Laser Ophthalmoscopy (SLO) images is very scarce. In addition, existing SLO image segmentation methods still have difficulty in balancing accuracy and model parameters. This paper proposes a SLO image segmentation model based on lightweight U-Net architecture called MBRNet, which solves the problems in the current research through Multi-scale Bottleneck Residual (MBR) module and attention mechanism. Concretely speaking, the MBR module expands the receptive field of the model at a relatively low computational cost and retains more detailed information. Attention Gate (AG) module alleviates the disturbance of noise so that the network can concentrate on vascular characteristics. Experimental results on two public SLO datasets demonstrate that by comparison to existing methods, the MBRNet has better segmentation performance with relatively few parameters.

Invited Session V: The eye as a window to systemic and neurodegenerative health: Early prediction of multiple sclerosis using scanning laser ophthalmoscopy (SLO) video sequence data with a Deep Learning (DL) based approach.

Multiple Sclerosis (MS) is a chronic immune-mediated inflammatory disease (IMID) of the central nervous system (CNS). Early identification of MS, especially as a screening method for at-risk individuals, is crucial to delay disease progression and improve patient outcomes by preventing future irreversible neurologic damage. In this work, we utilize well-validated tracking scanning laser ophthalmoscope (TSLO) image to predict MS compared to the unaffected controls. While traditional Machine Learning (ML) methods, such as Logistic Regression (LR), have demonstrated a strong predictive power [Mauro F. Pinto et al., 2020] in disease identification, we propose the use of a novel DL based model. Though the use of Deep Neural Network (DNN), this model can have a much higher learning capacity to capture latent features embedded in the retinal images. We hypothesize that such latent information, often hidden in ML feature engineering processes, plays an important role for the prediction of disease and can be well represented by DL models. To establish a DL based model capable of learning latent image features to provide predictive power for the presence of MS. Utilize a deep convolutional neural network to extract the retinal coding and implement a recurrent neural network to learn the temporal correlations in video sequences. Our approaches were tested using a 250-subject MS/control database collected at the UCSF. Patients with Expanded Disability Status Scale (EDSS)< 4 are compared to healthy subjects. Both raw retinal images and the frequency and spatial patterns of the eye motion are combined to construct a hybrid image, denoted as "retinal coding", and directly fed to the DL model for training and testing. Preliminary results on predictive power were measured using Area-under-Curve (AUC) of the Receiver Operating Characteristics (ROC) curve, sensitivity, and specificity, as well as an F-1 score. We observe an AUC of 0.920, sensitivity of 0.90, specificity of 0.89 and F-1 score of 0.89 using the DL model to distinguish MS from controls, which outperforms the baseline LR model by 24%. This work can be considered as a proof-of-concept concerning the possibility of identifying MS disease using a DL based approach. The results demonstrate the possibility of predicting early-stage MS and understanding disease's dynamics. Such end-to-end model could be generalizable and trained on other disease states.

Retinal Artery Angles in High Axial Myopia and Its Relationship With Visual Function.

To evaluate the retinal artery angles in high axial myopia and assess the correlation with other morphometric and functional parameters. This cross-sectional study included 112 eyes of 112 patients with high axial myopia. Based on axial length (AL), the participants were divided into three groups: group 1 (26 ≤ AL < 28 mm), group 2 (28 ≤ AL < 31 mm), and group 3 (≥31 mm). Scanning laser ophthalmoscopy imaging was used to analyze the retinal artery angle (Yugami correlated angle [YCA]). Retinal vascular densities (VDs) in both superficial capillary plexuses (SCPs) and deep capillary plexuses were evaluated. Fixation behavior, including retinal mean sensitivity (MS), macular fovea 2°, 4° fixation rate (P1, P2), and 68.2% bivariate contour ellipse area, were analyzed by microperimetry. Finally, the correlation between YCAs and AL, VDs, best-corrected visual acuity (BCVA), and fixation behavior was assessed. The YCAs showed significant differences among the three groups (all P < 0.001, respectively). Compared to group 1, the YCA decreased in group 2 (P < 0.001) and continued to decrease in group 3 (P = 0.043). The correlation analysis revealed that smaller YCAs (YCA, YCA1/2, YCA1/4) were positively correlated with the longer AL (ρ = 0.580, 0.545, 0.448, P < 0.001) and lower VDs in any sector in SCPs (all P ≤ 0.05). Furthermore, smaller YCAs were positively correlated with decreased BCVA (ρ = 0.392, 0.387, 0.262; all P < 0.001) and reduced MS (ρ= 0.300, 0.269, 0.244; all P < 0.05). Smaller YCAs were correlated with longer AL, lower VD in SCP, decreased BCVA, and reduced MS. The YCAs might reflect vascular deformation caused by axial elongation and could potentially be useful in predicting visual function in high axial myopia. The quantitative analysis of YCAs in fundus photography holds potential clinical value in predicting visual function in high axial myopia.

THE WAY PATIENTS SEE FLOATERS: Widefield Dynamic Scanning Laser Ophthalmoscopy Imaging of Vitreous Abnormalities.

To investigate the use of dynamic widefield scanning laser ophthalmoscopy (SLO) and B-scan ultrasonography in imaging vitreous abnormalities in patients with complaints of floaters. Twenty-one patients underwent both dynamic SLO and B-scan ultrasonography to image their vitreous abnormalities. After reviewing these videos, patients graded each imaging technique on a scale of 1 to 10, based on how closely it represented their visual perception of floaters. The mean age of the patients (12 women and nine men) was 47.7 ± 18.5 years. The patients graded a median score of nine for SLO imaging (mean = 8.43) compared with a median score of 5 (mean = 4.95) for ultrasound ( P = 0.001). Widefield SLO imaging demonstrated three-dimensional interconnectivity within the condensations of the formed vitreous that exhibited translational and rotational movements with eye saccades. Floaters are a common complaint, but it is difficult to know whether imaging findings of the vitreous correlate to what patients perceive. Widefield SLO seems to image vitreous abnormalities related to how patients perceive their own floaters better than B-scan ultrasonography. Despite the term "floaters", the vitreous abnormalities in the videos seemed to be manifestations of a complex three-dimensional degeneration of the vitreous framework.

Multimodal Deep Learning Classifier for Primary Open-angle Glaucoma Diagnosis Using Wide-field Optic Nerve Head Cube Scans in Eyes With and Without High Myopia.

An OCT-based multimodal deep learning classification model including texture information is introduced that outperforms single-modal models and multimodal models without texture information for glaucoma diagnosis in eyes with and without high myopia. To evaluate the diagnostic accuracy of a multimodal deep learning (DL) classifier using wide optical coherence tomography (OCT) optic nerve head cube scans in eyes with and without axial high myopia. 371 primary open-angle glaucoma (POAG) eyes and 86 healthy eyes, all without axial high myopia (axial length AL≤26mm) and 92 POAG eyes and 44 healthy eyes, all with axial high myopia (AL>26mm) were included. The multimodal DL classifier combined features of 3 individual VGG-16 models: 1) texture-based en face image, 2) retinal nerve fiber layer (RNFL) thickness map image and 3) confocal scanning laser ophthalmoscope (cSLO) image. Age, AL, and disc area adjusted area under the receiver operating curves (AUROC) were used to compare model accuracy. Adjusted AUROC for the multimodal DL model was 0.91 (95% CI = 0.87, 0.95). This value was significantly higher than the values of individual models (0.83 [0.79, 0.86] for texture based en face image; 0.84 [0.81, 0.87] for RNFL thickness map; and 0.68 [0.61, 0.74] for cSLO image; all P≤0.05). Using only highly myopic eyes, the multimodal DL model showed significantly higher diagnostic accuracy (0.89 [0.86 , 0.92]) compared to texture en face image (0.83 [0.78 , 0.85]), RNFL (0.85 [0.81 , 0.86]) and cSLO image models (0.69 [0.63 , 0.76]) (all P≤0.05). Combining OCT-based RNFL thickness maps with texture based en face images showed better ability to discriminate between healthy and POAG than thickness maps alone, particularly in high axial myopic eyes.

Retinal Vessel Caliber Measurement Bias in Fundus Images in the Presence of the Central Light Reflex.

To investigate the agreement between a fundus camera and a scanning laser ophthalmoscope in retinal vessel caliber measurements and to identify whether the presence of the central light reflex (CLR) explains potential discrepancies. For this cross-sectional study, we obtained fundus camera and scanning laser ophthalmoscope images from 85 eyes of 85 healthy individuals (aged 50-65 years) with different blood pressure status. We measured the central retinal artery equivalent (CRAE) and central retinal artery vein equivalent (CRVE) with the Knudtson-Parr-Hubbard algorithm and assessed the CLR using a semiautomatic grading method. We used Bland-Altman plots, 95% limits of agreement, and the two-way mixed effects intraclass correlation coefficient for consistency [ICC(3,1)] to describe interdevice agreement. We used multivariable regression to identify factors associated with differences in between-device measurements. The between-device difference in CRAE (9.5 µm; 95% confidence interval, 8.0-11.1 µm) was larger than the between-device difference in CRVE (2.9 µm; 95% confidence interval, 1.3-4.5 µm), with the fundus camera yielding higher measurements (both P < 0.001). The 95% fundus camera-scanning laser ophthalmoscope limits of agreement were -4.8 to 23.9 µm for CRAE and -12.0 to 17.8 µm for CRVE. The corresponding ICCs(3,1) were 0.89 (95% confidence interval, 0.83-0.92) and 0.91 (95% confidence interval, 0.86-0.94). The between-device CRAE difference was positively associated with the presence of a CLR (P = 0.002). Fundus cameras and scanning laser ophthalmoscopes yield correlated but not interchangeable caliber measurements. The CLR induces bias in arteriolar caliber in fundus camera images, compared with scanning laser ophthalmoscope images. Refined measurements could yield better estimates of the association between retinal vessel caliber and ophthalmic or systemic disease.