Imaging Of Human Retina Research Articles

Event Stream Denoising Method Based on Spatio-Temporal Density and Time Sequence Analysis.

An event camera is a neuromimetic sensor inspired by the human retinal imaging principle, which has the advantages of high dynamic range, high temporal resolution, and low power consumption. Due to the interference of hardware and software and other factors, the event stream output from the event camera usually contains a large amount of noise, and traditional denoising algorithms cannot be applied to the event stream. To better deal with different kinds of noise and enhance the robustness of the denoising algorithm, based on the spatio-temporal distribution characteristics of effective events and noise, an event stream noise reduction and visualization algorithm is proposed. The event stream enters fine filtering after filtering the BA noise based on spatio-temporal density. The fine filtering performs time sequence analysis on the event pixels and the neighboring pixels to filter out hot noise. The proposed visualization algorithm adaptively overlaps the events of the previous frame according to the event density difference to obtain clear and coherent event frames. We conducted denoising and visualization experiments on real scenes and public datasets, respectively, and the experiments show that our algorithm is effective in filtering noise and obtaining clear and coherent event frames under different event stream densities and noise backgrounds.

Polarization-insensitive optical coherence tomography using polarization maintaining fiber with a simple optical configuration

Compensation of polarization-variance-related artifacts is required to steadily obtain high-quality optical coherence tomography (OCT) images at various experimental conditions. Since most OCT systems utilize optical fiber to transfer the light easily and a polarized light source, the polarization state is arbitrarily changed in every different condition. In this study, we proposed polarization-maintaining-fiber-based polarization-insensitive OCT (PM-PI-OCT) with a simple optical configuration and a simple compensation process. The proposed PM-PI-OCT is not only theoretically proved by mathematical derivations but also evaluated by quantitative analysis of various fiber twisting angles. Moreover, the applicability and robustness of the proposed PM-PI-OCT were proved by human retina imaging using the customized handheld probe. Our proposed polarization-insensitive OCT requires no pre-calibration, no post-processing procedure, and no computational load for implementation and is able to be applied to universal fiber-based OCT systems. We believe that our simple and robust polarization-insensitive OCT system is able to be applied to various existing OCT setups for polarization state variance compensation with high versatility and applicability.

Optimization of handheld spectrally encoded coherence tomography and reflectometry for point-of-care ophthalmic diagnostic imaging.

Handheld optical coherence tomography (HH-OCT) systems enable point-of-care ophthalmic imaging in bedridden, uncooperative, and pediatric patients. Handheld spectrally encoded coherence tomography and reflectometry (HH-SECTR) combines OCT and spectrally encoded reflectometry (SER) to address critical clinical challenges in HH-OCT imaging with real-time en face retinal aiming for OCT volume alignment and volumetric correction of motion artifacts that occur during HH-OCT imaging. We aim to enable robust clinical translation of HH-SECTR and improve clinical ergonomics during point-of-care OCT imaging for ophthalmic diagnostics. HH-SECTR is redesigned with (1)optimized SER optical imaging for en face retinal aiming and retinal tracking for motion correction, (2)a modular aluminum form factor for sustained alignment and probe stability for longitudinal clinical studies, and (3)one-handed photographer-ergonomic motorized focus adjustment. We demonstrate an HH-SECTR imaging probe with micron-scale optical-optomechanical stability and use it for in vivo human retinal imaging and volumetric motion correction. This research will benefit the clinical translation of HH-SECTR for point-of-care ophthalmic diagnostics.

Optic Disc Segmentation in Human Retina Images Using a Meta Heuristic Optimization Method and Disease Diagnosis with Deep Learning

Glaucoma is a common eye disease that damages the optic nerve and leads to loss of vision. The disease shows few symptoms in the early stages, making its identification a complex task. To overcome the challenges associated with this task, this study aimed to tackle the localization and segmentation of the optic disc, as well as the classification of glaucoma. For the optic disc segmentation, we propose a novel metaheuristic approach called Grey Wolf Optimization (GWO). Two different approaches are used for glaucoma classification: a one-stage approach, in which the whole image without cropping is used for classification, and a two-stage approach. In the two-stage approach, the optic disc region is detected using the You Only Look Once (YOLO) detection algorithm. Once the optic disc region of interest (ROI) is identified, glaucoma classification is performed using pre-trained convolutional neural networks (CNNs) and vision transformation techniques. In addition, both the one-stage and the two-stage approaches are applied in combination with the pre-trained CNN using the Random Forest algorithm. In segmentation, GWO achieved an average sensitivity of 96.04%, a specificity of 99.58%, an accuracy of 99.39%, a DICE coefficient of 94.15%, and a Jaccard index of 90.4% on the Drishti-GS dataset. For classification, the proposed method achieved remarkable results with a high-test accuracy of 100% and 88.18% for hold-out validation and three-fold cross-validation for the Drishti-GS dataset, and 96.15% and 93.84% for ORIGA with hold-out and five-fold cross-validation, respectively. Comparing the results with previous studies, the proposed CNN model outperforms them. In addition, the use of the Swin transformer shows its effectiveness in classifying glaucoma in different subsets of the data.

Imaging of human parafoveal area with large field of view in adaptive optics line scanning ophthalmoscope

The parafoveal area, with its high concentration of photoreceptors and fine retinal capillaries, is crucial for central vision and often exhibits early signs of pathological changes. The current adaptive optics scanning laser ophthalmoscope (AOSLO) provides an excellent tool to acquire accurate and detailed information about the parafoveal area with cellular resolution. However, limited by the scanning speed of two-dimensional scanning, the field of view (FOV) in the AOSLO system was usually less than or equal to 2∘, and the stitching for the parafoveal area required dozens of images, which was time-consuming and laborious. Unfortunately, almost half of patients are unable to obtain stitched images because of their poor fixation. To solve this problem, we integrate AO technology with the line-scan imaging method to build an adaptive optics line scanning ophthalmoscope (AOLSO) system with a larger FOV. In the AOLSO, afocal spherical mirrors in pairs are nonplanar arranged and the distance and angle between optical elements are optimized to minimize the aberrations, two cylinder lenses are orthogonally placed before the imaging sensor to stretch the point spread function (PSF) for sufficiently digitizing light energy. Captured human retinal images show the whole parafoveal area with [Formula: see text] FOV, 60[Formula: see text]Hz frame rate and cellular resolutions. Take advantage of the 5∘ FOV of the AOLSO, only 9 frames of the retina are captured with several minutes to stitch a montage image with an FOV of [Formula: see text], in which photoreceptor counting is performed within approximately 5∘ eccentricity. The AOLSO system not only provides cellular resolution but also has the capability to capture the parafoveal region in a single frame, which offers great potential for noninvasive studying of the parafoveal area.

Neural Network Algorithm for Biometric Analysis of Human Retina Image

Neural Network Algorithm for Biometric Analysis of Human Retina Image

A dual-channel visible light optical coherence tomography system enables wide-field, full-range, and shot-noise limited human retinal imaging

Visible light optical coherence tomography (VIS-OCT) is an emerging ophthalmic imaging method featuring ultrahigh depth resolution, retinal microvascular oximetry, and distinct scattering contrast in the visible spectral range. The clinical utility of VIS-OCT is hampered by the fundamental trade-off between the imaging depth range and axial resolution, which are determined by the spectral resolution and bandwidth, respectively. To address this trade-off, here we developed a dual-channel VIS-OCT system with three major advancements including the first linear-in-K VIS-OCT spectrometer to decrease the roll-off, reference pathlength modulation to expand the imaging depth range, and per-A-line noise cancellation to remove excess noise, Due to these unique designs, this system achieves 7.2 dB roll-off over the full 1.74 mm depth range (water) with shot-noise limited performance. The system uniquely enables >60° wide-field imaging which would allow simultaneous imaging of the peripheral retina and optic nerve head, as well as ultrahigh 1.3 µm depth resolution (water). Benefiting from the additional near-infrared (NIR) channel of the dual-channel design, this system is compatible with Doppler OCT and OCT angiography (OCTA). The comprehensive structure-function measurement enabled by this dual-channel VIS-OCT system is an advance towards adoption of VIS-OCT in clinical applications.

Multiwavelength laser doppler holography (MLDH) in spatiotemporal optical coherence tomography (STOC-T)

Spatiotemporal optical coherence tomography (STOC-T) is the novel modality for high-speed, crosstalk- and aberration-free volumetric imaging of biological tissue in vivo. STOC-T extends the Fourier-Domain holographic Optical Coherence Tomography by the spatial phase modulation that enables the reduction of spatial coherence of the tunable laser. By reducing the spatial coherence of the laser, we suppress coherent noise, and, consequently, improve the imaging depth. Furthermore, we remove geometrical aberrations computationally in postprocessing. We recently demonstrated high-speed, high-resolution STOC-T of human retinal imaging in vivo. Here, we show that the dataset produced by STOC-T can be processed differently to reveal blood flow in the human retina in vivo. To render the blood flow, we first pre-process STOC-T holographic data to access the approximated information about the Doppler-shifted optical field backscattered from the sample. Then, we analyze it using methods from the laser Doppler flowmetry, namely, by analyzing the Doppler broadening caused by moving light scatterers (red blood cells). However, contrary to conventional approaches, we use multiple illumination wavelengths. This enables us to render the structural volumetric and blood flow images from the same dataset concurrently. Our method, denoted as multiwavelength laser Doppler holography (MLDH), links laser Doppler flowmetry with multiwavelength holographic detection to enable noninvasive visualization and possible blood flow quantification at different human retina layers at high speeds and high transverse resolution in vivo.

Multi-level spatial-temporal and attentional information deep fusion network for retinal vessel segmentation

Objective. Automatic segmentation of fundus vessels has the potential to enhance the judgment ability of intelligent disease diagnosis systems. Even though various methods have been proposed, it is still a demanding task to accurately segment the fundus vessels. The purpose of our study is to develop a robust and effective method to segment the vessels in human color retinal fundus images. Approach. We present a novel multi-level spatial-temporal and attentional information deep fusion network for the segmentation of retinal vessels, called MSAFNet, which enhances segmentation performance and robustness. Our method utilizes the multi-level spatial-temporal encoding module to obtain spatial-temporal information and the Self-Attention module to capture feature correlations in different levels of our network. Based on the encoder and decoder structure, we combine these features to get the final segmentation results. Main results. Through abundant experiments on four public datasets, our method achieves preferable performance compared with other SOTA retinal vessel segmentation methods. Our Accuracy and Area Under Curve achieve the highest scores of 96.96%, 96.57%, 96.48% and 98.78%, 98.54%, 98.27% on DRIVE, CHASE_DB1, and HRF datasets. Our Specificity achieves the highest score of 98.58% and 99.08% on DRIVE and STARE datasets. Significance. The experimental results demonstrate that our method has strong learning and representation capabilities and can accurately detect retinal blood vessels, thereby serving as a potential tool for assisting in diagnosis.

Surface-plasmon-enhanced MoS2 multifunctional optoelectronic memory for emulating human retinal imaging

As one of the most important members of the two-dimensional (2D) chalcogenide family, MoS2 plays a fundamental role in the development of 2D electronic and optoelectronic designs. However, MoS2-based optoelectronic devices are hindered by their weak light–matter interactions, which make it challenging to achieve excellent device performance in photoelectronic memory applications. Here, we developed a multifunctional optoelectronic memory by coupling Au nanoparticles with MoS2, where the presence of Au nanoparticles on the surface significantly enhanced the light absorption capacity of MoS2 through the surface-plasmon-enhanced effect. The device achieved a photoresponse capability with a light current-to-dark current ratio exceeding 103, surpassing the majority of values reported for comparable photoconductive detectors. Importantly, it exhibits excellent light writing, storage, and erasuring capabilities, with a storage time exceeding 1000 s. Based on this device, a 3 × 3 array hardware core is designed to mimic human retinal imaging under the irradiation of 660, 532, and 457 nm lasers by using R-CNN algorithm, reducing power consumption, and redundancy. These advancements have the potential to drive future developments in neuromorphic electronics, particularly in optical information sensing and learning.

Adaptive balanced detection spectral domain optical coherence tomography.

Balanced detection optical coherence tomography (BD-OCT) enables near-shot noise-limited imaging by suppressing wavelength-dependent relative intensity noise (RIN) originating from the light source. In spectral-domain BD-OCT (SD-BD-OCT), the level of RIN suppression relies on the co-registration accuracy of the spectra simultaneously captured by two independent spectrometers. However, existing matching methods require careful pre-calibration using a RIN-dominated dataset or subjective post-processing using a signal-dominated dataset. We developed an adaptive subpixel matching approach, referred to as adaptive balance, that can be applied to any SD-BD-OCT dataset regardless of RIN or signal level without the need for pre-calibration. We showed that adaptive balance performed comparable to or better than reported methods by imaging phantoms with varying spectrometer camera gain, exposure time, and supercontinuum laser repetition rate. We further demonstrated the benefits of adaptive balance in human retinal imaging.

A neural network classifier for detecting diabetic retinopathy from retinal images

With the spread of diabetes mellitus, diabetic retinopathy (DR) is becoming a major public health problem (especially in developing countries). The long-term complications resulting from DR have a significant impact on patients. Early diagnosis and subsequent treatment can reduce the damage to health. Predictive analytics can be based on the analysis of human retinal images using convolutional neural networks. In this paper, the research focuses on the development of an efficient method for DR detection based on the EfficientNet convolutional neural network, self-learning technology and data augmentation operations. As a result of the experiments, a neural network classifier based on convolutional neural networks is developed, recommendations for data augmentation operations are given. Experiments were performed on the public dataset and showed that it is possible to achieve the proportion of correctly classified objects equal to 97.14 % on the test set from the public dataset.

Retinal blood flow speed quantification at the capillary level using temporal autocorrelation fitting OCTA [Invited

Optical coherence tomography angiography (OCTA) can visualize vasculature structures, but provides limited information about blood flow speed. Here, we present a second generation variable interscan time analysis (VISTA) OCTA, which evaluates a quantitative surrogate marker for blood flow speed in vasculature. At the capillary level, spatially compiled OCTA and a simple temporal autocorrelation model, ρ(τ) = exp(-ατ), were used to evaluate a temporal autocorrelation decay constant, α, as the blood flow speed marker. A 600 kHz A-scan rate swept-source OCT prototype instrument provides short interscan time OCTA and fine A-scan spacing acquisition, while maintaining multi mm2 field of views for human retinal imaging. We demonstrate the cardiac pulsatility and assess repeatability of α measured with VISTA. We show different α for different retinal capillary plexuses in healthy eyes and present representative VISTA OCTA in eyes with diabetic retinopathy.

Sonar glass—Artificial vision: Comprehensive design aspects of a synchronization protocol for vision based sensors

Supporting visually impaired people during their navigation is a challenging task that involves localization, tracking, navigation, obstacle avoidance, and path guidance. Many researchers have experimented with Sonar, RFID, GPS, NFC, walking sticks, waist-based devices, and even computer vision modules for blind navigation. Using five sonar sensors for obstacle detection with direction and timestamp, we developed Sonar Glass. As humans see in left, right, front, top, and bottom directions based on eye angle and the head pose, the sonar glass is designed to provide obstacle information at the same angle. The head movement of the visually impaired person activates a pair of sensors for each module. Understanding human eye movement mechanisms and developing a synchronization protocol for each pair of visual sensors on sonar glass sitting on both eyes is the main goal of this paper. The major challenge lies in understanding and simulating the human vision mechanism and to realize how the field of Artificial Intelligence can be a contributor in producing technologies for the visually impaired. We are using log-polar transform to simulate human retinal image mapping. The Scale Invariant Feature Transform (SIFT) algorithm has also been implemented. It is the first time that both human eyes can be replaced by vision sensors. After comparing the estimated obstacle information from one sensor pair with that of the other sensors, the voice track is activated. The blind person uses the nearest obstacle information to avoid the obstacle and to extract spatial information about obstacles ahead of the user and provide an early warning. Unique in its design, the sonar Glass’s synchronization protocol for a pair of related sensors provides possible object information in that direction. The executions were simulated in MATLAB and the results obtained in real time are found to be promising as the tests were carried out both indoor and outdoor. The sonar Glass design is unique of its kind and the synchronization protocol for the pair of related sensors provides the possible vision about the object information at that particular direction.

Classification of diabetic retinopathy stages based on neural networks

Diabetic retinopathy is one of the main side effects of diabetes, which causes severe effects, including blindness. The main challenge is the early diagnosis of this disease for timely and effective treatment. Diabetic retinopathy can be detected much faster and more accurately by using machine learning methods for image analyzing of the human retina. The development of methods and algorithms for the detection and classification of this disease, the automation of this process are the actual and costeffective goals.The article focuses on the classification of the stages of diabetic retinopathy using neural networks based on human retinal images. Classification problem of diabetic retinopathy stages is described.The architecture of deep neural networks based on VGG16 and VGG19 with the addition of custom layers is proposed. Recommendations for the selection of the size of the initial retinal images and the preprocessing stage (cropping) are given As a result of the performed experimental research. Analysis of the dataset was performed. Neural network models were trained and results were evaluated with class imbalance taken into account.

Spatio-temporal optical coherence tomography provides full thickness imaging of the chorioretinal complex.

Despite the rapid development of optical imaging methods, high-resolution invivo imaging with penetration into deeper tissue layers is still a major challenge. Optical coherence tomography (OCT) has been used successfully for non-invasive human retinal volumetric imaging invivo, advancing the detection, diagnosis, and monitoring of various retinal diseases. However, there are important limitations of volumetric OCT imaging, especially coherent noise and the limited axial range over which high resolution images can be acquired. The limited range prevents simultaneous measurement of the retina and choroid with adequate lateral resolution. In this article, we address these limitations with a technique that we term spatio-temporal optical coherence tomography (STOC-T), which uses light with controlled spatial and temporal coherence and advanced signal processing methods. STOC-T enabled the acquisition of high-contrast and high-resolution coronal projection images of the retina and choroid at arbitrary depths.

Preferential Loss of Contrast Decrement Responses in Human Glaucoma.

PurposeThe purpose of this study was to determine whether glaucoma in human patients produces preferential damage to OFF visual pathways, as suggested by animal experimental models, patient electroretinogram (ERG), and retinal imaging data.MethodsSteady-state visual evoked potentials (SSVEPs) were recorded monocularly from 50 patients with glaucoma and 28 age-similar controls in response to equal Weber contrast increments and decrements presented using 2.73 hertz (Hz) sawtooth temporal waveforms.ResultsThe eyes of patients with glaucoma were separated into mild (better than −6 decibel [dB] mean deviation; n = 28) and moderate to severe (worse than −6 dB mean deviation, n = 22) groups based on their Humphrey 24-2 visual field measurements. Response amplitudes and phases from the two glaucoma-severity groups were compared to controls at the group level. SSVEP amplitudes were depressed in both glaucoma groups, more so in the moderate to severe glaucoma group. The differences between controls and the moderate-severe glaucoma groups were more statistically reliable for decrements than for increments. Mean responses to decremental sawtooth stimuli were larger than those to increments in controls and in the mild glaucoma but not in the moderate to severe glaucoma group at the first harmonic. OFF/decrement responses at the first harmonic were faster in controls, but not in patients.ConclusionsThe observed pattern of preferential loss of decremental responses in human glaucoma is consistent with prior reports of selective damage to OFF retinal ganglion cells in murine models and in data from human ERG and retinal imaging. These data motivate pursuit of SSVEP as a biomarker for glaucoma progression.

Optical disk segmentation in human retina images with golden eagle optimizer

In ophthalmology, examination of the optic nerve and disc of the retina is a common way to diagnose a variety of eye diseases such as aging, diabetes, optic neuritis, glaucoma, papilledema, and ischemic optic neuropathy. This study presents the implementation of three combined methods based on the Golden Eagle algorithm, backup vector machine and geometrically active contour method and image segmentation. These methods have been tested using the well-known retina database benchmark, RIM-ONE. This study uses preprocessing based on the geometrically active Shannon contour guided by the Golden Eagle algorithm and post-processing based on regular spaced cross-section segmentation. The Golden Eagle method is more efficient than approaches such as the geometrically active contour and the SVM method. The proposed method, compared to other methods, shows higher mean values in terms of image similarity and statistical index. It is confirmed that the proposed Golden Eagle method could be used in the future to assess the clinical fit of retinal images.

Characterization and Analysis of Retinal Axial Motion at High Spatiotemporal Resolution and Its Implication for Real-Time Correction in Human Retinal Imaging.

High-resolution ophthalmic imaging devices including spectral-domain and full-field optical coherence tomography (SDOCT and FFOCT) are adversely affected by the presence of continuous involuntary retinal axial motion. Here, we thoroughly quantify and characterize retinal axial motion with both high temporal resolution (200,000 A-scans/s) and high axial resolution (4.5 μm), recorded over a typical data acquisition duration of 3 s with an SDOCT device over 14 subjects. We demonstrate that although breath-holding can help decrease large-and-slow drifts, it increases small-and-fast fluctuations, which is not ideal when motion compensation is desired. Finally, by simulating the action of an axial motion stabilization control loop, we show that a loop rate of 1.2 kHz is ideal to achieve 100% robust clinical in-vivo retinal imaging.

Non-interferometric volumetric imaging in living human retina by confocal oblique scanning laser ophthalmoscopy.

Three-dimensional (3D) imaging of the human retina is instrumental in vision science and ophthalmology. While interferometric retinal imaging is well established by optical coherence tomography (OCT), non-interferometric volumetric imaging in the human retina has been challenging up to date. Here, we report confocal oblique scanning laser ophthalmoscopy (CoSLO) to fill that void and harness non-interferometric optical contrast in 3D. CoSLO decouples the illumination and detection by utilizing oblique laser scanning and oblique imaging to achieve ∼4x better axial resolution than conventional SLO. By combining remote focusing, CoSLO permits the acquisition of depth signals in parallel and over a large field of view. Confocal gating is introduced by a linear sensor array to improve the contrast and resolution. For the first time, we reported non-interferometric 3D human retinal imaging with >20° viewing angle, and revealed detailed features in the inner, outer retina, and choroid. CoSLO shows potential to be another useful technique by offering 3D non-interferometric contrasts.