Classification of Vogt-Koyanagi-Harada disease using feature selection and classification based on wide-field swept-source optical coherence tomography angiography.

Background: Vogt-Koyanagi-Harada (VKH) disease is a common and easily blinded uveitis entity, with choroid being the main involved site. Classification of VKH disease and its different stages is crucial because they differ in clinical manifestations and therapeutic interventions. Wide-field swept-source optical coherence tomography angiography (WSS-OCTA) provides the advantages of non-invasiveness, large-field-of-view, high resolution, and ease of measuring and calculating choroid, offering the potential feasibility of simplified VKH classification assessment based on WSS-OCTA. Methods: 15 healthy controls (HC), 13 acute-phase and 17 convalescent-phase VKH patients were included, undertaken WSS-OCTA examination with a scanning field of 15 × 9 mm2. 20 WSS-OCTA parameters were then extracted from WSS-OCTA images. To classify HC and VKH patients in acute and convalescent phases, two 2-class VKH datasets (HC and VKH) and two 3-class VKH datasets (HC, acute-phase VKH, and convalescent-phase VKH) were established by the WSS-OCTA parameters alone or in combination with best-corrected visual acuity (logMAR BCVA) and intraocular pressure (IOP), respectively. A new feature selection and classification method that combines an equilibrium optimizer and a support vector machine (called SVM-EO) was adopted to select classification-sensitive parameters among the massive datasets and to achieve outstanding classification performance. The interpretability of the VKH classification models was demonstrated based on SHapley Additive exPlanations (SHAP). Results: Based on pure WSS-OCTA parameters, we achieved classification accuracies of 91.61% ± 12.17% and 86.69% ± 8.30% for 2- and 3-class VKH classification tasks. By combining the WSS-OCTA parameters and logMAR BCVA, we achieved better classification performance of 98.82% ± 2.63% and 96.16% ± 5.88%, respectively. Through SHAP analysis, we found that logMAR BCVA and vascular perfusion density (VPD) calculated from the whole field of view region in the choriocapillaris (whole FOV CC-VPD) were the most important features for VKH classification in our models. Conclusion: We achieved excellent VKH classification performance based on a non-invasive WSS-OCTA examination, which provides the possibility for future clinical VKH classification with high sensitivity and specificity.

Oxygen-saturation-related functional parameter as a biomarker for diabetes mellitus-extraction method and clinical validation.

Purpose: To develop a computational method for oxygen-saturation-related functional parameter analysis of retinal vessels based on traditional color fundus photography, and to explore their characteristic alterations in type 2 diabetes mellitus (DM). Methods: 50 type 2 DM patients with no-clinically detectable retinopathy (NDR) and 50 healthy subjects were enrolled in the study. An optical density ratio (ODR) extraction algorithm based on the separation of oxygen-sensitive and oxygen-insensitive channels in color fundus photography was proposed. With precise vascular network segmentation and arteriovenous labeling, ODRs were acquired from different vascular subgroups, and the global ODR variability (ODRv) was calculated. Student's t-test was used to analyze the differences of the functional parameters between groups, and regression analysis and receiver operating characteristic (ROC) curves were used to explore the discrimination efficiency of DM patients from healthy subjects based on these functional parameters. Results: There was no significant difference in the baseline characteristics between the NDR and healthy normal groups. The ODRs of all vascular subgroups except the micro venule were significantly higher (p<0.05, respectively) while ODRv was significantly lower (p<0.001) in NDR group than that in healthy normal group. In the regression analysis, the increased ODRs except micro venule and decreased ODRv were significantly correlated with the incidence of DM, and the C-statistic for discrimination DM with all ODR is 0.777 (95% CI 0.687-0.867, p<0.001). Conclusion: A computational method to extract the retinal vascular oxygen-saturation-related optical density ratios (ODRs) with single color fundus photography was developed, and increased ODRs and decreased ODRv of retinal vessels could be new potential image biomarkers of DM.

Automatic vessel crossing and bifurcation detection based on multi-attention network vessel segmentation and directed graph search.

Analysis of the vascular tree is the basic premise to automatically diagnose retinal biomarkers associated with ophthalmic and systemic diseases, among which accurate identification of intersection and bifurcation points is quite challenging but important for disentangling complex vascular network and tracking vessel morphology. In this paper, we present a novel directed graph search-based multi-attentive neural network approach to automatically segment the vascular network and separate intersections and bifurcations from color fundus images. Our approach uses multi-dimensional attention to adaptively integrate local features and their global dependencies while learning to focus on target structures at different scales to generate binary vascular maps. A directed graphical representation of the vascular network is constructed to represent the topology and spatial connectivity of the vascular structures. Using local geometric information including color difference, diameter, and angle, the complex vascular tree is decomposed into multiple sub-trees to finally classify and label vascular feature points. The proposed method has been tested on the DRIVE dataset and the IOSTAR dataset containing 40 images and 30 images, respectively, with 0.863 and 0.764 F1-score of detection points and average accuracy of 0.914 and 0.854 for classification points. These results demonstrate the superiority of our proposed method outperforming state-of-the-art methods in feature point detection and classification.

Portable boom-type ultrahigh-resolution OCT with an integrated imaging probe for supine position retinal imaging.

To expand the clinical applications and improve the ease of use of ultrahigh-resolution optical coherence tomography (UHR-OCT), we developed a portable boom-type ophthalmic UHR-OCT operating in supine position that can be used for pediatric subjects, bedridden patients and perioperative conditions. By integrating the OCT sample arm probe with real-time iris display and automatic focusing electric lens for easy alignment, coupling the probe on a self-locking multi-directional manipulator to reduce motion artifacts and operator fatigue, and installing the OCT module on a moveable cart for system mobility, our customized portable boom-type UHR-OCT enables non-contact, high-resolution and high-stability retinal examinations to be performed on subjects in supine position. The spectral-domain UHR-OCT operates at a wavelength of 845 nm with 130 nm FWHM (full width at half maximum) bandwidth, achieving an axial resolution of ≈2.3µm in tissue with an A-line acquisition rate up to 128 kHz. A high-definition two-dimensional (2D) raster protocol was used for high-quality cross-sectional imaging while a cube volume three-dimensional (3D) scan was used for three-dimensional imaging and en-face reconstruction, resolving major layer structures of the retina. The feasibility of the system was demonstrated by performing supine position 2D/3D retinal imaging on healthy human subjects, sedated infants, and non-sedated awake neonates.

Quantitative analysis of morphological and functional features in Meibography for Meibomian Gland Dysfunction: Diagnosis and Grading.

BACKGROUND: To explore the performance of quantitative morphological and functional analysis in meibography images by an automatic meibomian glands (MGs) analyser in diagnosis and grading Meibomian Gland Dysfunction (MGD). METHODS: A cross-sectional study collected 256 subjects with symptoms related to dry eye and 56 healthy volunteers who underwent complete ocular surface examination was conducted between January 1, 2019, and December 31, 2020. The 256 symptomatic subjects were classified into MGD group (n = 195) and symptomatic non-MGD group (n = 61). An automatic MGs analyser was used to obtained multi-parametric measurements in meibography images including the MGs area ratio (GA), MGs diameter deformation index (DI), MGs tortuosity index (TI), and MGs signal index (SI). Adjusted odds ratios (ORs) of the multi-parametric measurements of MGs for MGD, and the area under the receiver operating characteristic (AUC-ROC) curves of multi-parametric measurements for MGD diagnosing and grading were conducted. FINDINGS: When consider age, sex, ocular surface condition together, the estimated ORs for DI was 1.62 (95% CI, 1.29-2.56), low-level SI was 24.34 (95% CI, 2.73-217.3), TI was 0.76(95% CI, 0.54-0.90), and GA was 0.86 (95% CI, 0.74-0.92) for MGD. The combination of DI-TI-GA-SI showed an AUC = 0.82 (P < 0.001) for discriminating MGD from symptomatic subjects. The DI had a higher AUC in identifying early-stage MGD (grade 1-2), while TI and GA had higher AUCs in moderate and advanced stages (grade 3-5). Merging DI-TI-GA showed the highest AUCs in distinguish MGD severities. INTERPRETATION: The MGs area ratio, diameter deformation, tortuosity and signal intensity could be considered promising biomarkers for MGD diagnosis and objective grading. FUNDING: This work was supported by the Key-Area Research and Development Program of Guangdong Province (No. 2019B010152001), the National Natural Science Foundation of China under Grant (81901788) and Guangzhou Science and Technology Program (202002030412).

A novel multiple subdivision-based algorithm for quantitative assessment of retinal vascular tortuosity.

Vascular tortuosity as an indicator of retinal vascular morphological changes can be quantitatively analyzed and used as a biomarker for the early diagnosis of relevant disease such as diabetes. While various methods have been proposed to evaluate retinal vascular tortuosity, the main obstacle limiting their clinical application is the poor consistency compared with the experts' evaluation. In this research, we proposed to apply a multiple subdivision-based algorithm for the vessel segment vascular tortuosity analysis combining with a learning curve function of vessel curvature inflection point number, emphasizing the human assessment nature focusing not only global but also on local vascular features. Our algorithm achieved high correlation coefficients of 0.931 for arteries and 0.925 for veins compared with clinical grading of extracted retinal vessels. For the prognostic performance against experts' prediction in retinal fundus images from diabetic patients, the area under the receiver operating characteristic curve reached 0.968, indicating a good consistency with experts' predication in full retinal vascular network evaluation.

An automated and multiparametric algorithm for objective analysis of meibography images.

BACKGROUND: Meibography is a non-contact imaging technique used by ophthalmologists and eye care practitioners to acquire information on the characteristics of meibomian glands. One of its most important applications is to assist in the evaluation and diagnosis of meibomian gland dysfunction (MGD). As the artificial qualitative analysis of meibography images can lead to low repeatability and efficiency, automated and quantitative evaluation would greatly benefit the image analysis process. Moreover, since the morphology and function of meibomian glands varies at different stages of MGD, multiparametric analysis offering more comprehensive information could help in discovering subtle changes to glands during MGD progression. Therefore, an automated and multiparametric objective analysis of meibography images is urgently needed. METHODS: An algorithm was developed to perform multiparametric analysis of meibography images with fully automatic and repeatable segmentation based on image contrast enhancement and noise reduction. The full architecture can be divided into three steps: (I) segmentation of the tarsal conjunctiva area as the region of interest (ROI); (II) segmentation and identification of glands within the ROI; and (III) quantitative multiparametric analysis including a newly defined gland diameter deformation index (DI), gland tortuosity index (TI), and gland signal index (SI). To evaluate the performance of this automated algorithm, the similarity index (k) and the segmentation error including the false-positive rate (rP ) and the false-negative rate (rN ) were calculated between the manually defined ground truth and the automatic segmentations of both the ROI and meibomian glands of 15 typical meibography images. RESULTS: The results of the performance evaluation between the manually defined ground truth and automatic segmentations were as follows: for ROI segmentation, the similarity index (k)=0.94±0.02, the false-positive rate (rP )=6.02%±2.41%, and the false-negative rate (rN )=6.43%±1.98%; for meibomian gland segmentation, the similarity index (k)=0.87±0.01, the false-positive rate (rP )=4.35%±1.50%, and the-false negative rate (rN )=18.61%±1.54%. The algorithm was successfully applied to process typical meibography images acquired from subjects of different meibomian gland health statuses, by providing the gland area ratio (GA), the gland length (L), gland width (D), gland diameter deformation index (DI), gland tortuosity index (TI), and gland signal index (SI). CONCLUSIONS: A fully automated algorithm was developed which demonstrated high similarity with moderate segmentation errors for meibography image segmentation compared with the manual approach, offering multiple parameters to quantify the morphology and function of meibomian glands for the objective evaluation of meibography images.