Deep Learning Segmentation, Visualization, and Automated 3D Assessment of Ciliary Body in 3D Ultrasound Biomicroscopy Images.

Purpose: This study aimed to develop a fully automated deep learning ciliary body segmentation and assessment approach in three-dimensional ultrasound biomicroscopy (3D-UBM) images. Methods: Each 3D-UBM eye volume was aligned to the optic axis via multiplanar reformatting. Ciliary muscle and processes were manually annotated, and Deeplab-v3+ models with different loss functions were trained to segment the ciliary body (ciliary muscle and processes) in both en face and radial images. Results: We trained and tested the models on 4320 radial and 3864 en face images from 12 cadaver eye volumes. Deep learning models trained on radial images with Dice loss achieved the highest mean F1-score (0.89) for ciliary body segmentation. For three-class segmentation (ciliary muscle, processes, and background), radial images with Dice loss achieved the highest mean F1-score (0.75 for the ciliary process and 0.82 for the ciliary muscle). Part of the ciliary muscle (10.9%) was misclassified as the ciliary process and vice versa, which occurred owing to the difficulty in differentiating the ciliary muscle-processes border, even by experts. Deep learning segmentation made further editing by experts at least seven times faster than a fully manual approach. In eight cadaver eyes, the average ciliary muscle, process, and body volumes were 56 ± 9, 43 ± 13, and 99 ± 18 mm3, respectively. The average surface area of the ciliary muscle, process, and body were 346 ± 45, 363 ± 83, and 709 ± 80 mm2, respectively. We performed transscleral cyclophotocoagulation in cadaver eyes to shrink the ciliary processes. Both manual and automated measurements from deep learning segmentation show a decrease in volume, surface area, and 360° cross-sectional area measurements. Conclusions: The proposed deep learning segmentation of the ciliary body and 3D measurements showed transscleral cyclophotocoagulation-related changes in the ciliary body. Translational Relevance: Automated ciliary body assessment using 3D-UBM has the translational potential for ophthalmic treatment planning and monitoring.

Predictive Assessment of Quantitative Ultra-Widefield Angiographic Features for Future Need for Anti-VEGF Therapy in Diabetic Eye Disease.

The objective of this study was to identify biomarkers that predict a future need for anti-VEGF therapy in diabetic retinopathy (DR). Eyes with DR that underwent ultra-widefield angiography (UWFA) and had at least a 1 year follow-up were grouped based on future anti-VEGF treatment requirements: (1) not requiring treatment, (2) immediate treatment (within 3 months of UWFA), and (3) delayed treatment (after 3 months of UWFA). Quantitative UWFA features and clinical factors were evaluated. Random forest models were built to differentiate eyes requiring immediate and delayed treatment from the eyes not requiring treatment. A total of 173 eyes were included. The mean follow-up was 22 (range: 11-43) months. The macular leakage index, panretinal leakage index, presence of DME, and visual acuity were significantly different in eyes requiring immediate (n = 38) and delayed (n = 34) treatment compared to eyes not requiring treatment (n = 101). Random forest model differentiating eyes requiring immediate treatment from eyes not requiring treatment demonstrated an AUC of 0.91 ± 0.07. Quantitative angiographic features have potential as important predictive biomarkers of a future need for anti-VEGF therapy in DR and may serve to guide the frequency of a follow-up.

Longitudinal assessment of quantitative ultra-widefield ischaemic and vascular parameters in sickle cell retinopathy.

PURPOSE: To evaluate longitudinal quantitative ischaemic and vasculature parameters, including ischaemic index, vessel area, length and geodesic distance in sickle cell retinopathy (SCR) on ultra-widefield fluorescein angiography (UWFA). METHODS: Optimal UWFA images from two longitudinal timepoints of 74 eyes from 45 patients with SCR were aligned and a common region of interest was determined. A deep-learning augmented ischaemia and vascular segmentation platform was used for feature extraction. Geodesic distance maps demonstrating the shortest distance within the vascular masks from the centre of the optic disc were created. Ischaemic index, vessel area, vessel length and geodesic distance were measured. Paired t-test and linear mixed effect model analysis were performed. RESULTS: Overall, 25 (44 eyes) patients with HbSS, 14 (19 eyes) with HbSC, 6 (11 eyes) with HbSthal and other genotypes were included. Mean age was 40.1±11.0 years. Mean time interval between two UWFA studies was 23.0±15.1 months (range: 3-71.3). Mean panretinal ischaemic index increased from 10.0±7.2% to 10.9±7.3% (p<0.005). Mean rate of change in ischaemic index was 0.5±0.7% per year. Mean vessel area (p=0.020) and geodesic distance (p=0.048) decreased significantly. Multivariate analysis demonstrated baseline ischaemic index and Goldberg stage are correlated with progression. CONCLUSION: Longitudinal ischaemic index and retinal vascular parameter measurements demonstrate statistically significant progression in SCR. The clinical significance of these relatively small magnitude changes remains unclear but may provide insights into the progression of retinal ischaemia in SCR.

LONGITUDINAL ELLIPSOID ZONE DYNAMICS AFTER MACULAR HOLE REPAIR IN THE DISCOVER STUDY: Structure-Function Assessment.

PURPOSE: To evaluate association of the baseline macular hole (MH) geometric features and longitudinal ellipsoid zone integrity with the visual acuity outcome after surgical repair. METHODS: This was a post-hoc analysis of eyes in the DISCOVER study undergoing vitrectomy repair for MH. Anatomical and functional data were collected through one year postoperatively. An automated retinal layer segmentation platform was used for the assessment of outer retinal metrics and volumetric reconstruction of MH. Association of longitudinal ellipsoid zone features and baseline MH height, width, and volume with VA outcomes were investigated. RESULTS: Eighty-four eyes with MH were included. The mean baseline VA was 20 of 114 and increased to 20 of 45 (P < 0.001) at postoperative Month 12 (N = 45). Successful MH closure was achieved in 98.8% of cases. Ellipsoid zone integrity metrics significantly improved from baseline (P = 0.002) and postoperative Month 1 (P < 0.001) to post-operative Month 12. Ellipsoid zone metrics independently correlated with VA at all follow-up visits (P < 0.05). Increased baseline MH width and volume negatively correlated with the VA at postoperative Month 12 (P < 0.001). Preoperative VA and EZ integrity on optical coherence tomography were predictors for postoperative VA. CONCLUSION: Baseline MH volumetric parameters and EZ parameters were associated with VA outcomes after repair.

iOCT-assisted macular hole surgery: outcomes and utility from the DISCOVER study.

BACKGROUND/AIMS: This study aimed to characterise the clinical outcomes and utility of intraoperative optical coherence tomography (iOCT)-assisted macular hole (MH) repair. METHODS: This was a post hoc analysis of eyes in the D etermination of feasibility of I ntraoperative S pectral domain microscope C ombined/integrated O CT V isualization during E n face R etinal and ophthalmic surgery (DISCOVER) study undergoing surgical MH repair with use of iOCT. Functional and surgical outcome data were collected through 12 months postoperatively. MH closure rate, postoperative visual acuity (VA), percentage of cases in which iOCT provided valuable feedback and altered surgical decision making were measured. RESULTS: Eighty-four eyes were included in this study. The mean preoperative VA measured 20/114. The mean postoperative VA improved to 20/68 (p<0.001) at month 1, 20/48 (p<0.001) at month 3 and 20/45 (p<0.001) at month 12 or later. In 43 cases (51%), surgeons reported that iOCT provided valuable information (eg, confirming release of vitreomacular traction and identification of occult residual membranes). In 10 cases (12%), iOCT data specifically altered surgical decision making. Postoperative day 1 transtamponade OCT confirmed tissue apposition and apparent hole closure in 74% of eyes (21/26). All five open holes on postoperative day 1 closed following positioning. Single-surgery MH closure was achieved in 97.6% of cases. One persistent MH was successfully closed with a subsequent surgical repair for a final overall closure rate of 98.8%. Due to chronicity and MH size, additional repair was not recommended for the single remaining persistent MH. CONCLUSION: This study suggests that iOCT may have important utility in MH surgery, including impacting surgical decision making. iOCT-assisted MH surgery resulted in significant improvement in VA and high single-surgery success rate.

Radiomics-based assessment of ultra-widefield leakage patterns and vessel network architecture in the PERMEATE study: insights into treatment durability.

AIM: To evaluate the potential of radiomics-based ultra-widefield fluorescein angiography (UWFA)-derived imaging biomarkers in retinal vascular disease for predicting therapeutic durability of intravitreal aflibercept injection (IAI). METHODS: The Peripheral and Macular Retinal Vascular Perfusion and Leakage Dynamics in Diabetic Macular Edema and Retinal Venous Occlusions During Intravitreal Aflibercept Injection (IAI) Treatment for Retinal Edema (PERMEATE) study prospectively evaluated quantitative UWFA dynamics in diabetic macular oedema or macular oedema secondary to retinal vascular occlusion. 27 treatment-naïve eyes were treated with 2 mg IAI q4 weeks for the first 6 months, and then administered q8 weeks. Morphological and graph-based attributes were used to model the spatial distribution of leakage areas, while tortuosity measures were used to model the vessel network disorder. Eyes were grouped based on functional tolerance of the first 8-week treatment interval challenge. 'Non-rebounders' (N=15) maintained/improved best-corrected visual acuity (BCVA) following the 8-week challenge. 'Rebounders' (N=12) exhibited worsened BVCA. The image biomarkers were used with a machine learning classifier to preliminarily evaluate their ability to predict BCVA stability. RESULTS: Two new UWFA image-derived biomarkers were identified and extracted. The cross-validated area under the receiver operating characteristic curve (AUC) was 0.77±0.14 using baseline leakage distribution features and 0.73±0.10 for the UWFA baseline tortuosity measures. Additionally, the change in vascular tortuosity between month 4 and baseline yielded an AUC of 0.73±0.08. Three baseline clinical features of letter score, macular volume and central subfield thickness yielded a corresponding AUC of 0.42±0.09. CONCLUSIONS: Two computer-extracted UWFA radiomics-based descriptors were identified as potential biomarkers for predicting treatment durability and tolerance of longer treatment intervals. Conventional treatment parameters were not significantly different between these same groups.

Automated Quality Assessment and Image Selection of Ultra-Widefield Fluorescein Angiography Images through Deep Learning.

Purpose: Numerous angiographic images with high variability in quality are obtained during each ultra-widefield fluorescein angiography (UWFA) acquisition session. This study evaluated the feasibility of an automated system for image quality classification and selection using deep learning. Methods: The training set was comprised of 3543 UWFA images. Ground-truth image quality was assessed by expert image review and classified into one of four categories (ungradable, poor, good, or best) based on contrast, field of view, media opacity, and obscuration from external features. Two test sets, including randomly selected 392 images separated from the training set and an independent balanced image set composed of 50 ungradable/poor and 50 good/best images, assessed the model performance and bias. Results: In the randomly selected and balanced test sets, the automated quality assessment system showed overall accuracy of 89.0% and 94.0% for distinguishing between gradable and ungradable images, with sensitivity of 90.5% and 98.6% and specificity of 87.0% and 81.5%, respectively. The receiver operating characteristic curve measuring performance of two-class classification (ungradable and gradable) had an area under the curve of 0.920 in the randomly selected set and 0.980 in the balanced set. Conclusions: A deep learning classification model demonstrates the feasibility of automatic classification of UWFA image quality. Clinical application of this system might greatly reduce manual image grading workload, allow quality-based image presentation to clinicians, and provide near-instantaneous feedback on image quality during image acquisition for photographers. Translational Relevance: The UWFA image quality classification tool may significantly reduce manual grading for clinical- and research-related work, providing instantaneous and reliable feedback on image quality.