Translational Vision Science & Technology最新文献

Purpose: Diabetic retinopathy (DR) is a typical complication in patients with diabetes. This study aimed to compare retinal blood flow and vascular resistance between eyes with DR and healthy eyes using laser speckle flowgraphy (LSFG).

Methods: In total, 50 normal eyes and 87 DR eyes were examined at Kagawa University Hospital. LSFG was used to measure the mean blur rate (MBR) and total capillary resistance (TCR) of large vessels in the optic papilla. These values were compared across normal eyes and all eyes with DR, moderate nonproliferative diabetic retinopathy (NPDR), severe NPDR, and proliferative diabetic retinopathy (PDR). A TCR receiver operating characteristic (ROC) curve was plotted, and the diagnostic ability of the TCR for DR was determined using the area under the curve. The TCR cutoff value was determined using the Youden index.

Results: No significant difference in MRB was observed between normal eyes and the other groups. TCR was significantly higher in all groups except the PDR group, compared to normal eyes. The TCR area under the ROC curve was 0.751, indicating moderate diagnostic accuracy for DR. Using the Youden index, the TCR cutoff value was 0.79 (sensitivity, 0.740; specificity, 0.701).

Conclusions: Measuring TCR, in addition to MBR, as diagnostic markers provides more detailed pathological information regarding DR.

Translational relevance: Comparison of values between groups would be useful in predicting DR onset and stage progression.

Purpose: The purpose of this study was to introduce and evaluate the femtosecond laser enhanced refractive outcome (FLERO) prediction method, an intraocular lens (IOL) calculator that augments Barrett Universal II (BUII) by integrating novel anterior segment optical coherence tomography (OCT) biometric predictors obtained during femtosecond laser-assisted cataract surgery (FLACS).

Methods: Two thousand, three hundred sixty-three (2363) eyes of 1720 patients (mean age = 71.33 years, 60.26% women) undergoing FLACS were analyzed. FLERO was developed by selecting the most predictive subset of OCT-derived biometry features using a "genetic algorithm" and combining them with BUII predictions in a linear model. Internal validation was performed through cross-validation, and prediction errors (PEs) were compared with BUII and Kane errors.

Results: Compared to BUII, FLERO increased the proportion of eyes achieving postoperative refraction within ±0.25 diopter (D), ±0.50 D, and ±1.00 D of target from 0.470 to 0.507, 0.781 to 0.824, and 0.962 to 0.970, respectively. Mean absolute error decreased from 0.345 D for BUII and 0.338 D for Kane to 0.315 D for FLERO. FLERO outperformed BUII and Kane across (short, medium, and long) eyes, where proportions of eyes achieving refraction within ±0.50 D were 0.696, 0.831, and 0.782 for FLERO, 0.468, 0.796, and 0.718 for BUII, and 0.595, 0.798, and 0.718 for Kane. Wilcoxon Signed-Rank testing indicated significant reductions in absolute PEs for FLERO versus BUII and Kane (P < 0.0001). PE regression revealed FLERO made significantly smaller errors.

Conclusions: FLERO enhances BUII by incorporating novel OCT-derived FLACS biometric parameters across short, medium, and long eyes.

Translational relevance: FLERO combines advanced FLACS-derived intraoperative biometry with established IOL formulae to refine refractive outcome prediction.

Purpose: This study aims to predict postoperative toric implantable collamer lens (TICL) rotation and its effect on refraction and vision using artificial intelligence (AI) models.

Methods: Data from 642 eyes from 371 patients undergoing TICL surgery were included. Regression models predicted rotation degrees, and classification models predicted rotation-related residual astigmatism, visual acuity loss, or the need for realignment surgery. Regression tasks were evaluated using root mean square error (RMSE) and coefficient of determination (R²), while classification tasks were assessed by accuracy and mean area under the curve (AUC). Subgroup analyses were conducted for low, medium, and high astigmatism. The cutoff value for rotation affecting astigmatism or visual acuity was determined using receiver operating characteristic curves.

Results: Tabular prior-data fitted network (TabPFN) was the most accurate regression model for predicting postoperative rotation, achieving RMSE (10.672 ± 5.880) and mean absolute error (5.643 ± 2.328) compared to traditional models. For predicting rotation-related complications, TabPFN consistently achieved the highest accuracy across all secondary outcomes (0.906-0.981), particularly excelling in realignment surgery prediction with near-perfect precision (0.990 ± 0.006) and the highest AUC (0.900 ± 0.078). Cutoff values were 7.50° (AUC = 0.65), 4.50° (AUC = 0.68), and 2.50° (AUC = 0.73) for residual astigmatism and 9.50° (AUC = 0.72), 7.50° (AUC = 0.80), and 4.50° (AUC = 0.93) for visual acuity loss in low, medium, and high astigmatism groups, respectively.

Conclusions: AI models effectively predict postoperative rotation stability, providing valuable references for ophthalmologists.

Translational relevance: This study investigated the quantitative relationship between rotation, astigmatism, and vision, bridging the gap between artificial intelligence and optical theory.

Purpose: The lamina cribrosa (LC) is a complex network of collagenous beams that maintains mechanical homeostasis in response to fluctuations in intraocular pressure, thus protecting retinal ganglion cell axons exiting the eye. Understanding the structure and function of the LC may provide new insights into glaucomatous neurodegeneration. The purpose of this study is to utilize a two-photon fabrication technique to fabricate a model of the human LC (mLC) to scale.

Methods: Segmented multiphoton microscopy images of human LC tissues from a prior study were used to two-photon polymerize three mLCs. The mLCs were subsequently imaged using micro-computed tomography. Regional and full-field structural anisotropies and global microstructure were compared between the input and micro-computed tomography mLC images.

Results: Structural analysis of the LC tissues and mLCs demonstrates that various characteristics were closely maintained after fabrication. There was variation in the parameters across samples. Pore eccentricity, structural anisotropy, and pore convexity were all closely recapitulated with an error of less than approximately 15%.

Conclusions: Generating a model of the human LC from segmented images is the first step toward a biomimetic approach to patient-specific modeling of the LC. Future work to improve the resolution and match the material properties of LC native tissues will generate a powerful model for mechanobiological studies. Mechanobiological experiments may be useful to understand the underlying mechanisms that drive glaucoma disease initiation and progression.

Translational relevance: This study introduces a novel method to fabricate the human LC, which can allow for patient-specific mechanobiological models of the LC in glaucoma.

Purpose: To assess the influence of varying concentrations of benzalkonium chloride (BAK), the predominant preservative utilized in ophthalmic formulations, on the barrier integrity and mitochondrial function of primary cultured human corneal epithelial cells (HCECs).

Methods: Primary HCEC monolayers were exposed to BAK at concentrations ranging from 0.02% to 0.00002%. The barrier function was monitored using electric cell-substrate impedance sensing (ECIS), where a decrease in electrical resistance signified a loss of barrier function. Mitochondrial function was evaluated after 24 hours of BAK exposure with the Seahorse XFe96 Flux Analyzer, which measured basal respiration, adenosine triphosphate (ATP) production, and maximal respiration.

Results: High BAK concentrations (≥0.02%) caused a rapid, dose-dependent decrease in resistance, exceeding 40% within 1 hour. In contrast, lower concentrations (0.00025%-0.002%) led to a delayed, gradual reduction. Specifically, 0.00025% BAK resulted in a 37% decrease in resistance by 72 hours, whereas 0.0001% caused a 26% reduction; concentrations ≤ 0.00005% had no significant effect. Increased capacitance accompanied the resistance loss, indicating membrane disturbance. Seahorse analysis revealed that BAK concentrations ≥ 0.00005% significantly reduced basal respiration and ATP production. Maximal respiration decreased at higher doses (≥0.0001%).

Conclusions: BAK induces concentration-dependent, cumulative toxicity in HCECs, causing rapid membrane disruption and irreversible barrier failure at or above its critical micelle concentration (CMC), along with ongoing sub-CMC toxicity through mitochondrial suppression at lower doses. These findings highlight the need for preservative strategies that reduce both acute and chronic epithelial damage in ophthalmic applications.

Translational relevance: Real-time impedance and mitochondrial assessments determine thresholds for BAK toxicity, guiding the development of safer ophthalmic formulations to protect the ocular surface.

Purpose: To compare the ability of conventional luminance-modulating perimetric stimuli and an area-modulation stimulus (AMS) designed to measure changes in complete spatial summation to identify physiological retinal ganglion cell density (RGCD) gradients in healthy observers.

Methods: Contrast thresholds were measured for Goldmann III (GIII; 0.43°, 200 ms) and V (GV; 1.72°, 200 ms) stimuli at 3° and 10° eccentricity in 100 healthy observers (median age, 43 years, range, 18-85 years), with mean spherical equivalent refractive errors ranging from -10.38 to +4.63 DS. Area thresholds were measured at the same locations using a fixed luminance stimulus (ΔL: 4.4 cd/m2, 200 ms). Colocalized RGCD estimates were determined using (i) optical coherence tomography (OCT) RGC layer thickness measures, and (ii) achromatic peripheral grating resolution acuity (PGRA) thresholds. Ratios of the difference in log energy threshold (ΔE) and log RGCD (ΔRGCD) between eccentricities were calculated (ΔE/|ΔRGCD|), with a value of 1 assumed to be the optimal relationship between functional thresholds and RGCD.

Results: ΔE/|ΔRGCD|) values (median and interquartile range [IQR]) were largest for AMS (OCT, 0.54 [IQR, 0.37-0.78]; PGRA, 0.71 [IQR, 0.46-1.19]), followed by GIII (OCT, 0.29 [IQR, 0.08-0.44]; PGRA, 0.33 [IQR, 0.07-0.54]; and GV (OCT, 0.16 [IQR, 0.02-0.29]; and PGRA, 0.19 [IQR, 0.02-0.44]). Interstimulus differences between all stimulus pairs were statistically significant (AMS vs GIII, both P < 0.001; AMS vs GV, both P < 0.001; GIII vs GV, both P < 0.05).

Conclusions: ΔE/|ΔRGCD| values were closest to 1 for AMS, suggesting this stimulus relates best to underlying physiological variations in RGCD.

Translational relevance: Thresholds measured with area modulation stimuli vary more proportionally with physiological changes in retinal ganglion cell density relative to conventional perimetric stimuli.

Purpose: To develop and validate a novel method for quantifying blood flow velocity in ocular vessels using speckle pattern analysis in optical coherence tomography (OCT) images.

Methods: We developed an OCT-based methodology that quantifies flow velocity by analyzing speckle structures within OCT images to derive a relative flow value (RFV) based on the horizontal derivative of speckle patterns. A model eye with 150-µm capillary tubing was used to evaluate speckle pattern analysis at controlled flow rates of human and porcine blood. We assessed the correlation between RFV and actual flow velocities and examined measurement consistency across vessel sectioning angles varying ±20° from perpendicular.

Results: RFV exhibited a power-law relationship with actual flow velocities, RFV = 169.6 × (Actual Flow)0.264 (r2 = 0.983), and maintained accuracy without saturation up to 70 mm/s. Measurements remained consistent across vessel sectioning angles. Intraluminal speckle structures changed dynamically with flow velocity, and averaged images revealed characteristic hourglass profiles at higher velocities. In a patient with central retinal vein occlusion, RFV detected altered pulsatility and reduced venous flow, aligning with clinical expectations.

Conclusions: This speckle-based OCT method provides a non-invasive, depth-resolved approach for quantifying ocular blood velocity with minimal computational requirements. It holds potential for diagnosing and monitoring vascular abnormalities in ocular and systemic diseases. Further validation is necessary across diverse patient populations.

Translational relevance: This method simplifies blood flow quantification, enhancing the role of OCT in hemodynamic assessment both ocularly and systemically.

Purpose: Ultra-widefield retinal imaging is critical for diagnosing peripheral retinal diseases, but discrepancies in field of view (FOV) definitions hinder system comparisons. This study aims to evaluate the FOV performance of six widefield or ultra-widefield retinal imaging systems using a simplified eye phantom model and in vivo demonstration in a healthy subject.

Methods: A custom eye phantom model with concentric ring patterns and geometric markers was used to assess FOV and distortion across six imaging systems, including Nikon Optos, Natus RetCam, Intalight Dream OCT, Siloam Vision's contact and non-contact iCam-OCT, and OHSU table-top UWF OCT. The phantom model eliminated anatomic variability, enabling controlled measurements. Imaging was also performed on a healthy emmetropic adult under identical conditions for qualitative validation.

Results: The evaluation revealed significant variations in peripheral coverage and distortion among systems. Nikon Optos and Siloam Vision contact iCam-OCT achieved the widest FOVs, whereas RetCam and Intalight Dream OCT had more limited coverage. Radial and tangential distortion increased with eccentricity, with Nikon Optos exhibiting the highest distortion. In vivo imaging confirmed phantom-based findings regarding FOV coverage and distortion across these systems.

Conclusions: FOV performance varies across imaging systems, and the measured FOVs were generally consistent with the manufacturers stated specifications. Distortion increased with eccentricity and differed among devices. The phantom model provided an objective method for quantifying FOV and distortion, with findings supported by in vivo imaging.

Translational relevance: This work offers a practical method for evaluating widefield imaging systems, supporting informed system selection, and emphasizing consistent evaluation criteria in clinical practice.

Purpose: To evaluate the performance of vision-language models (VLMs), in glaucoma detection and visual field (VF) mean deviation (MD) prediction tasks using optical coherence tomography (OCT) images.

Methods: A total of 27,610 SPECTRALIS OCT images from 1025 participants (1690 eyes), collected between 2008 and 2021 as part of the Diagnostic Innovations in Glaucoma Study (DIGS) and the African Descent and Glaucoma Evaluation Study (ADAGES), were included. Vision components of LLaVA and PaliGemma, as well as RETFound and ResNet-50 models, were fine-tuned for glaucoma classification and VF MD prediction. Models were trained using OCT circle scans centered on the optic nerve head. Three training configurations were compared. Performance was evaluated using area under the receiver operating characteristic curve (AUC), mean absolute error (MAE), and related metrics.

Results: The LLaVA model, when both vision encoder and multi-layer projector were fine-tuned, achieved the best performance with an AUC of 0.92 (95% confidence interval [CI], 0.86-0.95) for glaucoma classification and an MAE of 1.79 dB (95% CI, 1.55-2.00) for VF MD prediction. RETFound and PaliGemma also performed well, with AUCs of 0.91 and 0.90 and MAEs of 1.87 dB and 1.84 dB, respectively. Models with frozen vision encoders showed reduced accuracy. Stratified analysis showed better glaucoma classification in older individuals and moderate-to-advanced cases. VF MD prediction was more accurate in younger individuals, with higher errors in advanced glaucoma.

Conclusions: Fine-tuned VLMs demonstrated high performance in glaucoma detection and VF MD prediction, matching or exceeding specialized foundation models and traditional convolutional neural network (CNN)-based methods.

Translational relevance: This study highlights the potential of general-purpose AI models to be adapted for glaucoma care, enabling scalable decision support from OCT imaging.

Purpose: To propose a novel deep learning-based methodology for drusen detection and quantification in early age-related macular degeneration (AMD) using retinal multispectral images. The retinal multispectral images highlight features in several nonoverlapping spectral bands that the deep learning models leverage for automatic drusen detection and quantification in dry AMD.

Methods: The proposed novel methodology comprises quality assessment of retinal images, region of interest extraction, drusen segmentation, and drusen quantification stages. Different deep learning models (such as UNet++ convolutional neural network with EfficientNetV2 encoder) have been implemented for these stages. A total of 170 drusen and 150 nondrusen retinal images (single eye) were split into four training and validation data sets to analyze the performance of a deep learning model for drusen segmentation.

Results: The proposed methodology achieved an average score, recall, and precision of 0.691, 0.668, and 0.776, respectively, across all four validation sets. This work also analyzed the performance of the proposed deep learning model for discriminating drusen and drusen-like lesions, achieving a pixel-wise segmentation accuracy of 99.998%. The number and the diameter of the detected drusen were also computed. A Dice score distribution for drusen segmentation with different numbers and sizes of drusen per eye is also shown.

Conclusions: This work demonstrates that deep learning models applied to retinal multispectral images can provide accurate and clinically significant drusen segmentation and quantification, thereby facilitating early detection, longitudinal monitoring, and reduction of the risk of vision loss from AMD.

Translational relevance: Deep learning-assisted detection of drusen from multispectral retinal images will refine and improve clinical diagnosis of early nonexudative age-related macular degeneration.