Comparison of AI‐based retinal fluid monitoring in neovascular age‐related macular degeneration with manual assessment by different eye care professionals under optimized conditions

To assess the interobserver variability (IOV) in indicating retreatment for neovascular Age-related macular degeneration 4 weeks after three Ranibizumab loading doses using spectral domain OCT (SD-OCT) as the primary objective diagnostic tool. Four observers decided for or against 4th Ranibizumab injection in 108 patients by six different rating rounds (RR) based on the SD-OCT findings after the loading doses. Postoperative OCT images were supplemented consecutively with information from a chart review as the 'patients subjective estimation of vision (SE)', the course of best-corrected visual acuity (BCVA) and the preoperative OCT as well as all information collectively. Agreement rates (AR) and Kappa statistics were calculated. Based on post-treatment OCT findings only (RR1), mean reinjection rate of all observers was 37.5%. Adding supplementary information, mean reinjection rate decreased to 20% when all information was available reflecting the 'real' situation (RR 6). Interobserver agreement rates varied from 66.7% to 90.7% depending on rating rounds and interobserver pairs. Mean AR and Kappa values (KV) were as following: AR 81.6%, KV 0.61 (RR1: 'only post-OP OCT'); AR 76.7%, KV 0.33 (RR2: post-OP OCT + SE); AR 80.3%, KV 0.45 (RR3: post-OP OCT + BCVA); AR 80.7%, KV 0.46 (RR4: pre- and post-OP OCT); AR 82.2%, KV 0.49 (RR5: post-OP OCT + SE + BCVA); and finally AR 83.6%, KV 0.47 (RR6: pre- and post-OP OCT + SE + BCVA). The overall mean agreement rate was 80.9% with a Kappa of 0.47. IOV for indicating retreatment after three Ranibizumab loading doses reveals only moderate agreement in Kappa statistics, which seems to be too low considering the high costs for retreatments. More concise guidelines based on the post-treatment OCT scans as the presumably most sensitive and noninvasive objective tool to follow choroidal neovascularization activity by judging the course of sub- and intraretinal fluid are necessary.

Evaluation of contrast-modified optical coherence tomography (OCT) images as a tool for better macular neovascularization (MNV) classification in age-related macular degeneration (AMD) cases. Twenty-five OCT images obtained with SPECTRALIS (Heidelberg Engineering GmbH) from patients showing MNV (10 type 1, 10 type 2, and 5 type 3) were selected. Two retina specialists (RDM and RGP) classified the MNV lesions and then the same cases were classified by 37 ophthalmologists with different degree of training. A grading tool was designed to classify these cases using standard OCT images (contrast value: 12) followed by reassessment with contrast-modified OCT images (contrast value: 1). Thirty-seven ophthalmologists were involved: four 1st-year trainees, four 2nd-year trainees, four 3rd-year trainees, 12 4th-year trainees, and 13 consultants from different subspecialties. Average result for correct classification was 13.16 using standard images alone and 14.22 using contrast-modified images showing a significant improvement (P = 0.01). The accuracy was greater for type 1 lesions when using contrast-modified images, whereas there was no improvement with type 2 and 3 lesions (P = 0.039, P = 0.835, P = 0.193). The average correct answers was greater for type 2 (µ = 5.08) than type 1 (µ = 5.08) and type 3 (µ = 2) (P = 0.046). The graders (n; %) classified the utility of contrast-modified images as essential (1; 2,7%), very useful (29; 78.4%), indifferent (5; 13.5%), not very useful (1; 2.7%) and useless (1; 2.7%). In challenging cases, it might be easier to locate the retinal pigment epithelium (RPE), evaluate RPE integrity, classify MNV, and study subretinal hyperreflective material using contrast enhanced OCT images. The accuracy for MNV classification increases with contrast-modified images in type 1 MNV. It is an accessible tool that could show even better performance as the ophthalmologist gains more experience using it.

Aims/Purpose: The detection of subretinal fluid (SRF) and intraretinal fluid (IRF) in optical coherence tomography (OCT) images plays a crucial role in the diagnosis and treatment of certain eye diseases. In this study, we investigated the performance of a Transfer Learning artificial intelligence (AI) technique, which we used to perform the model adaptation from the domain of Heidelberg OCT images to Optopol devices.Methods: A data set of 133 Optopol OCT images. 69 images indicating subretinal or intraretinal fluid and 64 fluid‐free images, were analysed. The AI algorithm that was initially trained on OCT images from Heidelberg device was fine‐tuned to perform the biomarker segmentation in Optopol (Revo NX, SOCT Copernicus REVO, REVO FC) OCT images. Manual assessments by the medical experts were used to evaluate the algorithm's performance.Results: Our findings demonstrate the promising performance of the domain‐adapted model. in detecting SRF and IRF, when initially trained on OCT images from a different device. The algorithm achieved an accuracy of 80% while classifying the presence versus absence of the fluids on the balanced test data set.Conclusions: The study demonstrates a successful application of a transfer learning technique for detection of subretinal and intraretinal fluids in OCT images acquired with the Optopol devices, indicating its potential for enhancing eye diagnostics. Further studies utilizing larger data sets, further biomarkers and diverse OCT devices are warranted to validate the algorithm's robustness and generalizability.

The objective of this study was to build deep learning models with optical coherence tomography (OCT) images to classify normal and age related macular degeneration (AMD), AMD with fluid, and AMD without any fluid. In this study, 185 normal OCT images from 49 normal subjects, 535 OCT images of AMD with fluid, and 514 OCT mages of AMD without fluid from 120 AMD eyes as training data, while 49 normal images from 25 normal eyes, 188 AMD OCT images with fluid and 154 AMD images without any fluid from 77 AMD eyes as test data, were enrolled. Data augmentation was applied to increase the number of images to build deep learning models. Totally, two classification models were built in two steps. In the first step, a VGG16 model pre-trained on ImageNet dataset was transfer learned to classify normal and AMD, including AMD with fluid and/or without any fluid. Then, in the second step, the fine-tuned model in the first step was transfer learned again to distinguish the images of AMD with fluid from the ones without any fluid. With the first model, normal and AMD OCT images were classified with 0.999 area under receiver operating characteristic curve (AUC), and 99.2% accuracy. With the second model, AMD with the presence of any fluid, and AMD without fluid were classified with 0.992 AUC, and 95.1% accuracy. Compared with a transfer learned VGG16 model pre-trained on ImageNet dataset, to classify the three categories directly, higher classification performance was achieved with our notable approach. Conclusively, two classification models for AMD clinical practice were built with high classification performance, and these models should help improve the early diagnosis and treatment for AMD.

PurposeWe introduce a deep-learning-based biomarker proposal system for the purpose of accelerating biomarker discovery in age-related macular degeneration (AMD). DesignRetrospective analysis of a large dataset of retinal optical coherence tomography (OCT) images. ParticipantsA total of 3,456 adults aged between 51 and 102 years of whom OCT images were collected under the PINNACLE project. MethodsOur system proposes candidates for novel AMD imaging biomarkers in OCT. It works by first training a neural network using self-supervised contrastive learning to discover, without any clinical annotations, features relating to both known and unknown AMD biomarkers present in 46,496 retinal OCT images. To interpret the learned biomarkers, we partition the images into 30 subsets, termed clusters, that contain similar features. We conduct two parallel 1.5-hour semi-structured interviews with two independent teams of retinal specialists to assign descriptions in clinical language to each cluster. Descriptions of clusters achieving consensus can potentially inform new biomarker candidates. Main Outcome MeasuresWe checked if each cluster showed clear features comprehensible to retinal specialists, if they related to AMD and how many described established biomarkers used in grading systems as opposed to recently proposed or potentially new biomarkers. We also compared their prognostic value for late stage wet and dry AMD against an established clinical grading system and a demographic baseline model. ResultsOverall, both teams independently identified clearly distinct characteristics in 27 of 30 clusters, of which 23 were related to AMD. Seven were recognised as known biomarkers used in established grading systems and 16 depicted biomarker combinations or subtypes that are either not yet used in grading systems, were only recently proposed, or were unknown. Clusters separated incomplete from complete retinal atrophy, intraretinal from subretinal fluid and thick from thin choroids, and in simulation outperformed clinically-used grading systems in prognostic value. ConclusionsUsing self-supervised deep learning we were able to automatically propose AMD biomarkers going beyond the set used in clinically established grading systems. Without any clinical annotations, contrastive learning discovered subtle differences between fine-grained biomarkers. Ultimately, we envision that equipping clinicians with discovery-oriented deep-learning tools can accelerate discovery of novel prognostic biomarkers.

Use of a Neural Net to Model the Impact of Optical Coherence Tomography Abnormalities on Vision in Age-related Macular Degeneration

The purpose of this study was to quantitatively evaluate retinal pigment epithelium (RPE) changes in serous pigment epithelial detachment (PED) among patients with age-related macular degeneration by means of prototype multi-contrast optical coherence tomography (OCT), which is capable of simultaneous collection of OCT angiography, polarization-sensitive OCT, and standard OCT images. We evaluated 26 eyes of 21 patients with serous PED. RPE-melanin OCT images were calculated from the multi-contrast OCT dataset and compared with near-infrared autofluorescence images. An active RPE lesion was defined as an area of thickened RPE-melanin (≥ 70 μm; RPE70) on RPE-melanin OCT. Each PED area was divided into peak and slope regions. RPE70 area ratios were compared with the maximum PED height, PED area, PED volume, and slope area ratio (area of slope region/area of whole PED). RPE-melanin OCT images were consistent with near-infrared autofluorescence images. The RPE70 area ratio in the slope region was significantly negatively correlated with the slope area ratio. Development of active RPE lesions in the slope region was correlated with the PED configuration. Multi-contrast OCT is useful for objective evaluation of changes in the RPE in patients with age-related macular degeneration.

The purpose of this study was to evaluate focal damage in the retinal pigment epithelium (RPE) layer in serous retinal pigment epithelium detachment (PED) with multi-contrast optical coherence tomography (OCT), which is capable of simultaneous measurement of OCT angiography, polarization-sensitive OCT and standard OCT images. We evaluated 37 eyes with age-related macular degeneration that had serous PED. Focal RPE damage was indicated by hyper-transmission beneath the RPE-Bruch’s membrane band in standard OCT images. Distribution of RPE melanin was calculated using the dataset from multi-contrast OCT. Twenty-four points with hyper-transmission were detected in 21 of the 37 eyes. Standard OCT images failed to show disruption of the RPE-Bruch’s membrane band at 5 of the 24 hyper-transmission points. Conversely, multi-contrast OCT images clearly showed melanin defects in the RPE-Bruch’s membrane band at all points. Areas of melanin defects with disruption of the RPE-Bruch’s membrane band were significantly larger than those without disruption. The volume of intraretinal hyper-reflective foci was significantly larger in eyes with hyper-transmission than that in eyes without hyper-transmission. Multi-contrast OCT is more sensitive than standard OCT for displaying changes at the RPE-Bruch’s membrane band when there are small areas of RPE damage.

PurposeTo quantify ellipsoid zone (EZ) loss during anti‐VEGF therapy for neovascular age‐related macular degeneration (nAMD) and correlate these findings with nAMD disease activity using artificial intelligence‐based algorithms.MethodsSpectral domain optical coherence tomography (Spectralis, Heidelberg Engineering) images from nAMD treatment‐naïve patients from the Fight Retinal Blindness! (FRB!) Registry from Zürich, Switzerland were processed at baseline and over 3 years of follow‐up. An approved deep learning algorithm (Fluid Monitor, RetInSight) was used to automatically quantify intraretinal fluid (IRF), subretinal fluid (SRF) and pigment epithelial detachment (PED). An ensemble U‐net deep learning algorithm was used to automated quantify EZ integrity based on EZ layer thickness. The impact of fluid volumes on EZ thickness and late‐stages outcomes were calculated using Wilcoxon rank‐sum tests, a linear mixed model and a longitudinal panel regression model.ResultsTwo hundred and eleven eyes from 158 patients were included. The mean ± SD EZ loss area in the central 6 mm was 1.81 ± 2.68 mm2 at baseline and reached 6.21 ± 6.15 mm2 at month 36. Higher fluid volumes (top 25%) of IRF and PED in the central 1 and 6 mm of the macula were significantly associated with more advanced EZ thinning and loss compared to the low fluid volume subgroup. The high SRF subgroup in the linear regression model showed no statistically significant association with EZ integrity in the central macula; however, the longitudinal analysis revealed an increased EZ thickness with no additional loss.ConclusionsIntraretinal fluid and PED volumes and their resolution pattern have an impact on alteration of the underlying EZ layer. AI‐supported quantifications are helpful in quantifying early signs of macular atrophy and providing individual risk profiles as a basis for tailored therapies for optimized visual outcomes.

Purpose: To highlight the diagnostic challenge of eyes with hypotony maculopathy with concurrent macular diseases and to present optical coherence tomography (OCT) imaging findings in hypotony maculopathy. Methods: Retrospective review of 15 eyes of 12 patients with hypotony maculopathy at a single institution. Results: Low intraocular pressure (IOP) was caused by overfiltration from trabeculectomy in 9 eyes (of 15, 60%), overtreatment with topical aqueous suppressants after trabeculectomy in 3 eyes (of 15, 20%), bleb leak in 1 eye (of 15, 7%), and treatment with topical aqueous suppressants alone in 2 eyes (of 15, 13%). On OCT imaging, 7 eyes (of 15, 47%) had chorioretinal folds without intraretinal fluid (IRF) or subretinal fluid (SRF), 5 eyes (of 15, 33%) had either IRF or SRF with chorioretinal folds, and 3 eyes (of 15, 20%) had IRF or SRF without chorioretinal folds. The majority of eyes (11 of 15 eyes, 73%) had additional diagnoses apart from hypotony maculopathy that could have potentially caused IRF or SRF, which included epiretinal membrane, retinal vein occlusion, and age-related macular degeneration. Two eyes (of 15, 13%) were initially erroneously managed for a nonhypotony maculopathy pathology due to the diagnostic challenge in eyes with hypotony maculopathy. Conclusion: Hypotony maculopathy demonstrates nonspecific OCT findings such as IRF and SRF that are commonly seen in other macular diseases. In eyes with these macular diseases, diagnosing hypotony maculopathy may be challenging; thus, attention to past ocular history and IOP must always be part of evaluation.

IntroductionThe use of optical coherence tomography (OCT) images is increasing in the medical treatment of age-related macular degeneration (AMD), and thus, the amount of data requiring analysis is increasing. Advances in machine-learning techniques may facilitate processing of large amounts of medical image data. Among deep-learning methods, convolution neural networks (CNNs) show superior image recognition ability. This study aimed to build deep-learning models that could distinguish AMD from healthy OCT scans and to distinguish AMD with and without exudative changes without using a segmentation algorithm.MethodsThis was a cross-sectional observational clinical study. A total of 1621 spectral domain (SD)-OCT images of patients with AMD and a healthy control group were studied. The first CNN model was trained and validated using 1382 AMD images and 239 normal images. The second transfer-learning model was trained and validated with 721 AMD images with exudative changes and 661 AMD images without any exudate. The attention area of the CNN was described as a heat map by class activation mapping (CAM). In the second model, which classified images into AMD with or without exudative changes, we compared the learning stabilization of models using or not using transfer learning.ResultsUsing the first CNN model, we could classify AMD and normal OCT images with 100% sensitivity, 91.8% specificity, and 99.0% accuracy. In the second, transfer-learning model, we could classify AMD as having or not having exudative changes, with 98.4% sensitivity, 88.3% specificity, and 93.9% accuracy. CAM successfully described the heat-map area on the OCT images. Including the transfer-learning model in the second model resulted in faster stabilization than when the transfer-learning model was not included.ConclusionTwo computational deep-learning models were developed and evaluated here; both models showed good performance. Automation of the interpretation process by using deep-learning models can save time and improve efficiency.Trial RegistrationNo15073.

Age-related macular degeneration (ARMD) is a degenerative eye disease among people over the age of 50. This disorder affects the retina and macula of the eye. The macula is the central part of the retina. The function of the macular region is to give a sharp and clear vision. ARMD is mainly identified by the presence of yellow deposits called drusen and the structural changes in the retinal pigment epithelium (RPE) layer of the retina. If ARMD is not identified at an early stage, it will lead to permanent blindness. The ARMD is mainly diagnosed using Color fundus and Optical coherence tomography (OCT) images of the retina. OCT is an imaging technique that makes use of low-coherence light to capture two-and three-dimensional images from within optical scattering media. OCT images usually are at a micrometer resolution, affected by a noise called speckle noise. If speckle noise is present in an OCT image, diagnosis of ARMD is very difficult, irrespective of whether it is manually or automatically. So, removal of speckle-noise is significant for the accurate diagnosis and early detection of ARMD. In this paper, we evaluate the performance of eight popular image noise reduction filters for its ability to remove speckle noise in terms of different performance measures on a publicly available OCT dataset. Our evaluation shows that, out of the filters selected for assessment, a fast non-local mean filter outperforms all other filters regarding all the performance measures used.

Background: To explain the biological role of cytokines in the eye and the possible role of cytokines in the pathogenesis of neovascular age-related macular degeneration (nAMD) by comparing the correlation between cytokine of aqueous humor concentration and optical coherence tomography (OCT) retinal fluid.Methods: Spectral-domain OCT (SD-OCT) images and aqueous humor samples were collected from 20 nAMD patient’s three clinical visits. Retinal fluid volume in OCT was automatically quantified using deep learning--Deeplabv3+. Eighteen cytokines were detected in aqueous humor using the Luminex technology. OCT fluid volume measurements were correlated with changes in aqueous humor cytokine levels using Pearson’s correlation coefficient (PCC).Results: The patients with intraretinal fluid (IRF) showed significantly lower levels of cytokines, such as C-X-C motif chemokine ligand 2 (CXCL2) (p = 0.03) and CXCL11 (p = 0.009), compared with the patients without IRF. And the IRF volume was negatively correlated with CXCL2 (r = −0.407, p = 0.048) and CXCL11 (r = −0.410, p = 0.046) concentration in the patients with IRF. Meanwhile, the subretinal fluid (SRF) volume was positively correlated with vascular endothelial growth factor (VEGF) concentration (r = 0.299, p = 0.027) and negatively correlated with interleukin (IL)-36β concentration (r = −0.295, p = 0.029) in the patients with SRF.Conclusion: Decreased level of VEGF was associated with decreased OCT-based retinal fluid volume in nAMD patients, while increased levels of CXCL2, CXCL11, and IL-36β were associated with decreased OCT-based retinal fluid volume in nAMD patients, which may suggest a role for inflammatory cytokines in retinal morphological changes and pathogenesis of nAMD patients.

PurposeTo analyze the individual value and the contribution of color fundus photography (CFP) and optical coherence tomography (OCT) in the screening of age-related macular degeneration (AMD) of an unselected population.MethodsCFP and OCT images of 15957 eyes of 8069 subjects older than 55 years, obtained during a population-based screening for AMD using a single diagnostic non-mydriatic imaging device, were analyzed by a blinded examiner. The two techniques were preliminary evaluated considering the dichotomous parameter "gradable/ungradable", then gradable images were classified. CFP were graded according to the standardized classification of AMD lesions. OCT images were also categorized considering the presence of signs of early/intermediate AMD, late AMD, or other retinal diseases. Another blinded operator re-graded 1978 randomly selected images (for both CFP and OCT), to assess test reproducibility.ResultsOf the 15957 eyes, 8356 CFP (52.4%) and 15594 (97.7%) OCT scans were gradable. Moreover, most of the eyes with ungradable CFP (7339, 96.6%) were gradable at OCT. AMD signs were revealed in 7.4% of gradable CFP and in 10.4% of gradable OCT images. Moreover, at OCT, AMD signs were found in 1110 (6.9%) eyes whose CFP were ungradable or without AMD (847 and 263 eyes, respectively). The inter-operator agreement was good for the gradable versus ungradable parameter, and optimal for the AMD grading parameter of CFP. The agreement was optimal for all OCT parameters.ConclusionsOCT provided gradable images in almost all examined eyes, compared to limited CFP efficiency. Moreover, OCT images allowed to detect more AMD eyes compared to gradable photos. OCT imaging appears to significantly improve the power of AMD screening in a general, unselected population, compared to CFP alone.

To develop our proposed technology method to improve retinal pigment epithelium (RPE) detection in optical coherence tomography (OCT) images and compare its efficacy with Topcon’s automated segmentation algorithm across multiple retinal diseases and healthy eyes. OCT images from 88 patients with age-related macular degeneration (AMD) were used for our proposed technology model training and validation. For testing with separate images were obtained from patients with AMD (100), diabetic retinopathy (DR; 50), epiretinal membrane (ERM; 50), branch retinal vein occlusion (BRVO; 50), and healthy eyes (50). The proposed technology was used to identify RPE in OCT images using the Pyramid Scene Parsing Network on top of ResNet-50. The accuracy of the proposed technology method in RPE detection was measured using the mean absolute error (MAE) and compared with Topcon’s automated segmentation algorithm for each retinal condition. As compared with Topcon’s automated segmentation algorithm, the proposed technology showed significantly better MAEs across all conditions: AMD (2.18 vs. 4.79), DR (1.69 vs. 3.17), ERM (1.50 vs. 2.67), BRVO (1.86 vs. 2.98), and healthy eyes (1.59 vs. 2.28). Notably, the proposed technology’s superiority was most evident in the AMD group. The proposed technology method outperformed Topcon’s automated segmentation algorithm in accurately visualizing RPE in OCT images across all tested conditions, especially in AMD. Our results indicate the proposed technology’s potential to elevate the RPE segmentation which can lead to enhancing ophthalmology care by providing more accurate OCT imaging analyses.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-25221-y.