Mixed‐decomposed convolutional network: A lightweight yet efficient convolutional neural network for ocular disease recognition

Bartonella: a new etiological agent of feline ocular disease.

Climate change is an emerging public health crisis that extends beyond systemic diseases to greatly impact ocular health. Increasing ultraviolet (UV) radiation exposure, worsening air pollution, and the accumulation of environmental toxins are contributing to a rise in cataracts, dry eye syndrome (DES), and pterygium. Additionally, climate change is accelerating the spread of infectious eye diseases such as trachoma, onchocerciasis, and fungal keratitis, particularly in vulnerable populations with limited access to healthcare. Diagnosing climate-driven ocular diseases presents unique challenges, as early-stage manifestations are often subtle and overlap with other systemic or age-related conditions. Recent advancements in ophthalmic diagnostics, including AI-assisted imaging, anterior segment optical coherence tomography (AS-OCT), tear fluid biomarkers, and public health campaigns, offer promising solutions for early detection and monitoring of ocular diseases. AI-driven deep learning models have demonstrated potential in detecting cataract changes and environmentally related corneal pathologies with high accuracy. Furthermore, portable imaging technologies have been shown to facilitate screening in underserved regions most affected by climate change. This review explores the intricate relationship between climate change and ocular diseases, focusing on the diagnostic challenges posed by environmental factors and innovations aimed at enhancing patient outcomes. Addressing these issues through improved diagnostic strategies, global health policies, and interdisciplinary collaboration is essential to mitigate the long-term ocular health impacts of climate change.

The computational cost of many deep convolutional neural networks (CNNs) for the automatic target recognition proposed in the area of synthetic aperture radar (SAR) imagery recently is huge, and the limited SAR images are always insufficient for training the deep CNNs. To improve the computational efficiency, a new light but very efficient convolutional network architecture is designed using some novel techniques to get the better results. The authors apply the batch normalisation before each convolutional layer in order to reduce ‘internal covariate shift’ and use the drop-out strategy in the fully layer to avoid over-fitting. Additionally, concatenated ReLU is used as activation scheme specially instead of ReLU for preserving the negative phase information to get the double feature maps of the previous layer rather than to increase the depth of the filters that can lessen the parameters of the networks. The results of the experimental demonstrate that the authors’ CNNs can both achieve a state-of-the-art classification accuracy of 99.53% of the SAR target classification in the Moving and Stationary Target Acquisition and Recognition ten classes public dataset and perform well even when the training data is sparse.

Purpose: To investigate iris vasculature changes following scleral buckling (SB) surgery in eyes with rhegmatogenous retinal detachment (RRD) with anterior-segment (AS) optical coherence tomography angiography (OCTA). Methods: In this prospective study, enrolled subjects were imaged with an SS-OCTA system (PLEX Elite 9000, Carl Zeiss Meditec Inc., Dublin, CA, USA). Image acquisition of the iris was obtained using an AS lens and a manual focusing adjustment in the iris using the retina imaging software. The quantitative analysis was performed in eight different iris regions: (i) superior, (ii) supero-temporal, (iii) supero-nasal, (iv) nasal, (v) temporal, (vi) inferior, (vii) infero-temporal, (viii) infero nasal which, were defined as squares with area of 1.5 mm2. Results: Fifteen eyes of 15 patients (six females; nine males) were included. Anterior segment optical coherence tomography angiography (AS-OCTA) parameters of the iris were statistically compared at baseline (preoperatively), 1 week, 1 month and 6 months after SB. At post-operative 1 week, perfusion density (PD) showed a significant decrease from 66.8 ± 13.2% to 58.55 ± 12.0% in the iris supero-nasal region (p = 0.016). However, at the 1-month follow-up visit, iris PD was significantly lower in all the analyzed iris regions, apart from the superior one. Conclusions: This study is the first description of AS-OCTA in patients undergoing SB. Our results showed a uniform reduction of the iris vessel network at 1 month after surgery, supporting the clinical use of AS-OCTA to identify early iris perfusion changes as potential predictive biomarkers of vascular disorders.

AimsTo develop a generative adversarial network (GAN) capable of generating realistic high-resolution anterior segment optical coherence tomography (AS-OCT) images.MethodsThis study included 142 628 AS-OCT B-scans from the American University of...

1 - Classification of ocular diseases using transfer learning approaches and glaucoma severity grading

To evaluate the intra- and inter-grader concordance of anterior segment optical coherence tomography (ASOCT) grading for detection of endothelial plaque in microbial keratitis, and to compare endothelial plaque detection via ASOCT grading versus in-person slit lamp examination. Diagnostic concordance study of 150 consecutive patients with microbiologically confirmed bacterial or fungal keratitis at a high-volume tertiary eye hospital in India. Two masked ophthalmologist graders independently evaluated ASOCT images for presence of two morphologically distinct endothelial plaque subtypes noted during image review (round and flat plaques). We assessed intra-grader and inter-grader concordance for each endothelial plaque morphology and for presence of either morphology. Diagnostic agreement between ASOCT and in-person slit lamp examination was evaluated using percent agreement, Cohen's kappa, sensitivity, and specificity. Univariable and multivariable logistic regression was used to assess odds of disagreement between ASOCT and slit lamp examination for endothelial plaque detection. ASOCT detection showed near perfect inter-grader agreement for round endothelial plaques (kappa 0.88, 94.7% agreement), flat endothelial plaques (kappa 0.84, 92.0% agreement), and either plaque (kappa 0.88, 94.0% agreement). Intra-grader agreement was substantial to near perfect for both graders across all plaque types (kappa 0.70-0.86). Ophthalmologist slit lamp examination identified endothelial plaque in 6.0% eyes, while ASOCT detected round plaques in 32.7%, flat plaques in 43.3%, and either plaque in 55.3% of eyes. Using ASOCT as reference, slit lamp examination demonstrated sensitivity of 16.3% for round plaques, 6.2% for flat plaques, and 9.6% for either plaque, with specificity exceeding 94% for all. Poor visual acuity (logMAR ≥1.0) was associated with increased disagreement for round plaques (adjusted OR 5.04), flat plaques (adjusted OR 3.63), and either plaque (adjusted OR 3.98). Bacterial infection was associated with increased disagreement for any endothelial plaque (adjusted OR 4.56). Slit lamp examination substantially under-detects endothelial plaque compared to ASOCT, while ASOCT enables reproducible detection with excellent intra- and inter-grader agreement. These findings support incorporation of ASOCT imaging into microbial keratitis evaluation protocols. Differences in round and flat endothelial plaque morphologies warrant further investigation.

The presence of preoperative posterior capsular dehiscence in posterior polar cataracts (PPCs) increases the chances of intraoperative posterior capsular rent. Our study aims to demonstrate the efficacy of preoperative anterior segment optical coherence tomography (ASOCT) to predict intraoperative posterior capsular rupture (PCR) in PPC. This was an observational study including patients within 25-75 years of age, having PPC, undergoing phacoemulsification. ASOCT was done for evaluating preoperative posterior capsular status and was graded as "intact" or "dehiscent, and eventually ,intraoperative-posterior capsular status was noted. SPSS (version 27.0) was used, and P value of ≤ 0.05 was considered statistically significant. ASOCT of 44 eyes had 9 (20.5%) dehiscent posterior capsules and 35 (79.5%) intact. Out of these 9 eyes, 7 (77.8%) had intraoperative PCR and 2 (22.2%) did not. Out of these 35, 34 (97.1%) were actually intact intraoperatively and 1 (2.9%) had intraoperative PCR. The sensitivity and specificity of ASOCT for detecting dehiscence were 94.4% and 87.5%, respectively. The positive and negative predictive values were 97.1 and 77.8, respectively. The diagnostic accuracy of the test was 95.45% with a Chi-square value of 27.01 (P < 0.0001). The diagnostic validity of ASOCT grading for detecting preoperative dehiscence was confirmed. The association between intraoperative PCR and preoperative dehiscence was statistically significant (Chi-square test - P < 0.0001). Our objective was established. The demographics and clinical parameters were comparable with other studies. Two eyes with preoperative dehiscence on ASOCT had intact posterior capsule during surgery. The cause can be a dense opacity with increased optical density causing an obscured a clear capsule view on ASOCT, an artifact defect, giving a false impression of dehiscence. One eye who had nondehiscent PC on ASOCT was found to have PCR on the table, which can be due to an extremely thin PC or tightly adherent polar opacity to the capsule. According to our study, the diagnostic accuracy of ASOCT for detecting preoperative dehiscence was 94.4% (sensitivity)and of nondehiscence was 87.5%(specificity). 97.1% of eyes with dehiscence on ASOCT, had actual posterior capsular rent (positive predictive value). To conclude, ASOCT can be used as a reliable diagnostic modality.

Majority of the present day applications in computer vision use Convolutional neural networks (CNN) as the preferred engine of computation. For hardware acceleration of CNNs, z-first storage architectures like NVDLA have gained prominence which show better resource utilization and effective throughput for 3-D convolution operation. However, these architectures have low effective throughput for 2-D operations such as Depthwise Convolution (DWC) and Pooling due to poor resource utilization. The current trend in CNNs show that DWC has gained prominence in compact mobile-first CNNs like MnasNet, EfficientNet etc., which have classification accuracies comparable to state-of-the-art complex CNNs. In this work, an accelerator for accelerating 2-D operations based on z-first storage architectures is proposed. Proposed accelerator delivers power efficiency improvement of at least 1.46× and 1.89× for DWC and Pooling respectively over the baseline architecture, while improving effective throughput by 4×. Proposed IFM reuse scheme also reduces IFM reads from memory at least by a factor of 1.72×.

Millions of people around the world suffer from various ocular diseases often leading to blindness due to delayed detection and treatment. This has led to a demand for quick automated detection process from medical images including retinal fundus images. In this paper, an automatic approach for classifying normal and diseased cases from given retinal fundus images is developed based on ensembling of some suitable deep learning architectures in a transfer learning platform. Instead of directly using the raw images, it is shown that use of an enhancement technique based on adaptive histogram equalization followed by morphological operations can offer better class separation between the normal and diseased images. Some efficient deep convolutional neural network (CNN) based architectures are implemented utilizing the pre-trained weights obtained via transfer learning. In order to achieve significant improvement in the classification performance, the predictions obtained from some selected deep CNN architectures, namely ResNet50, InceptionResNetV2, EfficientNetB0 and EfficientNetB2 are combined. Comprehensive experimentation carried out on an extensive ophthalmic database show promising performance. The wide range of disease and diverse collection conditions of the fundus images affirm the suitability of the method for practical implementation.

Imaging has become indispensable in the diagnosis and management of diseases in the posterior part of the eye. In recent years, imaging techniques for the anterior segment are also gaining importance and are nowadays routinely used in clinical practice. Ocular surface disease is often synonymous with dry eye disease, but also refers to other conditions of the ocular surface, such as Meibomian gland dysfunction or keratitis and conjunctivitis with different underlying causes, i.e., allergies or infections. Therefore, correct differential diagnosis and treatment of ocular surface diseases is crucial, for which imaging can be a helpful tool. A variety of imaging techniques have been introduced to study the ocular surface, such as anterior segment optical coherence tomography, in vivo confocal microscopy, or non-contact meibography. The present review provides an overview on how these techniques can be used in the diagnosis and management of ocular surface disease and compares them to clinical standard methods such as slit lamp examination or staining of the cornea or conjunctiva. Although being more cost-intensive in the short term, in the long term, the use of ocular imaging can lead to more individualized diagnoses and treatment decisions, which in turn are beneficial for affected patients as well as for the healthcare system. In addition, imaging is more objective and provides good documentation, leading to an improvement in patient follow-up and education.

To evaluate the presentation of Kayser-Fleischer (KF) rings in the cornea using anterior segment optical coherence tomography (AS-OCT) in patients with Wilson disease (WD). A retrospective analysis was conducted on the clinical data from 31 patients (62 eyes) who were diagnosed with WD, were positive for KF rings, and who were admitted to our hospital between August 2020 and December 2022. All patients underwent slit-lamp and AS-OCT examinations, with the KF ring width measured on AS-OCT grayscale images. On AS-OCT, the KF ring appeared as a high-reflective band at the level of Descemet membrane. The color band on AS-OCT displayed the ring in shades of yellow-green, yellow, or orange. The width of the KF ring was measured accurately using AS-OCT. AS-OCT, when combined with slit-lamp examination, provides a reliable method for evaluating KF rings in patients with WD. The grayscale measurements obtained via AS-OCT offer precise quantification of the corneal KF ring width.

Purpose To show the use of anterior segment optical coherence tomography (AS‐OCT) in conjunctival, limbal, subconjunctival tumors and to compare the data with histopathologic specimens.Methods 15 patients (15 eyes) with conjunctival, limbal and subconjunctival lesions were examined using AS‐OCT (Visante OCT 2.0, Carl Zeiss Meditec, Dublin, CA). Eyes were scanned using Anterior Segment Single, High Resolution Corneal and Raw Image modes. Gray‐scale, OCT Color and Rainbow Color images were analyzed using built‐in software. All tumors were excised and histopathologicaly examined.Results In all small tumor cases AS‐OCT allowed to see the deeper lesion structure with complete penetration. In conjunctival melanoma, papilloma, subepithelial naevus and pterygia cases AS‐OCT showed tumor structure, layers and extension clearly. Underlying cornea was also visible. In larger and solid tumors (squamous cell carcinoma, fibroma and lipoma) anterior surfaces were hyper‐reflective, but the deeper structures of the lesions were incompletely penetrated. However, in all cases it was possible to differentiate cystic lesions from solid lesions. Histopathologic examination revealed the diagnosis and in small tumors confirmed the AS‐OCT data showing tumor epithelium and underling layers.Conclusion AS‐OCT may be a useful non‐invasive diagnostic tool for the evaluation of ocular surface tumors in selective cases. AS‐OCT may give useful information before planning the surgery.

PurposeConjunctival naevi are the most frequently diagnosed primary melanocytic lesions of the conjunctiva. The clinical manifestations are greatly variable which may result in diagnostic difficulties and differential diagnostic confusions. Therefore aims of the present study were: 1) to assess the morphologic features of conjunctival naevi; 2) to delineate the anterior segment optical coherence tomography (AS-OCT) characteristics of these lesions; 3) to compare AS-OCT and ultrasound biomicroscopy (UBM) as diagnostic tools in these alterations and 4) to correlate histological results with the AS-OCT pictures in case of surgically excised naevi.MethodsAll lesions were photo-documented. AS-OCT and UBM (over the age of 18 years) were performed. Surgically excised lesions were admitted to histological examinations.ResultsIn our series of 57 conjunctival naevi, 54.4% were highly pigmented, 15.8% proved to be amelanotic. AS-OCT could detect intralesional cysts in 61.4% of the naevi, while slit-lamp and UBM proved to be less sensitive (40.3% vs. 28.5%). UBM could visualize the posterior margins of all naevi, while AS-OCT proved to be less sensitive with the detection of 89.4% of posterior naevus margins. Thickness of the conjunctival epithelial layer could be measured with AS-OCT in case of subepithelial naevi, while no distinct epithelial layer could be detected in compound and junctional naevi.ConclusionsSuperiority of AS-OCT over UBM was demonstrated in visualizing internal structures of conjunctival naevi. UBM proved to be a better tool in highly pigmented and remarkably elevated naevi. Correlation was found between the histological type of the naevus and the thickness of the epithelial layer covering the lesion.

Infectious keratitis is an important cause of visual loss. The purpose of this study was to investigate anterior segmentoptical coherence tomography patterns in infectious keratitis and evaluate the role of this tool in the early management of this disorder. In this cross-sectional study, we included patients with proven infectious keratitis, either by culture or therapeutictrial. Subjects underwent baseline anterior segment optical coherence tomography (Spectralis® anterior segment module, HeidelbergEngineering, Germany) performed by the same operator. We used anterior segment optical coherence tomography vertical andhorizontal raster default scans with 6.0 mm scan lines. Twenty-five patients (14 men and 11 women) were included. The most common risk factors identified were ocular trauma (11cases) and contact lens wear (7 cases). Fifteen patients presented bacterial infection; three, fungal infection; two parasitic infection;and five cases presented a negative microbiological culture. Anterior segment optical coherence tomography depicted nine distinctmorphological patterns. Anterior segment optical coherence tomography allows the depth of corneal involvement to be assessed. When the onlypatterns identified were hyperreflective stromal lesion and stromal edema, the visual outcome was better. Cystic spaces were presentin severe bacterial keratitis. Anterior segment optical coherence tomography can complement biomicroscopy, allowing for a better characterization ofcorneal involvement at presentation that can help in staging and providing useful prognostic information.