Merging Information From Infrared and Autofluorescence Fundus Images for Monitoring of Chorioretinal Atrophic Lesions

Development of a deep learning model to distinguish the cause of optic disc atrophy using retinal fundus photography

The differential diagnosis for optic atrophy can be challenging and requires expensive, time-consuming ancillary testing to determine the cause. While Leber's hereditary optic neuropathy (LHON) and optic neuritis (ON) are both clinically significant causes for optic atrophy, both relatively rare in the general population, contributing to limitations in obtaining large imaging datasets. This study therefore aims to develop a deep learning (DL) model based on small datasets that could distinguish the cause of optic disc atrophy using only fundus photography. We retrospectively reviewed fundus photographs of 120 normal eyes, 30 eyes (15 patients) with genetically-confirmed LHON, and 30 eyes (26 patients) with ON. Images were split into a training dataset and a test dataset and used for model training with ResNet-18. To visualize the critical regions in retinal photographs that are highly associated with disease prediction, Gradient-Weighted Class Activation Map (Grad-CAM) was used to generate image-level attention heat maps and to enhance the interpretability of the DL system. In the 3-class classification of normal, LHON, and ON, the area under the receiver operating characteristic curve (AUROC) was 1.0 for normal, 0.988 for LHON, and 0.990 for ON, clearly differentiating each class from the others with an overall total accuracy of 0.93. Specifically, when distinguishing between normal and disease cases, the precision, recall, and F1 scores were perfect at 1.0. Furthermore, in the differentiation of LHON from other conditions, ON from others, and between LHON and ON, we consistently observed precision, recall, and F1 scores of 0.8. The model performance was maintained until only 10% of the pixel values of the image, identified as important by Grad-CAM, were preserved and the rest were masked, followed by retraining and evaluation.

Recent Advances in Imaging Macular Atrophy for Late-Stage Age-Related Macular Degeneration

Age-related macular degeneration (AMD) is a leading cause of blindness worldwide. In late-stage AMD, geographic atrophy (GA) of dry AMD or choroidal neovascularization (CNV) of neovascular AMD eventually results in macular atrophy (MA), leading to significant visual loss. Despite the development of innovative therapies, there are currently no established effective treatments for MA. As a result, early detection of MA is critical in identifying later central macular involvement throughout time. Accurate and early diagnosis is achieved through a combination of clinical examination and imaging techniques. Our review of the literature depicts advances in retinal imaging to identify biomarkers of progression and risk factors for late AMD. Imaging methods like fundus photography; dye-based angiography; fundus autofluorescence (FAF); near-infrared reflectance (NIR); optical coherence tomography (OCT); and optical coherence tomography angiography (OCTA) can be used to detect and monitor the progression of retinal atrophy. These evolving diverse imaging modalities optimize detection of pathologic anatomy and measurement of visual function; they may also contribute to the understanding of underlying mechanistic pathways, particularly the underlying MA changes in late AMD.

Measuring Inflammation in the Vitreous and Retina: A Narrative Review

Uveitis consists of a group of syndromes characterised by intraocular inflammation, accounting for up to 15% of visual loss in the western world and 10% worldwide. Assessment of intraocular inflammation has been limited to clinician-dependent, subjective grading. Developments in imaging technology, such as optical coherence tomography (OCT), have enabled the development of objective, quantitative measures of inflammatory activity. Important quantitative metrics including central macular thickness and vitreous signal intensity allow longitudinal monitoring of disease activity and can be used in conjunction with other imaging modalities enabling holistic assessment of ocular inflammation. Ongoing work into the validation of instrument-based measures alongside development of core outcome sets is crucial for standardisation of clinical trial endpoints and developing guidance for quantitative multi-modal imaging approaches. This review outlines methods of grading inflammation in the vitreous and retina, with a focus on the use of OCT as an objective measure of disease activity.

Structural Endpoints and Outcome Measures in Uveitis

Most uveitis entities are rare diseases but, taken together, are responsible for 5-10% of worldwide visual impairment which largely affects persons of working age. As with many rare diseases, there is a lack of high-level evidence regarding its clinical management, partly due to a dearth of reliable and objective quantitative endpoints for clinical trials. This review provides an overview of available structural outcome measures for uveitis disease activity and damage in an anatomical order from the anterior to the posterior segment of the eye. While there is a multitude of available structural outcome measures, not all might qualify as endpoints for clinical uveitis trials, and thorough testing of applicability is warranted. Furthermore, a consensus on endpoint definition, standardization, and "core outcomes" is required. As stipulated by regulatory agencies, endpoints should be precisely defined, clinically important, internally consistent, reliable, responsive to treatment, and relevant for the respective subtype of uveitis. Out of all modalities used for assessment of the reviewed structural outcome measures, optical coherence tomography, color fundus photography, fundus autofluorescence, and fluorescein/indocyanine green angiography represent current "core modalities" for reliable and objective quantification of uveitis outcome measures, based on their practical availability and the evidence provided so far.

Recent Advances and Clinical Application of Color Scanning Laser Ophthalmoscope

Scanning laser ophthalmoscopes (SLOs) have been available since the early 1990s, but they were not commonly used because their advantages were not enough to replace conventional color fundus photography. In recent years, color SLOs have improved significantly, and the colored SLO images are obtained by combining multiple SLO images taken by lasers of different wavelengths. A combination of these images of different lasers can create an image that is close to that of the real ocular fundus. One advantage of the advanced SLOs is that they can obtain images with a wider view of the ocular fundus while maintaining a high resolution even through non-dilated eyes. The current SLOs are superior to the conventional fundus photography in their ability to image abnormal alterations of the retina and choroid. Thus, the purpose of this review was to present the characteristics of the current color SLOs and to show how that can help in the diagnosis and the following of changes after treatments. To accomplish these goals, we will present our findings in patients with different types of retinochoroidal disorders.

Morphological Changes and Prognostic Factors before and after Photodynamic Therapy for Central Serous Chorioretinopathy

Central serous chorioretinopathy (CSC) is a disease of unknown etiology, but half-dose photodynamic therapy (hPDT) is well known to be effective for CSC. Infrared reflectance (IR) has been shown to be effective for detecting retinal pigmented epithelial and choroidal lesions, but no reports have focused on chorioretinal changes using IR images after as compared to before hPDT. This study aimed to clarify the features of IR images as well as retinal and choroidal morphological changes before and after treatment with verteporfin hPDT for CSC. We also examined prognostic factors associated with CSC treatment. This was a retrospective study that included 140 eyes of 140 patients (male/female ratio 122:18, mean age 53.4 ± 10.8 years) diagnosed with CSC who underwent hPDT in our hospital during the period from April 2015 to December 2018. We determined changes in visual acuity, therapeutic efficacy, central retinal thickness (CRT), central choroidal thickness (CCT), and IR images at one and three months after hPDT as compared to before treatment. Dry macula was defined as a complete resolution of serous retinal detachment after hPDT. History of smoking, disease duration, presence of drusen, presence of retinal pigment epithelium abnormalities, type of fluorescein angiographic leakage, and presence of choroidal vascular hyperpermeability were investigated as prognostic factors associated with treatment efficacy. CRT and CCT were measured using optical coherence tomography (Spectralis HRA-2; Heidelberg Engineering), and IR images after versus before treatment were compared using ImageJ software (version 1.52) to calculate the mean luminance for a 3 × 3 mm area in the macula. Compared with the values before treatment, CCT, CRT, and visual acuity showed significant improvements at one and three months after treatment, and the mean luminance of IR images was also significantly increased. Furthermore, the luminance on IR images tended to rise, though the values at one month and three months after treatment did not differ significantly. Disease duration was significantly associated with dry macula one month after treatment, and visual acuity and CRT before hPDT were both significantly related to dry macula three months after treatment. IR images tended to improve over time, from before treatment through one and three months after hPDT.

Deep learning-based classification of retinal atrophy using fundus autofluorescence imaging

PURPOSE

To automatically classify retinal atrophy according to its etiology, using fundus autofluorescence (FAF) images, using a deep learning model.

METHODS

In this study, FAF images of patients with advanced dry age-related macular degeneration (AMD), also called geographic atrophy (GA), and genetically confirmed inherited retinal diseases (IRDs) in late atrophic stages [Stargardt disease (STGD1) and Pseudo-Stargardt Pattern Dystrophy (PSPD)] were included. The FAF images were used to train a multi-layer deep convolutional neural network (CNN) to differentiate on FAF between atrophy in the context of AMD (GA) and atrophy secondary to IRDs. Three-hundred fourteen FAF images were included, of which 110 images were of GA eyes and 204 were eyes with genetically confirmed STGD1 or PSPD. In the first approach, the CNN was trained and validated with 251 FAF images. Established augmentation techniques were used and an Adam optimizer was used for training. For the subsequent testing, the built classifiers were then tested with 63 untrained FAF images. The visualization method was integrated gradient visualization. In the second approach, 10-fold cross-validation was used to determine the model's performance.

RESULTS

In the first approach, the best performance of the model was obtained using 10 epochs, with an accuracy of 0.92 and an area under the curve for Receiver Operating Characteristic (AUC-ROC) of 0.981. Mean accuracy was 87.30 ± 2.96. In the second approach, a mean accuracy of 0.79 ± 0.06 was obtained.

CONCLUSION

This study describes the use of a deep learning-based algorithm to automatically classify atrophy on FAF imaging according to its etiology. Accurate differential diagnosis between GA and late-onset IRDs masquerading as GA on FAF can be performed with good accuracy and AUC-ROC values.

Deep Learning-Based Classification of Inherited Retinal Diseases Using Fundus Autofluorescence

Background. In recent years, deep learning has been increasingly applied to a vast array of ophthalmological diseases. Inherited retinal diseases (IRD) are rare genetic conditions with a distinctive phenotype on fundus autofluorescence imaging (FAF). Our purpose was to automatically classify different IRDs by means of FAF images using a deep learning algorithm. Methods. In this study, FAF images of patients with retinitis pigmentosa (RP), Best disease (BD), Stargardt disease (STGD), as well as a healthy comparable group were used to train a multilayer deep convolutional neural network (CNN) to differentiate FAF images between each type of IRD and normal FAF. The CNN was trained and validated with 389 FAF images. Established augmentation techniques were used. An Adam optimizer was used for training. For subsequent testing, the built classifiers were then tested with 94 untrained FAF images. Results. For the inherited retinal disease classifiers, global accuracy was 0.95. The precision-recall area under the curve (PRC-AUC) averaged 0.988 for BD, 0.999 for RP, 0.996 for STGD, and 0.989 for healthy controls. Conclusions. This study describes the use of a deep learning-based algorithm to automatically detect and classify inherited retinal disease in FAF. Hereby, the created classifiers showed excellent results. With further developments, this model may be a diagnostic tool and may give relevant information for future therapeutic approaches.