Automated Segmentation and Quantification of Drusen in Fundus and Optical Coherence Tomography Images for Detection of ARMD

Artificial intelligence in chorioretinal pathology through fundoscopy: a comprehensive review

BACKGROUND

Applications for artificial intelligence (AI) in ophthalmology are continually evolving. Fundoscopy is one of the oldest ocular imaging techniques but remains a mainstay in posterior segment imaging due to its prevalence, ease of use, and ongoing technological advancement. AI has been leveraged for fundoscopy to accomplish core tasks including segmentation, classification, and prediction.

MAIN BODY

In this article we provide a review of AI in fundoscopy applied to representative chorioretinal pathologies, including diabetic retinopathy and age-related macular degeneration, among others. We conclude with a discussion of future directions and current limitations.

SHORT CONCLUSION

As AI evolves, it will become increasingly essential for the modern ophthalmologist to understand its applications and limitations to improve patient outcomes and continue to innovate.

Non-Neovascular Age-Related Macular Degeneration Assessment: Focus on Optical Coherence Tomography Biomarkers

The imagistic evaluation of non-neovascular age-related macular degeneration (AMD) is crucial for diagnosis, monitoring progression, and guiding management of the disease. Dry AMD, characterized primarily by the presence of drusen and retinal pigment epithelium atrophy, requires detailed visualization of the retinal structure to assess its severity and progression. Several imaging modalities are pivotal in the evaluation of non-neovascular AMD, including optical coherence tomography, fundus autofluorescence, or color fundus photography. In the context of emerging therapies for geographic atrophy, like pegcetacoplan, it is critical to establish the baseline status of the disease, monitor the development and expansion of geographic atrophy, and to evaluate the retina's response to potential treatments in clinical trials. The present review, while initially providing a comprehensive description of the pathophysiology involved in AMD, aims to offer an overview of the imaging modalities employed in the evaluation of non-neovascular AMD. Special emphasis is placed on the assessment of progression biomarkers as discerned through optical coherence tomography. As the landscape of AMD treatment continues to evolve, advanced imaging techniques will remain at the forefront, enabling clinicians to offer the most effective and tailored treatments to their patients.

On Machine Learning in Clinical Interpretation of Retinal Diseases Using OCT Images

Optical coherence tomography (OCT) is a noninvasive imaging technique that provides high-resolution cross-sectional retina images, enabling ophthalmologists to gather crucial information for diagnosing various retinal diseases. Despite its benefits, manual analysis of OCT images is time-consuming and heavily dependent on the personal experience of the analyst. This paper focuses on using machine learning to analyse OCT images in the clinical interpretation of retinal diseases. The complexity of understanding the biomarkers present in OCT images has been a challenge for many researchers, particularly those from nonclinical disciplines. This paper aims to provide an overview of the current state-of-the-art OCT image processing techniques, including image denoising and layer segmentation. It also highlights the potential of machine learning algorithms to automate the analysis of OCT images, reducing time consumption and improving diagnostic accuracy. Using machine learning in OCT image analysis can mitigate the limitations of manual analysis methods and provide a more reliable and objective approach to diagnosing retinal diseases. This paper will be of interest to ophthalmologists, researchers, and data scientists working in the field of retinal disease diagnosis and machine learning. By presenting the latest advancements in OCT image analysis using machine learning, this paper will contribute to the ongoing efforts to improve the diagnostic accuracy of retinal diseases.

OCT and OCT Angiography Update: Clinical Application to Age-Related Macular Degeneration, Central Serous Chorioretinopathy, Macular Telangiectasia, and Diabetic Retinopathy

Similar to ultrasound adapting soundwaves to depict the inner structures and tissues, optical coherence tomography (OCT) utilizes low coherence light waves to assess characteristics in the eye. Compared to the previous gold standard diagnostic imaging fluorescein angiography, OCT is a noninvasive imaging modality that generates images of ocular tissues at a rapid speed. Two commonly used iterations of OCT include spectral-domain (SD) and swept-source (SS). Each comes with different wavelengths and tissue penetration capacities. OCT angiography (OCTA) is a functional extension of the OCT. It generates a large number of pixels to capture the tissue and underlying blood flow. This allows OCTA to measure ischemia and demarcation of the vasculature in a wide range of conditions. This review focused on the study of four commonly encountered diseases involving the retina including age-related macular degeneration (AMD), diabetic retinopathy (DR), central serous chorioretinopathy (CSC), and macular telangiectasia (MacTel). Modern imaging techniques including SD-OCT, TD-OCT, SS-OCT, and OCTA assist with understanding the disease pathogenesis and natural history of disease progression, in addition to routine diagnosis and management in the clinical setting. Finally, this review compares each imaging technique's limitations and potential refinements.

Real-world effectiveness of screening programs for age-related macular degeneration: amended Japanese specific health checkups and augmented screening programs with OCT or AI

PURPOSE

To investigate the effectiveness of screening and subsequent intervention for age-related macular degeneration (AMD) in Japan.

STUDY DESIGN

Best-case-scenario analysis using a Markov model.

METHODS

The clinical effectiveness and cost-effectiveness of screening for AMD were assessed by calculating the reduction proportion of blindness and the incremental cost-effectiveness ratio (ICER). The Markov model simulation began at screening at the age of 40 years and ended at screening at the age of 90 years. The first-eye and second-eye combined model assumed annual state-transition probabilities in the development and treatment of AMD. Data on prevalence, morbidity, transition probability, utility value, and treatment costs were obtained from previously published reports. Sensitivity analysis was performed to assess the influence of the parameters.

RESULTS

In the base-case analysis, screening for AMD every 5 years, beginning at age 40 years and ending at age 74 years (reflecting the screening ages of the current Japanese legal "Specific Health Checkups") showed a decrease of 40.7% in the total number of blind patients. The screening program reduced the number of blind people more than did the additional AREDS/AREDS2 formula supplement intake. However, the ICER of screening versus no screening was ¥9,846,411/QALY, which was beyond what people were willing to pay (WTP) in Japan. Sensitivity analysis revealed that neither OCT nor AI improved the ICER, but the scenario in which the prevalence of smoking decreased by 30% improved the ICER (¥4,655,601/QALY) to the level under the WTP.

CONCLUSIONS

Ophthalmologic screening for AMD is highly effective in reducing blindness but is not cost-effective, as demonstrated by a Markov model based on real-world evidence from Japan.

Automatic detection of leakage point in central serous chorioretinopathy of fundus fluorescein angiography based on time sequence deep learning

PURPOSE

To detect the leakage points of central serous chorioretinopathy (CSC) automatically from dynamic images of fundus fluorescein angiography (FFA) using a deep learning algorithm (DLA).

METHODS

The study included 2104 FFA images from 291 FFA sequences of 291 eyes (137 right eyes and 154 left eyes) from 262 patients. The leakage points were segmented with an attention gated network (AGN). The optic disk (OD) and macula region were segmented simultaneously using a U-net. To reduce the number of false positives based on time sequence, the leakage points were matched according to their positions in relation to the OD and macula.

RESULTS

With the AGN alone, the number of cases whose detection results perfectly matched the ground truth was only 37 out of 61 cases (60.7%) in the test set. The dice on the lesion level were 0.811. Using an elimination procedure to remove false positives, the number of accurate detection cases increased to 57 (93.4%). The dice on the lesion level also improved to 0.949.

CONCLUSIONS

Using DLA, the CSC leakage points in FFA can be identified reproducibly and accurately with a good match to the ground truth. This novel finding may pave the way for potential application of artificial intelligence to guide laser therapy.

Validation of a Novel Automated Algorithm to Measure Drusen Volume and Area Using Swept Source Optical Coherence Tomography Angiography

Purpose

The purpose of this study was to validate a novel automated swept source optical coherence tomography angiography (SS-OCTA) algorithm to measure elevations of the retinal pigment epithelium (RPE) in eyes with nonexudative age-related macular degeneration (neAMD).

Methods

Patients with drusen were enrolled in a prospective optical coherence tomography (OCT) study and underwent both spectral domain OCT (SD-OCT) and SS-OCTA imaging at the same visit using the 6 × 6 mm scan patterns. The RPE elevation measurements (square root area and cube root volume) from the SS-OCTA algorithm were compared with the automated validated SD-OCT algorithm on the instrument. Standard deviations of drusen measurements from four repeated scans of another separate set were also calculated to evaluate the reproducibility of the SS-OCTA algorithm.

Results

A total of 53 eyes from 28 patients were scanned on both instruments. A very strong correlation was found between the measurements from the two algorithms (all r > 0.95), although the measurements of the drusen area and volume were all larger from the SS-OCTA instrument. The reproducibility of the new SS-OCTA algorithm was analyzed using a sample of 66 eyes from 43 patients. The intraclass correlation coefficient (ICC) was greater than 99% from different macular regions for both the square root area and cube root volume measurements.

Conclusions

A novel automated SS-OCTA algorithm for the quantitative assessment of drusen was validated against the SD-OCT algorithm and was shown to be highly reproducible.

Translational Relevance

This novel SS-OCTA algorithm provides a strategy to measure the area and volume of drusen to assess disease progression in neAMD.

RPE layer detection and baseline estimation using statistical methods and randomization for classification of AMD from retinal OCT

BACKGROUND AND OBJECTIVE

Age-related macular degeneration (AMD) is a condition of the eye that affects the aged people. Optical coherence tomography (OCT) is a diagnostic tool capable of analyzing and identifying the disease affected retinal layers with high resolution. The objective of this work is to extract the retinal pigment epithelium (RPE) layer and the baseline (natural eye curvature, particular to every patient) from retinal spectral-domain OCT (SD-OCT) images. It uses them to find the height of drusen (abnormalities) in the RPE layer and classify it as AMD or normal.

METHODS

In the proposed work, the contrast enhancement based adaptive denoising technique is used for speckle elimination. Pixel grouping and iterative elimination based on the knowledge of typical layer intensities and positions are used to obtain the RPE layer. Using this estimate, randomization techniques are employed, followed by polynomial fitting and drusen removal to arrive at a baseline estimate. The classification is based on the drusen height obtained by taking the difference between the RPE and baseline levels. We have used a patient, wise classification approach where a patient is classified diseased if more than a threshold number of patient images have drusen of more than a certain height. Since all slices of an affected patient will not show drusen, we are justified in adopting this technique.

RESULTS

The proposed method is tested on a public data set of 2130 images/slices, which belonged to 30 patient volumes (15 AMD and 15 Normal) and achieved an overall accuracy of 96.66%, with no false positives. In comparison with existing works, the proposed method achieved higher overall accuracy and a better baseline estimate.

CONCLUSIONS

The proposed work focuses on AMD/normal classification using a statistical approach. It does not require any training. The proposed method modifies the motion restoration paradigm to obtain an application-specific denoising algorithm. The existing RPE detection algorithm is modified significantly to make it robust and applicable even to images where the RPE is not very evident/there is a significant amount of perforations (drusen). The baseline estimation algorithm employs a powerful combination of randomization, iterative polynomial fitting, and pixel elimination in contrast to mere fitting techniques. The main highlight of this work is, it achieved an exact estimation of the baseline in the retinal image compared to the existing methods.

Feasibility of support vector machine learning in age-related macular degeneration using small sample yielding sparse optical coherence tomography data

PURPOSE

A retrospective pilot study is conducted to demonstrate the utility of a novel support vector machine learning (SVML) algorithm in a small three-dimensional (3D) sample yielding sparse optical coherence tomography (spOCT) data for the automatic monitoring of neovascular (wet) age-related macular degeneration (wAMD).

METHODS

From the anti-vascular endothelial growth factor injection database, 588 consecutive pairs of OCT volumes (57.624 B-scans) were selected in 70 randomly chosen wAMD patients treated with ranibizumab. The SVML algorithm was applied to 183 OCT volume pairs (17.934 B-scans) in 30 patients. Four independent, diagnosis-blinded retina specialists indicated whether wAMD activity was present between 100 pairs of consecutive OCT volumes (9800 B-scans) in the remaining 40 patients for comparison with the SVML algorithm and a non-complex baseline algorithm using only retinal thickness. The SVML algorithm was assessed using inter-observer variability and receiver operating characteristic (ROC) analyses.

RESULTS

The retina specialists showed an average Cohen's κ of 0.57 ± 0.13 (minimum: 0.41, maximum: 0.83). The average κ between the proposed algorithm and the retina specialists was 0.62 ± 0.05 and 0.43 ± 0.14 between the baseline algorithm and the retina specialists. Using each of the four retina specialists as the reference, the proposed method showed a superior area under the ROC curve of 0.91 ± 0.03 compared to the ROC 0.81 ± 0.05 shown by the baseline algorithm.

CONCLUSION

The SVML algorithm was as effective as the retina specialists were in detecting activity in wAMD. Support vector machine learning (SVML) may be a useful monitoring tool in wAMD suited for small samples that yield sparse OCT data possibly derived from self-measuring OCT-robots.

A Hybrid Geometric Spatial Image Representation for scene classification

The recent development in the technology has increased the complexity of image contents and demand for image classification becomes more imperative. Digital images play a vital role in many applied domains such as remote sensing, scene analysis, medical care, textile industry and crime investigation. Feature extraction and image representation is considered as an important step in scene analysis as it affects the image classification performance. Automatic classification of images is an open research problem for image analysis and pattern recognition applications. The Bag-of-Features (BoF) model is commonly used to solve image classification, object recognition and other computer vision-based problems. In BoF model, the final feature vector representation of an image contains no information about the co-occurrence of features in the 2D image space. This is considered as a limitation, as the spatial arrangement among visual words in image space contains the information that is beneficial for image representation and learning of classification model. To deal with this, researchers have proposed different image representations. Among these, the division of image-space into different geometric sub-regions for the extraction of histogram for BoF model is considered as a notable contribution for the extraction of spatial clues. Keeping this in view, we aim to explore a Hybrid Geometric Spatial Image Representation (HGSIR) that is based on the combination of histograms computed over the rectangular, triangular and circular regions of the image. Five standard image datasets are used to evaluate the performance of the proposed research. The quantitative analysis demonstrates that the proposed research outperforms the state-of-art research in terms of classification accuracy.

Deep learning approach for the detection and quantification of intraretinal cystoid fluid in multivendor optical coherence tomography

We developed a deep learning algorithm for the automatic segmentation and quantification of intraretinal cystoid fluid (IRC) in spectral domain optical coherence tomography (SD-OCT) volumes independent of the device used for acquisition. A cascade of neural networks was introduced to include prior information on the retinal anatomy, boosting performance significantly. The proposed algorithm approached human performance reaching an overall Dice coefficient of 0.754 ± 0.136 and an intraclass correlation coefficient of 0.936, for the task of IRC segmentation and quantification, respectively. The proposed method allows for fast quantitative IRC volume measurements that can be used to improve patient care, reduce costs, and allow fast and reliable analysis in large population studies.