Automated detection of severe diabetic retinopathy using deep learning method

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.

A novel deep learning model for diabetic retinopathy detection in retinal fundus images using pre-trained CNN and HWBLSTM

Diabetic retinopathy (DR) is a global visual indicator of diabetes that leads to blindness and loss of vision. Manual testing presents a more difficult task when attempting to detect DR due to the complexity and variances of DR. Early detection and treatment prevent the diabetic patients from visual loss. Also classifying the intensity and levels of DR is crucial to provide necessary treatment. This study develops a novel deep learning (DL) approach called He Weighted Bi-directional Long Short-term Memory (HWBLSTM) with an effective transfer learning technique for detecting DR from the RFI. The collected fundus images initially undergo preprocessing to improve their quality, which includes noise removal and contrast enhancement using a Hybrid Gaussian Filter and probability density Function-based Gamma Correction (HGFPDFGC) technique. The segmentation procedure divides the image into subgroups and is crucial for accurate detection and classification. The segmentation of the study initially removes the optical disk (OD) and blood vessels (BVs) from the preprocessed images using mathematical morphological operations. Next, it segments the retinal lesions from the OD and BV removed images using the Enhanced Grasshopper Optimization-based Region Growing Algorithm (EGORGA). Then, the features from the segmented retinal lesions are learned using a Squeeze Net (SQN), and the dimensionality reduction of the extracted features is done using the Modified Singular Value Decomposition (MSVD) approach. Finally, the classification is performed by employing the HWBLSTM approach, which classifies the DR abnormalities in datasets as non-DR (NDR), non-proliferative DR (NPDR), moderate NPDR (MDNPDR), and severe DR, also known as proliferative DR (PDR). The proposed approach is implemented on APTOS as well as MESSIDOR datasets. The outcomes proved that the proposed technique accurately identifies the DR with minimal computation overhead compared to the existing approaches.Communicated by Ramaswamy H. Sarma.

Deep Learning and Machine Learning Algorithms for Retinal Image Analysis in Neurodegenerative Disease: Systematic Review of Datasets and Models

Purpose

Retinal images contain rich biomarker information for neurodegenerative disease. Recently, deep learning models have been used for automated neurodegenerative disease diagnosis and risk prediction using retinal images with good results.

Methods

In this review, we systematically report studies with datasets of retinal images from patients with neurodegenerative diseases, including Alzheimer's disease, Huntington's disease, Parkinson's disease, amyotrophic lateral sclerosis, and others. We also review and characterize the models in the current literature which have been used for classification, regression, or segmentation problems using retinal images in patients with neurodegenerative diseases.

Results

Our review found several existing datasets and models with various imaging modalities primarily in patients with Alzheimer's disease, with most datasets on the order of tens to a few hundred images. We found limited data available for the other neurodegenerative diseases. Although cross-sectional imaging data for Alzheimer's disease is becoming more abundant, datasets with longitudinal imaging of any disease are lacking.

Conclusions

The use of bilateral and multimodal imaging together with metadata seems to improve model performance, thus multimodal bilateral image datasets with patient metadata are needed. We identified several deep learning tools that have been useful in this context including feature extraction algorithms specifically for retinal images, retinal image preprocessing techniques, transfer learning, feature fusion, and attention mapping. Importantly, we also consider the limitations common to these models in real-world clinical applications.

Translational Relevance

This systematic review evaluates the deep learning models and retinal features relevant in the evaluation of retinal images of patients with neurodegenerative disease.

Deep learning-based fundus image analysis for cardiovascular disease: a review

It is well established that the retina provides insights beyond the eye. Through observation of retinal microvascular changes, studies have shown that the retina contains information related to cardiovascular disease. Despite the tremendous efforts toward reducing the effects of cardiovascular diseases, they remain a global challenge and a significant public health concern. Conventionally, predicting the risk of cardiovascular disease involves the assessment of preclinical features, risk factors, or biomarkers. However, they are associated with cost implications, and tests to acquire predictive parameters are invasive. Artificial intelligence systems, particularly deep learning (DL) methods applied to fundus images have been generating significant interest as an adjunct assessment tool with the potential of enhancing efforts to prevent cardiovascular disease mortality. Risk factors such as age, gender, smoking status, hypertension, and diabetes can be predicted from fundus images using DL applications with comparable performance to human beings. A clinical change to incorporate DL systems for the analysis of fundus images as an equally good test over more expensive and invasive procedures may require conducting prospective clinical trials to mitigate all the possible ethical challenges and medicolegal implications. This review presents current evidence regarding the use of DL applications on fundus images to predict cardiovascular disease.

Ocular Torsion in Children with Horizontal Strabismus or Orthophoria

PURPOSE

To report the rate of ocular torsion in children with horizontal strabismus or orthophoria.

METHODS

A retrospective study design was used. Nineteen children were included in the study, including seven girls, aged 4-16 years. All patients were examined for strabismus and 12 were scheduled for surgical intervention. All participants had digital fundus photos (DRSplus, Padova, Italy) of both eyes at presentation, and 5 of 12 also had fundus photos following the strabismus operation. Patient files were reviewed for age, demographic data, type of strabismus, clinical symptoms and signs, orthoptic exams, subjective and objective reports of torsion, inferior oblique overaction, and V pattern. Fundus photos were analyzed for torsion by ImageJ software [ImageJ 1.54f, National Institute of Health, USA]. The disc-foveal angle was calculated for ocular torsion. Disc-foveal angle was defined as the angle formed between a line passing through the center of the optic disc to the fovea and another horizontal line passing through the center of the optic disc, using fundus photographs.

RESULTS

Of the 19 children, 18 had horizontal strabismus: 9 with exotropia and 9 with esotropia. One child was orthophoric with torsional strabismus. Inferior oblique overaction was detected in all but 3 children, while V pattern was documented in 10. Visual acuity was reduced (under 6/12) in four eyes of four children. None were symptomatic for ocular torsion. Although extorsion was documented clinically in 3 of 19 children, it was measurable on fundus photos in all patients before surgery with a mean of 8.7 ± 8.5 degrees and 8.5 ± 9.7 degrees in the right and left eyes, respectively. The mean extorsion in both eyes was 19.7 ± 10.1 degrees and improved to a mean of 15.3 ± 7.9 degrees in the children who were operated on and had documented postoperative fundus photographs.

CONCLUSIONS

Extorsion was detected on fundus photos at a significantly higher rate than in clinical examination. Notably, inferior oblique overaction was mainly associated with torsion. This study demonstrated that torsion is underdiagnosed in clinical examinations, as the children are often asymptomatic, but fundus photos which are easily obtained can improve its detection.

Research progress on diagnosing retinal vascular diseases based on artificial intelligence and fundus images

As the only blood vessels that can directly be seen in the whole body, pathological changes in retinal vessels are related to the metabolic state of the whole body and many systems, which seriously affect the vision and quality of life of patients. Timely diagnosis and treatment are key to improving vision prognosis. In recent years, with the rapid development of artificial intelligence, the application of artificial intelligence in ophthalmology has become increasingly extensive and in-depth, especially in the field of retinal vascular diseases. Research study results based on artificial intelligence and fundus images are remarkable and provides a great possibility for early diagnosis and treatment. This paper reviews the recent research progress on artificial intelligence in retinal vascular diseases (including diabetic retinopathy, hypertensive retinopathy, retinal vein occlusion, retinopathy of prematurity, and age-related macular degeneration). The limitations and challenges of the research process are also discussed.

Progress of Imaging in Diabetic Retinopathy-From the Past to the Present

Advancement of imaging technology in retinal diseases provides us more precise understanding and new insights into the diseases' pathologies. Diabetic retinopathy (DR) is one of the leading causes of sight-threatening retinal diseases worldwide. Colour fundus photography and fluorescein angiography have long been golden standard methods in detecting retinal vascular pathology in this disease. One of the major advancements is macular observation given by optical coherence tomography (OCT). OCT dramatically improves the diagnostic quality in macular edema in DR. The technology of OCT is also applied to angiography (OCT angiograph: OCTA), which enables retinal vascular imaging without venous dye injection. Similar to OCTA, in terms of their low invasiveness, single blue color SLO image could be an alternative method in detecting non-perfused areas. Conventional optical photography has been gradually replaced to scanning laser ophthalmoscopy (SLO), which also make it possible to produce spectacular ultra-widefield (UWF) images. Since retinal vascular changes of DR are found in the whole retina up to periphery, it would be one of the best targets in UWF imaging. Additionally, evolvement of artificial intelligence (AI) has been applied to automated diagnosis of DR, and AI-based DR management is one of the major topics in this field. This review is trying to look back on the progress of imaging of DR comprehensively from the past to the present.

Measurement of the Lid Margin Thickness in Meibomian Gland Dysfunction with Vernier Micrometer

Introduction

To investigate the lid margin thickness (LMT) from the posterior lash line to the mucocutaneous junction at the middle position in adults with and without meibomian gland dysfunction (MGD) by vernier micrometer (VM).

Methods

This is a cross-sectional, observational study. A hundred eyes from 100 volunteers aged 20 to 79, including 56 normal participants and 44 participants with MGD, were recruited. Measurements of the LMT by VM were performed by the same person.

Results

The mean age of 56 normal subjects (24 males and 32 females) and 44 MGD subjects (16 males and 28 females) was 40.0 ± 13.2 years and 42.7 ± 17.1 years, respectively. There was a significant difference in the upper LMT between normal and MGD subjects (1.36 ± 0.25 vs. 1.60 ± 0.27 mm, P < 0.001), but not in the lower LMT (1.0 ± 0.23 vs. 1.10 ± 0.28 mm, P = 0.07). In both normal and MGD subjects, the upper or lower LMT was significantly positively correlated with age (P < 0.05), and the upper LMT was greater than the lower LMT (P < 0.001). In addition, the lower LMT in MGD subjects was significantly positively correlated with meibum expressibility (r_s = 0.35, P = 0.02).

Conclusions

The LMT was closely related to age and could be an important indicator for detecting MGD. Furthermore, we found that the upper LMT was greater than the lower LMT, and the lower LMT in MGD subjects seemed to be related to meibum expressibility.