Using Deep Learning to Segment Retinal Vascular Leakage and Occlusion in Retinal Vasculitis

Dhirachaikulpanich, Dhanach; Xie, Jianyang; Chen, Xiuju; Li, Xiaoxin; Madhusudhan, Savita; Zheng, Yalin; Beare, Nicholas A. V.

doi:10.1080/09273948.2024.2305185

Cited by 3 publications

(1 citation statement)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Some previous works have attempted automated segmentation of retinal vascular leakage in wide-field fluorescein angiography images [9][10][11]. Li et al [12] and Dhirachaikulpanich et al [13] use deep learning approaches to segment retinal leakages in fluorescein angiography images. However, only a few works so far have focused on the specific task of HF segmentations: Zhou et al [14] combine diagnostic descriptions and extracted visual features through Mutual-aware Feature Fusion to improve HF segmentations in gray fundus images.…”

Section: Introductionmentioning

confidence: 99%

When Two Eyes Don’t Suffice—Learning Difficult Hyperfluorescence Segmentations in Retinal Fundus Autofluorescence Images via Ensemble Learning

Santarossa,

Beyer,

Scharf

et al. 2024

J. Imaging

View full text Add to dashboard Cite

Hyperfluorescence (HF) and reduced autofluorescence (RA) are important biomarkers in fundus autofluorescence images (FAF) for the assessment of health of the retinal pigment epithelium (RPE), an important indicator of disease progression in geographic atrophy (GA) or central serous chorioretinopathy (CSCR). Autofluorescence images have been annotated by human raters, but distinguishing biomarkers (whether signals are increased or decreased) from the normal background proves challenging, with borders being particularly open to interpretation. Consequently, significant variations emerge among different graders, and even within the same grader during repeated annotations. Tests on in-house FAF data show that even highly skilled medical experts, despite previously discussing and settling on precise annotation guidelines, reach a pair-wise agreement measured in a Dice score of no more than 63–80% for HF segmentations and only 14–52% for RA. The data further show that the agreement of our primary annotation expert with herself is a 72% Dice score for HF and 51% for RA. Given these numbers, the task of automated HF and RA segmentation cannot simply be refined to the improvement in a segmentation score. Instead, we propose the use of a segmentation ensemble. Learning from images with a single annotation, the ensemble reaches expert-like performance with an agreement of a 64–81% Dice score for HF and 21–41% for RA with all our experts. In addition, utilizing the mean predictions of the ensemble networks and their variance, we devise ternary segmentations where FAF image areas are labeled either as confident background, confident HF, or potential HF, ensuring that predictions are reliable where they are confident (97% Precision), while detecting all instances of HF (99% Recall) annotated by all experts.

show abstract

Section: Introductionmentioning

confidence: 99%

When Two Eyes Don’t Suffice—Learning Difficult Hyperfluorescence Segmentations in Retinal Fundus Autofluorescence Images via Ensemble Learning

Santarossa,

Beyer,

Scharf

et al. 2024

J. Imaging

View full text Add to dashboard Cite

show abstract

Managing a patient with uveitis in the era of artificial intelligence: Current approaches, emerging trends, and future perspectives

Rojas-Carabali,

Cifuentes-González,

Gutierrez-Sinisterra

et al. 2024

Asia-Pacific Journal of Ophthalmology

View full text Add to dashboard Cite

Applications of Artificial Intelligence in Vasculitides: A Systematic Review

Omar,

Agbareia,

Naffaa

et al. 2024

Preprint

View full text Add to dashboard Cite

Background and Aim: Vasculitides are rare inflammatory disorders that sometimes can be difficult to diagnose due to their diverse presentations. This review examines the use of Artificial Intelligence (AI) to improve diagnosis and outcome prediction in vasculitis. Methods: A systematic search of PubMed, Embase, Web of Science, IEEE Xplore, and Scopus identified relevant studies from 2000 to 2024. AI applications were categorized by data type (clinical, imaging, textual) and by task (diagnosis or prediction). Studies were assessed for risk of bias using PROBAST and QUADAS-2 tools. Results: Forty-six studies were included. AI models achieved high diagnostic performance in Kawasaki Disease, with sensitivities up to 92.5% and specificities up to 97.3%. Predictive models for complications, such as IVIG resistance in Kawasaki Disease, showed AUCs between 0.716 and 0.834. Other vasculitis types, especially those using imaging data, were less studied and often limited by small datasets. Conclusion: The current literature shows that AI algorithms can enhance vasculitis diagnosis and prediction, with deep and machine learning models showing promise in Kawasaki Disease. However, broader datasets, more external validation, and the integration of newer models like LLMs are needed to advance their clinical applicability across different vasculitis types. Keywords: Vasculitis, Artificial Intelligence, Machine Learning, Deep Learning, Natural Language Processing.

show abstract

Using Deep Learning to Segment Retinal Vascular Leakage and Occlusion in Retinal Vasculitis

Cited by 3 publications

References 31 publications

When Two Eyes Don’t Suffice—Learning Difficult Hyperfluorescence Segmentations in Retinal Fundus Autofluorescence Images via Ensemble Learning

When Two Eyes Don’t Suffice—Learning Difficult Hyperfluorescence Segmentations in Retinal Fundus Autofluorescence Images via Ensemble Learning

Managing a patient with uveitis in the era of artificial intelligence: Current approaches, emerging trends, and future perspectives

Applications of Artificial Intelligence in Vasculitides: A Systematic Review

Contact Info

Product

Resources

About