Non-Invasive to Invasive: Enhancing FFA Synthesis from CFP with a Benchmark Dataset and a Novel Network DOI Open Access
Hongqiu Wang, Zhaohu Xing,

Weitong Wu

и другие.

Опубликована: Окт. 28, 2024

Fundus imaging is a pivotal tool in ophthalmology, and different modalities are characterized by their specific advantages. For example, Fluorescein Angiography (FFA) uniquely provides detailed insights into retinal vascular dynamics pathology, surpassing Color Photographs (CFP) detecting microvascular abnormalities perfusion status. However, the conventional invasive FFA involves discomfort risks due to fluorescein dye injection, it meaningful but challenging synthesize images from non-invasive CFP. Previous studies primarily focused on synthesis single disease category. In this work, we explore multiple diseases devising Diffusion-guided generative adversarial network, which introduces an adaptive dynamic diffusion forward process discriminator adds category-aware representation enhancer. Moreover, facilitate research, collect first multi-disease CFP paired dataset, named Multi-disease Paired Ocular Synthesis (MPOS) with four fundus diseases. Experimental results show that our network can generate better compared state-of-the-art methods. Furthermore, introduce paired-modal diagnostic validate effectiveness of synthetic diagnosis diseases, synthesized real have higher accuracy than synthesizing Our research bridges gap between FFA, thereby offering promising prospects enhance ophthalmic patient care, focus reducing harm patients through procedures. dataset code will be released support further field (https://github.com/whq-xxh/FFA-Synthesis).

Язык: Английский

Advancing UWF-SLO Vessel Segmentation with Source-Free Active Domain Adaptation and a Novel Multi-center Dataset DOI
Hongqiu Wang, Xiangde Luo, Chen Wu

и другие.

Lecture notes in computer science, Год журнала: 2024, Номер unknown, С. 75 - 85

Опубликована: Янв. 1, 2024

Язык: Английский

Процитировано

8
Evaluation of Convolutional Neural Networks (CNNs) in Identifying Retinal Conditions Through Classification of Optical Coherence Tomography (OCT) Images DOI Open Access

Rohin R. Teegavarapu,

Harshal A. Sanghvi,

Triya Belani

и другие.

Cureus, Год журнала: 2025, Номер unknown

Опубликована: Янв. 7, 2025

Introduction Diabetic retinopathy (DR) is a leading cause of blindness globally, emphasizing the urgent need for efficient diagnostic tools. Machine learning, particularly convolutional neural networks (CNNs), has shown promise in automating diagnosis retinal conditions with high accuracy. This study evaluates two CNN models, VGG16 and InceptionV3, classifying optical coherence tomography (OCT) images into four categories: normal, choroidal neovascularization, diabetic macular edema (DME), drusen. Methods Using 83,000 OCT across categories, CNNs were trained tested via Python-based libraries, including TensorFlow Keras. Metrics such as accuracy, sensitivity, specificity analyzed confusion matrices performance graphs. Comparisons dataset sizes evaluated impact on model accuracy tools deployed JupyterLab. Results InceptionV3 achieved between 85% 95%, peaking at 94% outperforming (92%). Larger datasets improved sensitivity by 7% all highest normal drusen classifications. like positively correlated size. Conclusions The confirms CNNs' potential diagnostics, achieving classification Limitations included reliance grayscale computational intensity, which hindered finer distinctions. Future work should integrate data augmentation, patient-specific variables, lightweight architectures to optimize clinical use, reducing costs improving outcomes.

Язык: Английский

Процитировано

1
Artificial intelligence for diagnosing exudative age-related macular degeneration DOI
Chaerim Kang, Jui‐En Lo, Helen Zhang

и другие.

Cochrane library, Год журнала: 2024, Номер 2024(10)

Опубликована: Окт. 17, 2024

Язык: Английский

Процитировано

3
A multi-modal multi-branch framework for retinal vessel segmentation using ultra-widefield fundus photographs DOI Creative Commons

Qihang Xie,

Xuefei Li, Yuanyuan Li

и другие.

Frontiers in Cell and Developmental Biology, Год журнала: 2025, Номер 12

Опубликована: Янв. 8, 2025

Vessel segmentation in fundus photography has become a cornerstone technique for disease analysis. Within this field, Ultra-WideField (UWF) images offer distinct advantages, including an expansive imaging range, detailed lesion data, and minimal adverse effects. However, the high resolution low contrast inherent to UWF present significant challenges accurate using deep learning methods, thereby complicating analysis context. To address these issues, study introduces M3B-Net, novel multi-modal, multi-branch framework that leverages fluorescence angiography (FFA) improve retinal vessel images. Specifically, M3B-Net tackles accuracy caused by inherently of Additionally, we propose enhanced UWF-based network specifically designed fine vessels. The includes Selective Fusion Module (SFM), which enhances feature extraction within integrating features generated during FFA process. further high-resolution images, introduce Local Perception (LPFM) mitigate context loss cut-patch Complementing this, Attention-Guided Upsampling (AUM) performance through convolution operations guided attention mechanisms. Extensive experimental evaluations demonstrate our approach significantly outperforms existing state-of-the-art methods image segmentation.

Язык: Английский

Процитировано

0
DME-MobileNet: Fine-Tuning nnMobileNet for Diabetic Macular Edema Classification DOI

Xuanzhao Dong,

Yalin Wang, Yanxi Chen

и другие.

Lecture notes in computer science, Год журнала: 2025, Номер unknown, С. 155 - 164

Опубликована: Янв. 1, 2025

Язык: Английский

Процитировано

0
Non-Invasive to Invasive: Enhancing FFA Synthesis from CFP with a Benchmark Dataset and a Novel Network DOI Open Access
Hongqiu Wang, Zhaohu Xing,

Weitong Wu

и другие.

Опубликована: Окт. 28, 2024

Fundus imaging is a pivotal tool in ophthalmology, and different modalities are characterized by their specific advantages. For example, Fluorescein Angiography (FFA) uniquely provides detailed insights into retinal vascular dynamics pathology, surpassing Color Photographs (CFP) detecting microvascular abnormalities perfusion status. However, the conventional invasive FFA involves discomfort risks due to fluorescein dye injection, it meaningful but challenging synthesize images from non-invasive CFP. Previous studies primarily focused on synthesis single disease category. In this work, we explore multiple diseases devising Diffusion-guided generative adversarial network, which introduces an adaptive dynamic diffusion forward process discriminator adds category-aware representation enhancer. Moreover, facilitate research, collect first multi-disease CFP paired dataset, named Multi-disease Paired Ocular Synthesis (MPOS) with four fundus diseases. Experimental results show that our network can generate better compared state-of-the-art methods. Furthermore, introduce paired-modal diagnostic validate effectiveness of synthetic diagnosis diseases, synthesized real have higher accuracy than synthesizing Our research bridges gap between FFA, thereby offering promising prospects enhance ophthalmic patient care, focus reducing harm patients through procedures. dataset code will be released support further field (https://github.com/whq-xxh/FFA-Synthesis).

Язык: Английский

Процитировано

2