DRSegNet: A cutting-edge approach to Diabetic Retinopathy segmentation and classification using parameter-aware Nature-Inspired optimization DOI Creative Commons

Sundreen Asad Kamal,

Youtian Du,

Majdi Khalid

et al.

PLoS ONE, Journal Year: 2024, Volume and Issue: 19(12), P. e0312016 - e0312016

Published: Dec. 5, 2024

Diabetic retinopathy (DR) is a prominent reason of blindness globally, which diagnostically challenging disease owing to the intricate process its development and human eye’s complexity, consists nearly forty connected components like retina, iris, optic nerve, so on. This study proposes novel approach identification DR employing methods such as synthetic data generation, K- Means Clustering-Based Binary Grey Wolf Optimizer (KCBGWO), Fully Convolutional Encoder-Decoder Networks (FCEDN). achieved using Generative Adversarial (GANs) generate high-quality transfer learning for accurate feature extraction classification, integrating these with Extreme Learning Machines (ELM). The substantial evaluation plan we have provided on IDRiD dataset gives exceptional outcomes, where our proposed model 99.87% accuracy 99.33% sensitivity, while specificity 99. 78%. why outcomes presented can be viewed promising in terms further diagnosis, well creating new reference point within framework medical image analysis providing more effective timely treatments.

Language: Английский

A novel CatractNetDetect deep learning model for effective cataract classification through data fusion of fundus images DOI Creative Commons
Walaa N. Ismail, Hessah A. Alsalamah

Discover Artificial Intelligence, Journal Year: 2024, Volume and Issue: 4(1)

Published: Aug. 13, 2024

Cataracts are common eye disorders characterized by the clouding of lens, preventing light from passing through and impairing vision. Various factors, including changes in lens's hydration or alterations its proteins, may contribute to their development. Regular examinations conducted an ophthalmologist optometrist imperative for detecting cataracts other ocular conditions early on. Manual checks caregivers pose several problems, subjectivity, human error, a lack expertise. Biomedical fusion involves combining linking various characteristics specific certain diseases different medical imaging resources. The primary objectives this approach disease classification reduce error rate increase number retrieved features. aim study is evaluate outcomes associated with fusing visual features related left right cataract characteristics. Additionally, we investigate impact limited variability deep learning models, specifically fundus versus normal images. To address issue, introduces CataractNetDetect, innovative multi-label system that fuses feature representations pairs images (e.g., eyes) automatic diagnosis disorders. Our focus on achieving improved performance stacking discriminative combine two into unified representation. Several architectures utilized as descriptors, ResNet-50, DenseNet-121, Inception-V3, enhancing resilience quality representations. Fine-tuning these DL using ImageNet dataset, followed integrated Inception-V3 models. model trained publicly available ODIR-5k which includes 5000 left/right depicting eight conditions, ranging healthy states uncommon ailments such cataracts, glaucoma, age-related macular degeneration (AMD), diabetes, hypertension, myopia abnormalities. Moreover, extensive preprocessing performed, data augmentation, noise reduction, contrast enhancement, scaling, circular border cropping. CataractNetDetect demonstrates F1-scores, AUC, maximum validation scores 98.0%, 97.9%, 100%, respectively. This ensemble-based distinguishes itself surpassing conventional established methodologies, thereby underscoring efficacy diagnostic applications.

Language: Английский

Citations

3
An automatic glaucoma grading method based on attention mechanism and EfficientNet-B3 network DOI Creative Commons
Xu Zhang,

Fuji Lai,

Weisi Chen

et al.

PLoS ONE, Journal Year: 2024, Volume and Issue: 19(8), P. e0296229 - e0296229

Published: Aug. 16, 2024

Glaucoma infection is rapidly spreading globally and the number of glaucoma patients expected to exceed 110 million by 2040. Early identification detection particularly important as it can easily lead irreversible vision damage or even blindness if not treated with intervention in early stages. Deep learning has attracted much attention field computer been widely studied especially recognition diagnosis ophthalmic diseases. It challenging efficiently extract effective features for accurate grading a limited dataset. Currently, algorithms, 2D fundus images are mainly used automatically identify disease not, but do distinguish between late stages; however, clinical practice, treatment same, so more proceed achieve glaucoma. This study uses private dataset containing modal data, images, 3D-OCT scanner therein an triple classification (normal, early, moderately advanced) optimal performance on various measures. In view this, this paper proposes automatic method based mechanism EfficientNetB3 network. The network ResNet34 built fuse respectively, classification. proposed auto-classification minimizes feature redundancy while improving accuracy, incorporates two-branch model, which enables convolutional neural focus its main eye discard meaningless black background region image improve model. combined cross-entropy function achieves highest accuracy up 97.83%. Since ensures reliable decision-making screening, be second opinion tool doctors, greatly reduce missed misdiagnosis buy time patient’s treatment.

Language: Английский

Citations

2
Retinal blood vessel segmentation using density-based fuzzy C-means clustering and vessel neighborhood connected component DOI
Kittipol Wisaeng

Measurement, Journal Year: 2024, Volume and Issue: 242, P. 116229 - 116229

Published: Nov. 14, 2024

Language: Английский

Citations

1
WITHDRAWN: LSAC-Net: A lightweight scale-aware CNN with densely connected focal modulation for retinal blood vessel segmentation DOI Creative Commons
Mufassir Matloob Abbasi, Imran Shafi, Jamil Ahmad

et al.

Heliyon, Journal Year: 2024, Volume and Issue: unknown, P. e33515 - e33515

Published: July 1, 2024

Language: Английский

Citations

0
DRSegNet: A cutting-edge approach to Diabetic Retinopathy segmentation and classification using parameter-aware Nature-Inspired optimization DOI Creative Commons

Sundreen Asad Kamal,

Youtian Du,

Majdi Khalid

et al.

PLoS ONE, Journal Year: 2024, Volume and Issue: 19(12), P. e0312016 - e0312016

Published: Dec. 5, 2024

Diabetic retinopathy (DR) is a prominent reason of blindness globally, which diagnostically challenging disease owing to the intricate process its development and human eye’s complexity, consists nearly forty connected components like retina, iris, optic nerve, so on. This study proposes novel approach identification DR employing methods such as synthetic data generation, K- Means Clustering-Based Binary Grey Wolf Optimizer (KCBGWO), Fully Convolutional Encoder-Decoder Networks (FCEDN). achieved using Generative Adversarial (GANs) generate high-quality transfer learning for accurate feature extraction classification, integrating these with Extreme Learning Machines (ELM). The substantial evaluation plan we have provided on IDRiD dataset gives exceptional outcomes, where our proposed model 99.87% accuracy 99.33% sensitivity, while specificity 99. 78%. why outcomes presented can be viewed promising in terms further diagnosis, well creating new reference point within framework medical image analysis providing more effective timely treatments.

Language: Английский

Citations

0