Automatic early glaucoma detection by extracting parapapillary atrophy and optic disc from fundus image using SVM

Kumar, Shailesh; Kumar, Basant

doi:10.1007/s11042-021-11023-7

Cited by 6 publications

(1 citation statement)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Authors experimented [19] DL for improved retinal vascular segmentation (RVS) to automatically learn complex hierarchical patterns. RVS as a boundary detection problem and use fully CNNs to build a vessel probability map.…”

Section: Related Workmentioning

confidence: 99%

WITHDRAWN: HHO-UNet-IAA: Harris Hawks Optimization based Novel UNet-Inception Attention Architecture for Glaucoma Segmentation

khan,

Banerjee,

Narula

et al. 2024

Preprint

View full text Add to dashboard Cite

An accurate diagnosis is significant in preventing blindness because ophthalmology is one of the most prevalent causes of blindness worldwide. Glaucoma is diagnosed with a scanning-laser-tomography examination of the optic nerve head. The surface topography of the optic nerve head is captured and analyzed by this technique. The manual delineation of the optic nerve head by a skilled ophthalmologist is required for the analysis that results in a diagnosis of the illness. Transfer learning based computerized glaucoma classification has achieved effective outcomes. This study presents and validates a novel heuristic-based UNet-Inception attention framework for glaucoma optic nerve segmentation and classification. The suggested technique is based on the concept of merging the benefits of two distinct networks, such as the UNet and Inception framework for automatic segmentation of Glaucoma, along with the blend of the Harris Hawks approach for selecting impact features and a hybrid loss function. Using a database of fundus images, the proposed design was evaluated and compared to several state-of-the-art methods. The integration of the UNet-Inception architecture with the Haris Hawk Optimisation technique in the field of ophthalmology image segmentation has shown potential for substantial enhancements in terms of accuracy, efficiency, and adaptability based on the performance evaluation of this study. Using their combined abilities, these components proposed an efficient, automated ophthalmic classification framework. The fitness evaluation based on Intersection over Union (IoU) indicates that the model's performance is in accordance with the segmentation task. With a median DSC of 0.9892 for the optical nerve head (ONH) segments and 0.98718 for delineating the optic nerve head layer (RNFL), the findings show that the suggested technique performs better than the other techniques in terms of segmented images.

show abstract

Section: Related Workmentioning

confidence: 99%