We found a match
Your institution may have rights to this item. Sign in to continue.
- Title
Computer-aided diagnostic model for retinal vascular diseases using graph-based attention mechanism.
- Authors
Sivapriya, G.; Manjula Devi, R.; Keerthika, P.
- Abstract
Retinal vascular segmentation and classification plays a substantial role in predicting ocular pathologies in medical applications. This work aims in segmenting the retinal blood vessels and predicts the type of disease associated with it. The clinical features require higher segmentation performance with lesser computational time. Learning approaches based on unsupervised learning helps to reduce the inconsistency that exists in various existing approaches. Here, the learning-based automated approach intends to enhance the classification outcomes. Moreover, the existing learning approaches consider differences in marginal distributions between the source and destination. Still, the conditional distributions seem to be fixed, which needs to be noted in various databases. To handle these issues, a novel graph-based attention model with the fusion of conventional neural networks (GA-CNN) is designed to diminish the conditional and marginal distribution differences. The proposed model employs attention mechanism where the model intends to establish association among the vectors to measure the features related with retinal vessels. The contextual data are gathered from the retinal vessels to highlight the fundus image. The performance of the anticipated GA-CNN model is compared with other approaches where the analysis is done with benchmark datasets (DRIVE, STARE, and Kaggle). The classification accuracy is substantially improved compared to different methods. The model outperforms various methods and establishes a better trade-off compared to UNet, DUnet, RUnet, R2Unet, and BiLSTM. It is proven that the proposed method achieves the accuracy of 90.13% and represented with appropriate graphs and tables. The prediction accuracy is substantially higher than other methods.
- Subjects
RETINAL blood vessels; MARGINAL distributions; RETINAL diseases; VASCULAR diseases; ATTENTION
- Publication
Journal of Supercomputing, 2024, Vol 80, Issue 2, p2862
- ISSN
0920-8542
- Publication type
Article
- DOI
10.1007/s11227-023-05581-w