A lightweight deep learning model with attention mechanisms for hypertensive retinopathy classification

Hypertensive Retinopathy is a common complication of hypertension, requiring timely diagnosis to prevent vision loss. Traditional diagnosis relies on manual examination of fundus images, which is time-consuming and subjective. To address this, we propose MA-DNet, a lightweight deep learning model combining DenseNet with channel and spatial attention mechanisms. Our approach leverages feature enhancement and data balancing techniques to improve classification accuracy. Evaluated on the OIA-ODIR dataset, MA-DNet achieves 95.8 % accuracy, outperforming existing methods. This study demonstrates the potential of attention mechanisms in enhancing HR classification for clinical decision support.