Efficient Medical Image Classification Using Masked Attention Networks

Abstract

Medical image classification plays a crucial role in early disease detection, clinical decision support, and automated healthcare systems. However, traditional deep learning models often suffer from high computational complexity and limited focus on diagnostically relevant regions within medical images. This study proposes an Efficient Masked Attention Network (EMANet) for medical image classification. The model integrates masked self-attention mechanisms with lightweight convolutional feature extractors to improve efficiency while enhancing focus on clinically significant regions. The masked attention mechanism suppresses irrelevant spatial features and enhances discriminative learning in medical imaging tasks such as tumor detection, organ segmentation, and disease classification. The proposed framework is evaluated on standard medical imaging datasets, and performance is measured using accuracy, precision, recall, F1-score, and computational efficiency. Experimental results demonstrate that the proposed method achieves superior classification accuracy while significantly reducing computational cost compared to conventional CNN and transformer-based models.