Starattm: Starnet Hybrid Attention Mechanism for Pathological Medical Image Analysis

Abstract:
Deep learning can predict the health status and disease risk of patients by analyzing large amounts of medical image data, which can help reduce the incidence and mortality of diseases. This article proposes a hybrid attention mechanism based on StarNet. StarNet can focus on key areas in medical images, reducing the interference of background noise. The attention module uses global average pooling and convolution operations to enhance the features of disease areas in medical images, and captures remote contextual information through a specific network structure, further improving the ability to extract key features. SCConv convolution is used for feature extraction to reduce computational complexity and improve model performance. Through testing on medical image datasets, the experimental results show that this method achieves significant results in both detection accuracy and precision.