Attention U-Net Model for Accurate Thoracic Disease Detection, Segmentation and Localization

Lung cancer, pneumonia, and other chest disorders including TB continue to be major global public health issues. For prompt and successful treatment, it is essential to identify these disorders early and accurately in medical imaging, particularly in chest X-rays. Deep learning approaches have produced encouraging outcomes in medical image processing applications in recent years. To enhance chest disease identification and segmentation, this research study presents a novel method that combines the strength of the U-Net layout with attention processes. The Attention U-Net model takes advantage of both feature extraction and attention processes to draw attention to key areas in the pictures, improving task performance for segmenting and identifying chest diseases. The suggested Attention U-Net model's design, implementation details, evaluation criteria, and experimental findings are presented in this work, along with the localization of the boundary regions of the chest disease. It has good generalizability to CXR images from unknown datasets with various patients.