Introducing Novel Radon Based Transform for Disease Detection From Chest X-Ray Images

Medical imaging technologies, such as chest X-rays (CXR), have demonstrated their utility in predicting diseases with high accuracy using deep learning algorithms. These models are crucial for identifying critical lung conditions. Nevertheless, the challenge lies in the resemblance of disease patterns and symptoms, which may cause misdiagnoses and critical mistakes. In our research, we introduce a novel technique for feature extraction from CXR images using an advanced version of the Radon transform, named the RadEx Transform. This method, by integrating the extracted features with CXR images, significantly enhances the learning capability of the models. We focus our study on the COVID-19 radiography dataset. The results indicate that our approach of feature extraction markedly increases accuracy beyond that achieved with raw images alone, surpassing conventional techniques by significant margins in terms of x, y, and z. Our research underscores the effectiveness of augmenting RadEx features with images in elevating the accuracy of lung disease detection. This approach holds considerable promise for advancing medical image analysis and diagnostic processes, marking a significant step forward in the domain.