A Fuzzy based CNN Method for Covid 19 Classification in Noisy Chest X-ray Images

The COVID-19 pandemic has caused a global health crisis. Chest X-ray (CXR) imaging is most commonly utilized techniques for diagnosing COVID-19. However, accuracy of CXR-based diagnosis is limited by the presence of noisy labels. Noisy labels are labels that are incorrect or inaccurate. They can be caused by a variety of factors, such as human error, image artifacts, and variations in imaging protocols. In this paper, a novel method is introduced for COVID-19 CXR image classification in the presence of noisy labels. The primary objective of this research is to improve the accuracy of COVID-19 detection in X-ray images that are characterized by the presence of noise. To achieve this, proposed method incorporates direction-based fuzzy rules into our methodology, which allows to effectively identify and analyse edges and textures within the X-ray images. The labels are then subjected to undergone manifold learning and multidimensional scaling to eliminate noise. High confidence samples are further chosen as training set for label-recovered dataset in order to increase constancy and precision of a classification framework without compromising its intrinsic performance. The proposed method is evaluated on three publicly available COVID-19 CXR datasets. The results showed that our method can achieve an accuracy of 96.59% on the large dataset, 97.12% on the medium dataset, 98.79% on the small dataset. These results are comparable to the state-of-the-art methods for COVID-19 CXR image classification. This can help to improve the early diagnosis and treatment of COVID-19 patients.