A Multi-Chest Disease Detection Using Multi-Scale Alignment Graph Capsule Binary Light Spectrum Dual Attention Neural Network
DOI:
https://doi.org/10.64252/kdd0bh58Keywords:
Chest x-rays, Coronavirus, Multi-class diseases, Multi-scale Alignment, Semantic segmentationAbstract
Internationally, the COVID-19 disease has badly affected both the healthcare system and the market. The complexity arises from the similarity in symptoms between COVID-19 and other chest diseases like pneumonia and lung cancer, making accurate diagnosis challenging. Dedicated frontline medical professionals and researchers are actively striving to develop a rapid and automated method for the initial stage recognition of COVID-19, with the goal of saving lives. Nevertheless, the clinical diagnosis of coronavirus remains subjective and variable. To tackle these issues, in this research propose a novel Multi-scale Alignment graph Capsule Binary light spectrum Dual attention neural Network (MACBDN) technique for efficiently classifying multi-chest diseases, including pneumothorax, pneumonia, tuberculosis, lung cancer and COVID-19. The methodology begins with the utilization of the adaptive guided multi-layer side window box filter for preprocessing, aiming to eliminate noise while preserving crucial details. Subsequently, a lightweight multi SegNet is employed for semantic segmentation. S-transform and fast discrete orthonormal transform are then applied for efficient feature extraction. The proposed MACBDN effectively classifies diseases using the binary light spectrum optimizer.