Multi-Modal Lung Cancer Detection Using Pyramidal Cascade Neuro-Fuzzy Fractional Network.

Abstract

Lung cancer detection (LCD) is a process of identifying an occurrence of lung cancer (LC) or irregularities in the lungs. Early detection of lung cancer is crucial for improving patient survival and enabling effective treatment. Computed Tomography (CT) images and Positron emission tomography (PET) are employed for screening and detecting LC. These methods offer full cross-sectional images of the lungs to detect smaller lesions. Several techniques are developed for LCD, but they often fall into uncertainty. Therefore, a Pyramidal Cascade Neuro-Fuzzy Fractional Network (PCNFFN) is introduced for LCD utilizing CT and PET images. Initially, PET and CT images are pre-processed employing a Bilateral filter (BF). Then, lung lobes are segmented from both images utilizing Dual-Attention V-Network (DAV-Net). Thereafter, Black Hole Entropic Fuzzy Clustering (BHEFC) is employed to segment tumor locations from both lung lobe segmented images. Next, features are extracted from tumor location segmented images. Lastly, LCD is performed by PCNFFN. However, PCNFFN is a combination of Deep Pyramidal residual Network (PyramidNet) and Cascade Neuro-Fuzzy Network (NFN) with Fractional Calculus (FC). In addition, PCNFFN achieved an accuracy of about 91.002%, a true negative rate (TNR) of about 90.504% and a true positive rate (TPR) of about 92.571%.