PULMO-NET: blockchain-integrated lung cancer classification using golden jackal optimization and GoogleNet

Abstract

Lung cancer (LC) is a malignant disease caused by uncontrolled cell growth in the lungs, often associated with smoking and environmental factors. However, accurate LC classification is particularly challenging due to poor image quality, variability in imaging conditions, and noise artifacts in medical scans. In this work, a novel PULMO-NET is proposed for classifying LC using dual-modality imaging (CXR and CT). The dual-modality images are preprocessed using an adaptive trilateral (ADT) filter and segmented using the golden jackal optimization. The segmented lung regions are refined using the dragonfly algorithm (DA) which enables accurate extraction of diamond-shaped tumor patterns. Additionally, a blockchain-based system with local nodes is integrated to collect real-time patient data. GoogleNet uses inception modules to capture multi-scale features, enabling accurate classification of lung images into normal, non-small cell lung cancer (NSCLC), and small cell lung cancers (SCLC). The proposed PULMO-NET achieves the classification accuracy (AC) of 98.91% and F1 score of 96.51%. The PULMO-NET model improves the overall AC by 1.91%, 7.78%, and 4.33% better than Inception-v3, TPOT_SVM, and LeNet–DenseNet respectively.