Enhancing Precision in Lung Cancer Detection through CapsuleNet-ResNet Fusion Model

Abstract

Lung cancer is a big problem in global health, hence there has to be improvement in methods for detecting it early. In this study, we introduce a novel approach to enhancing the precision of lung cancer detection by merging ResNet and CapsuleNet into a fusion model. The strength of CapsuleNet in capturing hierarchical properties and the experience of ResNet in tackling vanishing gradient challenges are combined to create a more robust solution for lung cancer diagnosis. The suggested CapsuleNet-ResNet fusion model makes use of CapsuleNet's distinctive capsule structure to effectively describe spatial hierarchies within lung imagery. Dynamic routing can capture complex patterns more efficiently by using capsules instead of regular neurons. Using ResNet's residual learning to address issues caused by deep neural networks, we further enhance the model's feature extraction. We train ResNet and CapsuleNet independently after pre-processing the lung image collection. Afterwards, the learned representations from both networks are combined using a well-planned fusion procedure. More discriminative detection of lung cancer is achieved by the model through the combination of local and global data. We put our suggested method through its paces using benchmark lung cancer datasets in a battery of tests. We test the suggested CapsuleNet-ResNet fusion model against state-of-the-art methods, CapsuleNet and ResNet models on their own, and more. Our CapsuleNet-ResNet fusion model revealed significant results with 98% accuracy, 97.2% precision, 98.5% recall rate and 97.8% F1 Score. These results surpass those of fundamental algorithms such as VGG16, CNN, and ResNet. It reveal that our fusion model has the potential to be used for early detection of lung cancer because of its improved detection accuracy and resilience. By accurately depicting lung images and making use of the unique properties of ResNet and CapsuleNet, our proposed method enhances diagnostic abilities of lung cancer.