<p>The classification of lung cancer is often hindered by several challenges, including limited training data, high dimensionality, image complexity, and similarity between classes. These challenges frequently lead to inadequate classification performance. The utilisation of capsule networks enables the retention of the hierarchical structure among distinct components of an object depicted in an image through the substitution of scalar illustrations with vectors, thereby offering a potential solution to the aforementioned challenges. Inspired by the Capsule Network, this article introduces the Lung Capsule Network (LungCaps), a novel end-to-end deep learning architecture designed for cancer classification. The LungCaps system utilises a pair of convolutional neural networks that are encapsulated to extract spatial and spectral features at a higher level. The LungCaps model was subjected to rigorous testing on benchmark LIDC-IDRI datasets to evaluate its performance. The results of the evaluation indicate that LungCaps outperformed both the basic Capsule Net and ConvCaps Network. Experimental results on the LIDC-IDRI dataset demonstrate that the proposed LungCaps model achieves an accuracy of 95.85% and a precision of 99.75%, outperforming conventional Capsule Network and ConvCaps models.</p>