International Research Journal of Education and Technology

The classification of lung cancer raises substantial difficulties because of a scarcity of training data, a large number of dimensions, intricate imaging features, and similarities within classes, frequently leading to less than optimal diagnostic precision. The Capsule Neural Network preserves the hierarchical links between distinct image components by converting scalar feature representations into vectors, providing a novel solution to these problems. This article presents an innovative deep-learning framework designed specifically for classifying lung cancer. It utilizes two convolutional neural networks to improve the extraction of spatial and spectral characteristics. We assessed the efficacy of the CapsNet model by evaluating it on the benchmark LIDC-IDRI datasets. The results indicate that LungCaps outperforms standard Capsule Networks and ConvCaps in terms of performance, highlighting its potential to enhance lung cancer diagnosis in CT imaging.

Keywords—Lung Cancer, Capsule Network, Computed Tomography, Image Classification, Deep Learning