Integrated deep learning paradigm for comprehensive lung cancer segmentation and classification using mask R-CNN and CNN models

Abstract

This paper presents an integrated deep learning approach that combines Mask Region-based Convolutional Neural Networks (Mask R-CNN) for lung nodule segmentation with Convolutional Neural Networks (CNN) for malignancy classification in CT scan images. A key innovation of this study is the design of a streamlined hybrid pipeline that automates both localization and diagnosis, reducing reliance on manual preprocessing. The model was trained on a curated dataset from publicly available repositories, with careful preprocessing and data augmentation techniques applied to enhance generalization. The proposed method achieved 95.6% accuracy, with a precision of 94.8%, recall of 94.3%, F1-score of 97.5%, and AUC of 94.5%, outperforming individual CNN and Mask R-CNN models. Comparative analysis with state-of-the-art methods in literature demonstrates the effectiveness of this hybrid approach. The results suggest that the model is not only accurate but also scalable for clinical implementation in automated lung cancer diagnosis. Future work will consider gathering more wide and diverse datasets, including various stages of lung cancer and other lung-related disorders, to further improve model accuracy and robustness.