Lung Cancer Detection Using Convolutional Neural Networks

Abstract

Lung cancer is among the deadliest malignancies worldwide and remains a significant global health concern due to its high mortality rate and late-stage diagnosis. Despite advances in diagnostic imaging and treatment strategies, early and accurate identification of lung cancer is still a challenging clinical task. Computed Tomography (CT) imaging is the most common diagnostic tool for detecting pulmonary nodules. However, manual interpretation by radiologists is often time-intensive and subject to human bias and fatigue. Inter-observer variability is also a concern. The complexity of lung anatomy, subtle differences between benign and malignant nodules, and the large volume of CT data all increase the need for automated, reliable, and scalable computer-aided diagnosis (CAD) systems.

Introduction

Lung cancer is a leading cause of cancer-related deaths worldwide, responsible for approximately 18% of global cancer fatalities. Its high mortality rate is largely due to late diagnosis, as early-stage symptoms are often undetectable. Computed Tomography (CT) scans are an effective tool for identifying lung nodules, which may indicate cancer, but manual interpretation is time-consuming, mentally exhausting, and prone to human error. Small or subtle nodules are easily overlooked, making early and accurate diagnosis challenging.

To address these limitations, the use of Artificial Intelligence (AI), particularly deep learning techniques such as Convolutional Neural Networks (CNNs), has gained prominence. CNNs excel at automatically extracting hierarchical spatial features from CT images, allowing them to distinguish between benign and malignant nodules without manual feature engineering. This capability makes CNNs ideal for developing automated computer-aided diagnostic (CAD) systems for lung cancer detection.

Literature Survey highlights several recent studies:

• 3D attention-gated CNNs improve feature extraction from complex CT images, especially in cases with fibrosis, but require large datasets and high computational power.

• Hybrid models combining CNNs, radiomics, and clinical data enhance diagnostic accuracy and interpretability but increase model complexity.

• CNNs for ultra-low-dose CT scans facilitate faster and safer detection, especially in emergency settings, though image quality may limit performance.

• Ensemble CNNs and radiomics approaches improve robustness and accuracy across diverse datasets but demand extensive preprocessing and computational resources.

Problem Definition: Manual CT analysis is prone to errors, fatigue, and delays, making early lung cancer detection difficult. Traditional image processing methods rely on handcrafted features and struggle with complex patterns. There is an urgent need for automated, efficient, and accurate systems to assist radiologists.

Methodology: The study proposes a CNN-based framework for lung cancer detection. Key steps include:

1. Data Collection: Gathering annotated CT scan datasets with labeled benign and malignant nodules.

2. Preprocessing: Standardizing image formats, resizing, normalizing pixel intensities, lung segmentation, and data augmentation.

3. Model Design: CNN architecture with convolutional, pooling, and fully connected layers, using ReLU activations and binary cross-entropy loss optimized with Adam.

4. Training and Validation: Dataset split into training, validation, and test sets; model learns to classify nodules automatically.

5. Deployment: Integration into a CAD system for assisting radiologists in early diagnosis.

System Architecture consists of six layers:

1. Data Acquisition: Collects CT scans from datasets like LIDC-IDRI or LUNA16.

2. Preprocessing: Enhances image quality, resizes, segments lungs, and applies augmentation.

3. Feature Extraction (CNN Module): Automatically learns spatial features from images.

4. Classification Layer: Performs binary classification of nodules as benign or malignant.

5. Output and Visualization: Displays predictions with performance metrics (accuracy, precision, recall, F1-score, ROC curve).

Conclusion

The proposed lung cancer detection system, developed using a customized Convolutional Neural Network and deployed through an interactive web interface, demonstrates the capability of deep learning to transform medical image analysis into an efficient, reliable, and clinically supportive process. By integrating automated preprocessing, hierarchical feature extraction, and probabilistic classification, the system provides a consistent and objective method for interpreting CT scan images. The real-time prediction functionality, combined with both single-image and batch-processing modes, ensures that the application can be used effectively in diagnostic workflows that require rapid evaluation of multiple patient scans. The generated prediction scores, confidence indicators, and structured results files enable clinicians to review outcomes with transparency and integrate the model’s insights into broader clinical judgement. Beyond its diagnostic accuracy, the system highlights the practical value of AI-driven tools in reducing manual workload and offering standardized interpretations across diverse CT images. The inclusion of downloadable batch results, intuitive visualization, and model-driven probability scoring further enhances usability, making the solution suitable for research, telemedicine, and preliminary screening applications. While the current model operates on 2D CT slices, the framework establishes a strong foundation for future expansion toward 3D volumetric analysis, multi-modal datasets, and advanced architectures such as transformers and hybrid CNN models.

References

[1] D. Shen, G. Wu, and H. Suk, “Deep learning in medical image analysis,” Annual Review of Biomedical Engineering, vol. 19, no. 1, pp. 221–248, 2023. [2] A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet classification with deep convolutional neural networks,” Communications of the ACM, vol. 60, no. 6, pp. 84–90, 2017. [3] A. Setio, F. Ciompi, G. Litjens, P. Gerke, and C. Jacobs, “Pulmonary nodule detection in CT images: False positive reduction using multi-view convolutional networks,” IEEE Transactions on Medical Imaging, vol. 35, no. 5, pp. 1160–1169, 2021. [4] H. Wang, J. Zhao, and L. Xu, “Residual attention convolutional networks for lung cancer classification using CT images,” Medical Image Analysis, vol. 78, pp. 102–114, 2021. [5] M. Dou, H. Chen, and P.-A. Heng, “3D convolutional neural networks for automatic pulmonary nodule detection from volumetric CT scans,” IEEE Transactions on Medical Imaging, vol. 37, no. 5, pp. 1478–1490, 2022. [6] G. Geethu and A. Nagaraj, “Lung cancer detection and classification using optimized CNN features and Squeeze–Inception–ResNeXt model,” Biomedical Signal Processing and Control, vol. 93, pp. 105–131, 2025. [7] A. Mohammed, R. Khalid, and M. Hossain, “Transfer learning for lung cancer detection using deep convolutional networks,” Computers in Biology and Medicine, vol. 165, pp. 107–118, 2024. [8] L. Zhou, Y. Qian, and T. Zhang, “Explainable deep learning for medical imaging: A survey of methods and applications,” IEEE Access, vol. 12, pp. 3345–3362, 2023. [9] J. Alaa, K. Nassar, and F. Al-Khafaji, “Hybrid CNN–RNN architecture for early lung cancer prediction from CT scans,” Pattern Recognition Letters, vol. 167, pp. 140–149, 2023. [10] Y. Lecun, Y. Bengio, and G. Hinton, “Deep learning,” Nature, vol. 521, no. 7553, pp. 436–444, 2015. [11] M. Anthimopoulos, S. Christodoulidis, L. Ebner, A. Christe, and S. Mougiakakou, “Lung pattern classification for interstitial lung diseases using a deep convolutional neural network,” IEEE Transactions on Medical Imaging, vol. 35, no. 5, pp. 1207–1216, 2020. [12] K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 770–778, 2016. [13] G. Litjens, T. Kooi, B. Bejnordi, and A. A. Setio, “A survey on deep learning in medical image analysis,” Medical Image Analysis, vol. 42, pp. 60–88, 2022. [14] C. Lundervold and A. Lundervold, “An overview of deep learning in medical imaging focusing on MRI,” Zeitschrift für Medizinische Physik, vol. 29, no. 2, pp. 102–127, 2023. [15] J. Chen, H. Lu, and J. Zhang, “A transformer-based approach for lung cancer prediction using electronic health records and CT imaging,” IEEE Journal of Biomedical and Health Informatics, vol. 29, no. 1, pp. 205–214, 2024. [16] M. Tan and Q. Le, “EfficientNet: Rethinking model scaling for convolutional neural networks,” Proceedings of the 36th International Conference on Machine Learning (ICML), pp. 6105–6114, 2019. [17] M. Abadi et al., “TensorFlow: Large-scale machine learning on heterogeneous distributed systems,” arXiv preprint arXiv:1603.04467, 2016. [18] The Cancer Imaging Archive (TCIA), “The Lung Image Database Consortium (LIDC-IDRI),” Public Dataset Repository, 2023. [Online]. Available: https://www.cancerimagingarchive.net [19] N. Srivastava, G. Hinton, A. Krizhevsky, and R. Salakhutdinov, “Dropout: A simple way to prevent neural networks from overfitting,” Journal of Machine Learning Research, vol. 15, no. 1, pp. 1929–1958, 2014. [20] World Health Organization (WHO), “Global Cancer Observatory: Lung cancer fact sheet,” International Agency for Research on Cancer (IARC), 2025. [Online]. Available: https://gco.iarc.fr

Copyright

Copyright © 2025 Dr. Vidhya K, Monisha K, Poojashree A, Prajwal K T, Thanusha K M. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.