Enhancing Lung Cancer Diagnosis: A Classification Based Machine Learning approach

Abstract

Out of all types of cancer in the world, lung cancer is one of the biggest killers and is classified as either non-small cell (NSCLC) or small cell (SCLC). Making sure lung cancer is properly staged is necessary for making a good treatment plan and for understanding prognosis. Most lung cancer staging today involves seeing and reading CT images by hand which requires much time, may lead to errors and strongly depends on how skilled the radiologist is. Such limitations frequently mean that patients are diagnosed later and treated in suboptimal ways. The project here suggests an innovative framework to automatically detect and stage lung cancer. Radiomics is used in the system to measure features like intensity, shape and texture featured inside a tumor from CT scans. PCA is introduced to lighten the computational stress by selecting and maintaining the most important features. Input from the reduced features feeds a custom neural network, the TNMClassifier which is developed to categorize Tumor (T), Node (N) and Metastasis (M) stages using the TNM staging system. The design employs fully connected layers along with dropout layers, so it isn’t likely to overfit and performs well. Classifying the TNM stages of lung cancer is accurate at 98% with the proposed system which helps address the weaknesses of existing solutions. This new method may help simplify work for clinicians, reduce chances of wrong diagnoses and provide better patient results

Introduction

The lungs are cone-shaped organs located in the chest, separated by the heart. The right lung is larger, while the left lung has an indentation (cardiac impression) to accommodate the heart. Lung cancer detection typically starts with imaging tests like X-rays and CT scans, with CT scans providing detailed views of small lesions and cancer progression.

This project proposes an AI-powered system using deep learning and radiomics to automatically detect, stage, and recommend treatment for lung cancer from CT images, reducing reliance on manual radiologist input. The system employs the TNM staging method (Tumor size, Node involvement, Metastasis) through a deep learning classifier (TNMClassifier) that extracts subtle tumor features missed by human eyes.

The solution includes a user-friendly web application where clinicians can upload CT scans, receive automatic diagnosis and staging, and get personalized treatment recommendations based on cancer stage, patient history, and medical guidelines. The app features secure data storage, role-based dashboards for admins, doctors, and patients, and supports integration into hospital networks.

Data preprocessing involves resizing images, noise removal, lung region isolation, and feature extraction using radiomics. The model is trained and validated on annotated public datasets, with data augmentation to balance classes and improve performance. The final system provides real-time, accurate lung cancer detection, staging, localization, and treatment suggestions, enhancing diagnostic efficiency and patient care.

Conclusion

Finally, the project has been built to help with automated lung cancer detection and staging in easy-to-use form. Combining Python, Flask, MySQL, TensorFlow, OpenCV, Pillow, Pandas, NumPy, Seaborn and Bootstrap, the system allows for fast processing and correct identification of lung cancer stages. With the End User Dashboard, admins, doctors and patients can each manage their tasks, transfer images, analyze studies instantly and receive specific guidance tailored to their needs. This system relies on the LunCan Model which handles the pipeline using three steps and a classifier: first, the data is preprocessed; second, features are extracted using Radiomics; third, PCA reduces these features; and finally, the TNMClassifier identify types of lung cancer in the data. Thanks to this model, doctors can reliably stage cancers because it handles every type of cancer earlier in the staging process. Using CT scan images, the Cancer Predictor module identifies lung cancer regions, marks tumor areas, recognizes stage of the cancer and helps diagnose and plan treatment quickly. Furthermore, the Recommendation System helps medical staff design proper treatments with reference to the cancer stage that was found. Using machine learning and medical imaging helps connect AI and clinical oncology by making the system much more efficient and intelligent for diagnosis. The project is set to revolutionize existing lung cancer diagnosis, so more cases are found early, patients have better outcomes and healthcare processes are streamlined.

References

[1] S. Albahli, H. T. Rauf, M. Arif, M. T. Nafis, and A. Algosaibi, ‘‘Identification of thoracic diseases by exploiting deep neural networks,’’ Comput., Mater. Continua, vol. 66, no.3, pp. 3139–3149, 2021. [2] A. Naik and D. R. Edla, ‘‘Lung Cancer classification on computed tomography images using deep learning,’’ Wireless Pers. Commun., vol. 116, pp. 655–690, Jan. 2021. [3] S. Albahli, ‘‘AI-driven deep CNN approach for multi-label pathology classification using chest X-rays,’’ PeerJ Comput. Sci., vol. 7, p. e495, Apr. 2021. [4] T. Meraj et al., ‘‘Lung Cancers detection using semantic segmentation and classification with optimal features,’’ Neural Comput. Appl., vol. 33, pp. 10737–10750, 2020. [5] W. Zuo, F. Zhou, and Y. He, ‘‘An embedded multi-branch 3D convolution neural network for false positive reduction in lung Cancer detection,’’ J. Digit. Imag., vol. 33, no. 4, pp. 846–857, Aug. 2020. [6] C.-J. Lin and Y.-C. Li, ‘‘Lung Cancer classification using Taguchi-based convolutional neural networks for computer tomography images,’’ Electronics, vol. 9, no. 7, p. 1066, Jun. 2020. [7] P. Zhai, Y. Tao, H. Chen, T. Cai, and J. Li, ‘‘Multi-task learning for lung Cancer classification on chest CT,’’ IEEE Access, vol. 8, pp. 180317–180327, 2020.

Copyright

Copyright © 2025 Janakiraman S, Priyadharshini K. 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.