Optimizing double-layered convolutional neural networks for efficient lung cancer classification through...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Lung cancer is a major global health concern with prognosis tied to early detection.
• Current diagnostic methods present limitations in accuracy, invasiveness, and scalability.
• Advanced computational tools are needed to improve lung cancer diagnosis.

Purpose of the Study:

• To develop and evaluate a machine learning model for enhanced lung cancer stage classification using CT images.
• To overcome limitations of traditional diagnostic methods with a faster, non-invasive tool.
• To improve the accuracy and reliability of lung cancer diagnosis.

Main Methods:

• Utilized the IQ-OTHNCCD lung cancer dataset for training.
• Applied preprocessing techniques: resizing, normalization, and Gaussian blurring.
• Employed a Convolutional Neural Network (CNN) and addressed class imbalance with SMOTE.

Main Results:

• Achieved a classification accuracy of 99.64% for lung cancer staging.
• Demonstrated high precision, recall, and F1-scores exceeding 98%.
• SMOTE effectively improved the classification of underrepresented classes.

Conclusions:

• Machine learning, specifically CNNs, shows significant potential for transforming lung cancer diagnostics.
• The developed model offers a highly accurate and reliable tool for early lung cancer detection.
• Improved diagnostic capabilities can lead to tailored treatment plans and better patient outcomes.