Using Deep Learning for Classification of Lung Nodules on Computed Tomography Images

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Lung cancer is a leading cause of cancer death, often diagnosed late.
• Computed Tomography (CT) aids early lung cancer detection, but diagnosis relies heavily on radiologist experience.
• Deep learning offers a powerful approach for medical image analysis and diagnosis.

Purpose of the Study:

• To design and evaluate deep neural networks for lung cancer calcification detection in CT images.
• To compare the performance of Convolutional Neural Networks (CNNs), Deep Neural Networks (DNNs), and Stacked Autoencoders (SAEs) for classifying lung nodules.
• To assess the efficacy of these models in distinguishing between benign and malignant lung nodules.

Main Methods:

• Three deep neural network architectures (CNN, DNN, SAE) were designed and adapted for lung nodule classification.
• The networks were trained and evaluated on the publicly available Lung Image Database Consortium (LIDC-IDRI) database.
• Performance metrics included accuracy, sensitivity, and specificity for classifying lung nodules.

Main Results:

• The Convolutional Neural Network (CNN) achieved the highest performance among the evaluated models.
• CNN demonstrated an accuracy of 84.15%, sensitivity of 83.96%, and specificity of 84.32%.
• The CNN model outperformed both DNN and SAE in the lung nodule classification task.

Conclusions:

• Deep learning, particularly CNNs, shows significant promise for improving the accuracy and efficiency of lung cancer detection from CT scans.
• CNNs offer a robust tool for assisting radiologists in differentiating benign from malignant lung nodules, potentially reducing diagnostic errors.
• Further research and validation of deep learning models can enhance early lung cancer diagnosis and patient outcomes.