Deep learning for lung cancer diagnosis: a comparative artificial intelligence study

Abstract

Early and accurate lung cancer diagnosis (LCD) is critical, yet traditional imaging methods such as X-rays and computed tomography (CT) scans are costly, invasive, and heavily reliant on expert interpretation. This study investigates artificial intelligence (AI)-driven diagnostics by comparing deep learning models (convolutional neural network (CNN), residual network (ResNet), and visual geometry group 16 (VGG16)) with traditional machine learning algorithms (logistic regression (LR) and support vector machine (SVM)), using lung image database consortium and image database resource initiative (LIDC-IDRI) and Kaggle lung CT scan datasets. Performance was evaluated across multiple metrics: accuracy, sensitivity, specificity, and area under the curve-receiver operating characteristic (AUC-ROC). Among the models, ResNet achieved the highest performance, with an accuracy of 94%, sensitivity of 95%, specificity of 93%, and AUC-ROC close to 1. CNN and VGG16 also showed superior metrics compared to LR and SVM, highlighting the robustness of deep learning techniques. These results demonstrate that deep learning models not only achieve higher diagnostic accuracy but also significantly reduce false detections compared to traditional approaches. The findings support the potential of AI to automate and enhance LCD, thereby improving accessibility, consistency, and speed in clinical settings. Future work will emphasize clinical validation, address ethical challenges, and focus on integrating AI models into real-world healthcare workflows to improve patient outcomes.