The Application Value of Artificial Intelligence in Pulmonary Nodule Detection and Benign-Malignant Classification

Abstract

Artificial intelligence demonstrates significant value in detecting pulmonary nodules and determining whether they are benign or malignant. With the ever-growing applications of low-dose CT screening, the detection speed of pulmonary nodules has increased substantially. However, manual diagnosis faces challenges including missed detections and inconsistent interpretations among different radiologists. This article provides a systematic review of artificial intelligence (AI) applications in lung cancer screening. It focuses on the automated identification, segmentation, and diagnostic classification of pulmonary nodules. We specifically examine the performance of deep learning models like CNN and U-Net, comparing results from multiple clinical studies. Research findings indicate that AI achieves high sensitivity of up to 98.98% in nodule detection and reaches AUC values between 0.88 and 0.936 in benign-malignant classification. These results are comparable to, and in some cases exceed, the performance of experienced radiologist teams. Additionally, AI shows potential in predicting pathological subtypes and improves diagnostic accuracy through multi-modal integration. By providing quantitative feature analysis and model integration, AI technology offers more objective and consistent diagnostic support. Its high sensitivity and substantial value as a diagnostic aid suggest that AI will likely become an essential component of standard clinical workflows in the near future.