Artificial intelligence in automated detection of lung nodules: a narrative review

Abstract

Lung cancer remains a leading cause of cancer-related mortality worldwide, and early detection is essential for improving patient outcomes. This study evaluates the role of artificial intelligence (AI) in lung nodule detection, focusing on its potential to enhance the accuracy of early lung cancer diagnosis. We assess the performance of AI tools, particularly convolutional neural networks (CNNs), in identifying and segmenting lung nodules from computed tomography (CT) and X-ray images. Our findings indicate that AI systems achieve a sensitivity of 70-90, comparable to that of experienced radiologists, while reducing false-positive rates. In pulmonary nodule detection on CT scans, AI demonstrated over 95 sensitivity with fewer than one false-positive per scan. The implementation of AI as a "second reader" significantly improved detection accuracy. Despite these advancements, challenges remain, including high false-positive rates, issues with generalizability across diverse populations, regulatory concerns, and skepticism among healthcare professionals. This study highlights the promise of AI in supporting radiologists and improving lung cancer screening while emphasizing the need for further research to enhance specificity and address existing limitations.