Artificial Intelligence in Radiology: Deep Learning for Lung Nodule Detection

Abstract

This multicenter validation study assessed a deep learning algorithm for pulmonary nodule detection on 18,432 CT scans from 6 academic medical centers. The AI system achieved sensitivity of 94.6% and specificity of 91.3% at the nodule level, outperforming radiologist performance (sensitivity: 87.2%, specificity: 89.1%) for nodules <6mm. The algorithm reduced radiologist reading time by 31.4% while maintaining equivalent diagnostic accuracy for nodules ≥6mm.

Keywords:

artificial intelligence radiology lung nodule deep learning CT scan