AI-Enhanced 3D Gradient Echo MRI: A Radiation-Free Alternative to CT for Lung Nodule Detection and Lung-RADS Classification

This study aims to evaluate the capabilities of a 3D gradient echo MRI sequence for the detection and classification of pulmonary nodules, specifically in relation to the lung CT screening reporting and data system (Lung-RADS). In a prospective trial, 75 patients (mean age 65±12years; 44% women) with benign and malignant lung nodules (March 2022-July 2024) underwent chest CT and 3D gradient echo MRI using parallel imaging, compressed sensing, and AI acceleration (CSAI). Three radiologists (experience: 4, 9, and 10 years) assessed detection rates, nodule size, morphology, and Lung-RADS classification in a blinded study. Intra- and inter-rater reliability were evaluated using intraclass correlation coefficient (ICC) for size and Cohen's/Fleiss' kappa for Lung-RADS and morphology. The CSAI gradient echo sequence achieved a detection rate of 96.3%, missing 5 out of 135 nodules across all readers. Nominal scan time was 3:53 min. MRI nodule diameter deviated from CT by 0.1 mm (1.96 SD: ±5.87mm). Excellent inter-rater agreement was observed for nodule size and morphology (Size: ICC-CT: 0.995; ICC-CSAI: 0.993; Morphology: Fleiss' kappa CT: 0.95; Fleiss' kappa MR: 0.88). Intra-rater morphology agreement between CT and MRI ranged from moderate to substantial (Cohen's-kappa: 0.55-0.71; p<.001). Lung-RADS agreement between CT and MRI was almost perfect for two readers (k=0.86, 0.90), and substantial for one (k=0.69), with near perfect inter-rater agreement (Fleiss' kappa CT:0.94; MR:0.87). Accelerated 3D gradient echo MRI showed high detection rates for pulmonary nodules with Lung-RADs scores and morphological assessments comparable to CT.