AI-accelerated 3D gradient echo versus ultrashort echo time MRI for lung nodule detection and measurement

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Ultrashort echo time (UTE) sequences offer good lung parenchyma signal in MRI but face limitations in scan time and reconstruction complexity.
• Pulmonary MRI is crucial for lung nodule detection and characterization.

Purpose of the Study:

• To compare an AI-enhanced 3D gradient echo sequence (CSAI-GRE) with a UTE sequence for lung nodule detection and Lung-RADS classification.
• To evaluate the diagnostic performance of these MRI sequences against CT as the reference standard.

Main Methods:

• A single-center observational study included 54 patients with CT-detected lung nodules who underwent pulmonary MRI using CSAI-GRE and UTE sequences.
• Three radiologists assessed image quality, nodule detection, and Lung-RADS classification, with CT serving as the reference standard.
• Statistical analyses included paired tests, mixed-effects modeling for detection, and kappa statistics for Lung-RADS agreement.

Main Results:

• CSAI-GRE demonstrated superior subjective image quality (4.30 vs 3.91) and higher lesion-level nodule detection (96.9% vs 92.8%) compared to UTE.
• CSAI-GRE showed higher conditional detection odds (OR, 10.2; P=.002).
• MRI-CT agreement for Lung-RADS categorization was substantial to almost perfect (κ range, 0.79-0.94). CSAI-GRE had lower Lung-RADS reclassification rates (12.3%) than UTE (20.8%) compared to CT.

Conclusions:

• Both CSAI-GRE and UTE sequences enable reliable pulmonary nodule detection, with CSAI-GRE offering better image quality and detection performance.
• Despite promising results, MRI-based Lung-RADS categorization requires further validation before routine clinical application due to reclassification rates.