Conjugate gradient and deep learning reconstructions: reduced time without affecting image quality and nodule...

Area of Science:

• Magnetic Resonance Imaging
• Medical Imaging Technology
• Radiology

Background:

• Lung MRI with ultrashort echo time (UTE-MRI) is crucial for nodule detection.
• Reducing scan time is essential for improving patient comfort and throughput.
• Traditional grid reconstruction (Grid Recon) may limit scan time reduction.

Purpose of the Study:

• To evaluate the effectiveness of conjugate gradient reconstruction (CG Recon) and deep learning reconstruction (DLR) in reducing UTE-MRI scan time.
• To assess if CG Recon and DLR maintain image quality and nodule detection capabilities compared to Grid Recon.
• To determine the utility of these advanced reconstruction methods for lung UTE-MRI.

Main Methods:

• UTE-MRI scans were performed on a NEMA phantom and 35 patients with pulmonary nodules.
• Scans utilized reduced spoke numbers (UTE1/2, UTE1/4) reconstructed with CG Recon and DLR.
• Image quality, signal-to-noise ratios (SNR), and nodule detection were quantitatively and qualitatively assessed.
• Receiver operating characteristic (ROC) analysis compared standard protocol (Grid Recon, UTEoriginal) with alternative protocols.

Main Results:

• All CG Recon and DLR protocols showed significant differences in SNR compared to the standard protocol (p < 0.05).
• Overall image quality differed significantly between the standard protocol and UTE1/4 protocols (p < 0.05).
• CG Recon with UTEoriginal achieved a significantly larger area under the curve in ROC analysis than the standard protocol (p < 0.05).

Conclusions:

• CG Recon and DLR effectively reduce scan time in lung UTE-MRI.
• These reconstruction techniques preserve critical image quality and nodule detection capabilities.
• CG Recon and DLR show significant potential for clinical application in lung MR imaging.