Investigating Simultaneity for Deep Learning-Enhanced Actual Ultra-Low-Dose Amyloid PET/MR Imaging.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neuroscience

Background:

• Diagnostic-quality amyloid PET imaging is crucial for Alzheimer's disease diagnosis.
• Deep learning models can enhance ultra-low-dose PET images using simultaneous structural MR imaging.
• The necessity of simultaneous MR imaging for this enhancement is yet to be fully explored.

Purpose of the Study:

• To investigate whether simultaneous MR imaging is required for deep learning-based enhancement of ultra-low-dose amyloid PET images.
• To determine if MR imaging-assisted ultra-low-dose PET imaging can be performed with separate PET/CT and MR imaging acquisitions.

Main Methods:

• Recruited 48 participants for PET/MR imaging studies.
• Trained U-Net-based deep networks using ultra-low-dose PET and either simultaneous or nonsimultaneous MR imaging.
• Evaluated image quality using quantitative metrics, standardized uptake value ratio correlation, and clinical reads.

Main Results:

• Enhanced PET images were qualitatively similar to standard-dose images in both simultaneous and nonsimultaneous conditions.
• Quantitative metrics showed significant improvements with no differences attributed to simultaneity.
• High standardized uptake value ratio correlation and similar clinical reads for enhanced images.

Conclusions:

• Accurate amyloid PET images can be generated using enhanced ultra-low-dose PET with either nonsimultaneous or simultaneous MR imaging.
• This finding broadens the applicability of ultra-low-dose amyloid PET imaging.
• Deep learning-based image enhancement offers a flexible approach for amyloid PET acquisition.