MR-assisted PET respiratory motion correction using deep-learning based short-scan motion fields

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Radiology

Background:

• Respiratory motion during Positron Emission Tomography (PET) scans can degrade image quality and lesion detectability.
• Motion correction (MoCo) techniques aim to mitigate these artifacts.
• Integrating magnetic resonance imaging (MRI) data for PET MoCo offers potential for improved accuracy.

Purpose of the Study:

• To evaluate the impact of PET respiratory motion correction (MoCo) using MRI-derived motion vector fields (MVFs).
• To assess a novel PET MoCo approach utilizing MVFs from a deep-learning reconstructed short MRI scan.
• To compare the performance of different MRI acquisition times for PET MoCo.

Main Methods:

• PET MoCo was evaluated in a respiratory motion phantom with varying lesion sizes and tumor-to-background ratios (TBRs).
• MRI-based MVFs were derived from either 2000 spokes (5-6 min) or 200 spokes (30-40 s) using deep-learning reconstruction.
• Performance was assessed using quantitative metrics (SUVmax, SUVpeak, SUVmean, lesion volume) and blinded radiological review in six patients with hepatic lesions.

Main Results:

• MRI-assisted PET MoCo methods yielded results comparable to static scans in the phantom study.
• Both MoCo approaches significantly improved SUVmax, SUVpeak, SUVmean, and reduced lesion volume compared to non-MoCo PET.
• No significant statistical difference was observed between the two MoCo methods in quantitative metrics.
• Radiological reviewers consistently found improved lesion conspicuity with both MoCo methods.

Conclusions:

• MRI-assisted PET MoCo is effective, validated by phantom studies and patient data.
• A PET MoCo method using only 30-40 seconds of MRI motion modeling data significantly enhances quantitative and qualitative metrics in patients with hepatic lesions.