Artificial Intelligence-Assisted Compressed Sensing Technique Accelerates Magnetic Resonance Imaging Simulation for...

Area of Science:

• Medical Imaging
• Radiation Oncology
• Artificial Intelligence

Background:

• Magnetic resonance imaging (MRI) simulation is crucial for precise radiation therapy planning in head and neck cancers.
• Conventional parallel imaging (PI) techniques can lead to prolonged acquisition times, impacting patient comfort and workflow efficiency.

Purpose of the Study:

• To compare the efficacy of artificial intelligence-assisted compressed sensing (ACS) with conventional parallel imaging (PI) for MRI simulation in head and neck cancer radiation therapy.
• To evaluate differences in acquisition time, image quality, and target volume delineation accuracy between ACS and PI.

Main Methods:

• Fifty-two head and neck cancer patients underwent MRI simulation using both ACS and PI techniques on a 3.0-T system.
• Acquisition parameters, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and image quality scores were assessed.
• Tumor target volumes from fused CT-MRI images were compared using Dice similarity coefficients.

Main Results:

• ACS significantly reduced total MRI simulation acquisition time by 45.52% compared to PI (378.50s vs. 694.78s).
• Image quality metrics (SNR, CNR, lesion detection, margin sharpness, artifacts, overall quality) were comparable between ACS and PI.
• Tumor target volumes and Dice similarity coefficients for primary tumors and lymph nodes showed no significant difference between ACS and PI.

Conclusions:

• ACS offers a substantial acceleration of MRI simulation for head and neck cancer radiation therapy.
• The ACS technique achieves this speed improvement without compromising diagnostic image quality or the accuracy of tumor target delineation.