Clinical evaluation of k-space correlation informed motion artifact detection in segmented multi-slice MRI

Area of Science:

• Medical Imaging
• Artificial Intelligence in Radiology
• Biomedical Signal Processing

Background:

• Motion artifacts degrade diagnostic image quality in Magnetic Resonance Imaging (MRI).
• Artifacts necessitate patient recalls, increasing costs and negatively impacting patient experience.
• Real-time detection of motion artifacts during scanning is crucial for workflow optimization.

Purpose of the Study:

• To develop and validate a method for detecting motion artifacts directly from raw k-space data.
• To assess the performance of a deep learning model in predicting the severity of motion artifacts.
• To enable immediate corrective actions during MRI acquisition, thereby reducing the need for rescans.

Main Methods:

• Utilized cross-correlation analysis on adjacent phase-encoding lines in k-space data.
• Trained a split-attention residual network for motion artifact severity prediction.
• Validated the model on simulated data and tested its efficacy on a clinical k-space dataset from multi-shot, multi-slice scans.

Main Results:

• Demonstrated the feasibility of detecting motion artifacts directly from raw k-space data.
• The trained split-attention residual network effectively predicted motion artifact severity.
• The proposed method shows potential for real-time artifact detection within the MRI scanner.

Conclusions:

• Direct k-space analysis combined with deep learning offers a promising approach for real-time motion artifact detection in MRI.
• This technique can significantly improve radiology workflow efficiency and reduce associated costs.
• Early detection and correction of motion artifacts enhance diagnostic accuracy and patient satisfaction.