Artificial neural network for Slice Encoding for Metal Artifact Correction (SEMAC) MRI.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neuroscience

Background:

• Metal artifacts significantly degrade MRI quality, particularly after neurosurgery involving metallic implants.
• Slice Encoding for Metal Artifact Correction (SEMAC) is a technique used to reduce these artifacts.
• Accelerating SEMAC MRI is crucial for improving patient throughput and scan efficiency.

Purpose of the Study:

• To develop novel artificial neural networks (ANNs) for accelerating SEMAC MRI.
• To enhance metal artifact correction capabilities in MRI scans.
• To investigate the efficacy of ANNs in reducing artifacts caused by metallic neuro-plating instruments.

Main Methods:

• Development of Multilayer Perceptron (MLP) and Convolutional Neural Network (CNN) models.
• Training ANNs using SEMAC factors (4 or 6) as input and a higher factor (12) as the label.
• Acquisition of phantom and in vivo patient data (T1-, T2-weighted, proton-density) at 3T.
• Comparison of CNN performance against parallel imaging and compressed sensing.

Main Results:

• Both MLP and CNN models effectively suppressed artifacts, yielding results visually and quantitatively comparable to label images.
• CNN demonstrated superior artifact suppression compared to MLP, parallel imaging, and compressed sensing (P < .01).
• Significant metal artifact reduction was observed in both phantom and patient datasets, including T1-weighted images.

Conclusions:

• ANNs offer an effective method for accelerating SEMAC MRI while preserving high-quality metal artifact suppression.
• The developed CNN models show feasibility for clinical use in neuroimaging, warranting further investigation.
• This AI-driven approach holds potential for improving diagnostic accuracy in patients with metallic implants.