Multi-contrast generation and quantitative MRI using a transformer-based framework with RF excitation embeddings.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Neuroscience

Background:

• Magnetic Resonance Imaging (MRI) is crucial for clinical diagnosis, relying on radiofrequency (RF) excitation.
• Current MRI protocols require multiple RF sequences for comprehensive biophysical data, resulting in lengthy examination times.
• There is a need for faster MRI techniques to improve patient comfort and diagnostic efficiency.

Purpose of the Study:

• To develop a novel, accelerated MRI framework using a vision transformer.
• To generate a wide range of quantitative image contrasts from limited calibration data.
• To validate the framework's speed and diagnostic potential across different subjects and imaging sites.

Main Methods:

• Development of a vision transformer-based framework (TBMF) utilizing RF excitation information.
• Acquisition of per-subject calibration data within 28.2 seconds.
• Generation of quantitative molecular, water relaxation, and magnetic field maps.

Main Results:

• The TBMF successfully generated diverse quantitative MRI contrasts.
• Validation across healthy subjects and a cancer patient at two imaging sites.
• The framework demonstrated a 94% acceleration compared to alternative MRI protocols.

Conclusions:

• The transformer-based MRI framework (TBMF) offers a significant speed improvement for MRI examinations.
• This accelerated approach can generate valuable quantitative data for diagnosing pathologies.
• TBMF has the potential to enhance the study of human brain tissue composition in various diseases.