Advancing In Vivo Molecular Bioimaging With Optimal Frequency Offset Selection and Deep Learning Reconstruction for...

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Chemical Exchange Saturation Transfer (CEST) Magnetic Resonance Imaging (MRI) is a promising non-invasive molecular imaging technique.
• Current CEST MRI methods require lengthy data acquisition due to the need for multiple frequency offsets.
• Accelerating CEST MRI data acquisition is crucial for advancing pre-clinical research and clinical applications.

Purpose of the Study:

• To develop a method for accelerating CEST MRI data acquisition.
• To improve the efficiency of molecular imaging using CEST MRI.
• To maintain high-quality CEST MRI data despite reduced acquisition times.

Main Methods:

• A two-step approach was employed: optimization for sparse frequency offset selection and deep learning for Z-spectrum reconstruction.
• An optimization algorithm identified a minimal set of optimal sparse frequency offsets for data collection.
• A deep learning algorithm reconstructed high-resolution CEST MRI Z-spectra from the acquired low-resolution data.

Main Results:

• The optimization technique reduced data collection to 10% of the total frequency offset points.
• Deep learning accurately reconstructed dense Z-spectra, with low RMSE (avg. 0.0094) and MAE.
• The combined approach achieved a 10-fold acceleration in CEST MRI acquisition while preserving data quality.

Conclusions:

• The proposed method significantly accelerates CEST MRI data acquisition through optimal sparse sampling and deep learning reconstruction.
• This accelerated CEST MRI approach maintains high data quality, expanding its utility in biomedical research and clinical settings.
• The technique holds potential for advancing in vivo molecular bioimaging for both basic science and clinical research.