Simulated 2D RASER MRI dataset for AI-driven artefact correction

Data for AI-driven artefact correction in 2D RASER MRI images.

Random images are generated with basic shapes and image transformations. 30 projections of each image are taken, and undergo a RASER (Radiowave amplification by the stimulated emission of radiation) [1] simulation in MATLAB. The data is divided into 3 subsets:

• 10k_images.7z --> standard random images
• 10k_images_WithPump.zip --> projections experience parahydrogen pumping
• 1k_images_20TPI.zip --> high total population inversion (TPI) variations of +/- 20%

File format

Folder structure: {subset}/image{#}/{TPI value}/{filename.csv}

Each folder contains the following files:

• A(0).csv --> Signal amplitude
• d(0).csv --> TPI evolution
• meta.csv --> Meta information
• output(Real and Imag).csv --> Simulated RASER signal
• Phi(0).csv --> Signal phase

Data loading

Scripts for data loading are provided with the code at github.com/mobecks/raser-mri-ai.

References

[1] Sören Lehmkuhl et al., RASER MRI: Magnetic resonance images formed spontaneously exploiting cooperative nonlinear interaction.Sci. Adv.8,eabp8483(2022). DOI:10.1126/sciadv.abp8483