MRS Synthetic Data Working Group :: Best Practices for Simulating Data

JavaScript isn't enabled in your browser, so this file can't be opened. Enable and reload.

Welcome to another survey in by the Data & Code Sharing Committee's MRS Synthetic Data Working Group.

This short survey seeks to understand what's important to have in a synthetic data generation framework. Some experimental factors and artifacts may not be applicable to every clinical scenario. To make the results generalizable to a broader range of applications, please assume a hypothetical scenario in which all factors are applicable.

Thank you for taking time to help us with this endeavor!

Email *

When using a synthetic data generation method, how important are the following features to you?
*

Not at all important

Slightly important

Important

Fairly important

Very important

No opinion

Having a ground truth quantification

Having full control on various parameters (linewidth, SNR, noise etc.)

Generating more data, similar to your in-vivo dataset

Generating more data, varying from your in-vivo dataset

Generating data with realistic artifacts

Generating data with realistic residual artifacts (post-correction)

Not at all important

Slightly important

Important

Fairly important

Very important

No opinion

Having a ground truth quantification

Having full control on various parameters (linewidth, SNR, noise etc.)

Generating more data, similar to your in-vivo dataset

Generating more data, varying from your in-vivo dataset

Generating data with realistic artifacts

Generating data with realistic residual artifacts (post-correction)

Please indicate which of the following experimental factors you think SHOULD BE included when generating your synthetic data: *

Pulse sequence type (PRESS, STEAM, etc.)

Exact pulse timings (i.e. TE1 and TE2 in PRESS)

Exact RF pulse shapes (idealized shapes)

Exact RF pulse shapes (imperfect shapes)

Independent T2 relaxation rate for each metabolite

Independent T1 relaxation rate for each metabolite

Independent T2/T1 relaxation rates for each spin-system sub-unit (moiety)

Mixed line shapes (not purely gaussian or lorentzian)

Variation of lineshape values for metabolites (sampling from a range instead of a fixed value)

Macromolecules (non-lipid signals)

Lipid signals

Baseline distortion (not arising from macromolecules or lipids)

Other:

Required

Please indicate which of the following acquisition artifacts you think SHOULD BE included when generating your synthetic data: *

Line shape distortions from eddy current effects

Line shape distortions from other effects (e.g. large susceptibility effects)

Residual water contamination

Spurious echoes / unwanted coherences

Metabolite-specific frequency shifts, naive

Metabolite-specific frequency shifts, physics-informed

Post-processing residual artifacts

Zero-order phase offsets (phase drifts for transients) (uncorrected or residual offsets)

First-order phase offsets (phase drifts for transients) (uncorrected or residual offsets)

Frequency drift for transients (uncorrected or residual offsets)

Coil-sensitivity scaling for transients

Individual SNR values for transients

Overall SNR per spectrum

Other:

Required

What type of spectra would you find useful? *

Standard, pre-processed

Raw, unprocessed

Relaxation series

Diffusion-weighted series

Edited spectra

2D COSY

2D J-resolved

Multi-coil transients (of any type of spectra)

Other:

Required

Given a standardized simulation framework, would any of the following programming languages hinder your adoption of such a software package?

No there are not

Matlab

Python

C/C++

Fortran

Other:

Clear selection

Are there any other features you would like to see in a data generation framework? (If yes, please describe this/these feature(s))

Submit

Clear form

Never submit passwords through Google Forms.

This content is neither created nor endorsed by Google. Report Abuse - Terms of Service - Privacy Policy

Forms