Repository logo
 

Open source code for the generation of digital reference objects for dynamic contrast-enhanced MRI analysis software validation.

Accepted version
Peer-reviewed

No Thumbnail Available

Type

Article

Change log

Authors

Gallagher, Ferdia A 
Graves, Martin J 

Abstract

OBJECTIVES: Dynamic contrast-enhanced MR images can be analyzed through the application of a wide range of kinetic models. This process is prone to variability and a lack of standardization that can affect the measured metrics. There is a need for customized digital reference objects (DROs) for the validation of DCE-MRI software packages that undertake kinetic model analysis. DROs are currently available only for a small subset of the kinetic models commonly applied to DCE-MRI data. This work aimed to address this gap. METHODS: Code was written in the MATLAB programming environment to generate customizable DROs. This modular code allows the insertion of a plug-in to describe the kinetic model to be tested. We input our generated DROs into three commercial and open-source analysis packages and assessed the agreement of kinetic model parameter values output with the 'ground-truth' values used in the DRO generation. RESULTS: For the five kinetic models tested, the concordance correlation coefficient values were >98%, indicating excellent agreement of the results with 'ground-truth'. CONCLUSIONS: Testing our DROs on three independent software packages produced concordant results, strongly suggesting our DRO generation code is correct. This implies that our DROs can be used to validate other third party software for the kinetic model analysis of DCE-MRI data. ADVANCES IN KNOWLEDGE: This work extends published work of others to allow customized generation of test objects for any applied kinetic model and allows the incorporation of B1 mapping into the DRO for application at higher field strengths.

Description

Keywords

Humans, Software Validation, Contrast Media, Magnetic Resonance Imaging, Software

Journal Title

Br J Radiol

Conference Name

Journal ISSN

0007-1285
1748-880X

Volume Title

Publisher

Oxford University Press (OUP)
Sponsorship
Cancer Research UK (C12912/A27150)