Steps toward a Simulator for Magnetic Resonance Images of the Neonatal Brain

Purpose: With the increased interest in computer-aided analysis of magnetic resonance images (MRI) from neonates, objective methods for evaluation of image processing algorithms become indispensable. Due to anatomical differences between neonates and adults, it is not reasonable to use an existing adults' brain simulator (i.e. Brainweb) for this purpose. In this paper, we present our approach to implement a MRI simulator especially for images of neonatal brain tissues.

Materials and Methods: Basically, the simulator combines the anatomical complexity provided by a neonatal digital brain phantom with the simulation of MRI signal intensity. Our phantom consists of fuzzy models for scalp, fat, muscle, skull, dura mater, gray matter, myelinated white matter, non-myelinated white matter and cerebrospinal fluid. It has been created by automatic segmentation using SPM5 followed by manual correction of multispectral 3D MR images of a selected subject. In the simulator, the MR signal intensity is computed for the different types of neonatal neurocranium tissue by applying a discrete-event Bloch equation with the corresponding phenomenological T1/T2/T2* relaxation times and proton density. Then, the simulated tissue intensities are mapped to the 3D brain phantom. Finally, the simulated image is generated by adding noise to the raw image.

Results: Our simulator is based on a phantom which is composed of voxels sized 1×1×1mm³ and has been normalized to the stereotaxic space defined by the neonatal brain template 'GRAMFC_T39–42'. It has been implemented in MATLAB with an interface which allows the user to select different resoluteons and different acquisition sequences as spin echo, inversion recovery and gradient echo.

Conclusion: The presented simulator allows to create standard MR images of neonate tissue for objective evaluation and validation of image processing algorithms.