GAME (Growing and Adaptive MEshes)

Luca Ferrarini, Julien Milles

Supported by STW (Project number 06122)

Background

This project started officially on November the 16th 2003. Goal of the project was to develop an algorithm for automatic shape modeling and analysis of challenging 3-dimensional objects. Particular emphasis during the development was given to the analysis of shape changes in brain ventricles, due to atrophy of periventricular structures induced by Alzheimer disease (AD) and other neurodegenerative conditions.

Other methods for shape analysis recently presented in literature are often constrained to the modeling of objects with spherical topology. GAME was developed to overcome this limitation (see Fig. 1), providing Point Distribution Models of non-trivial objects. The shape analysis part was implemented via surface-based non-parametric statistical tests (permutation tests): given two populations of similar objects (e.g., brain ventricles of healthy controls and patients with AD), permutation tests were used to highlight locations on the surface which are significantly different between the two populations (see Fig. 2).

Finally, locations identified by the permutation tests as highlight discriminative can be used to train an intelligent machine in automatically recognize a patient with AD. These possibilities have been explored for the ventricles, and are currently under investigation for other brain structures, such as the hippocampi

GAME Applications Starting from November 2003, we have developed several applications based on our GAME method. The main results of our studies can be found in different articles which have been published over the years. The major results are summarized here:

1. Population-based comparison of Controls and AD. We applied GAME to the analysis of ventricular local shape differences in AD due to atrophy in periventricular structures.

2. Use of Support Vector Machine for automatic detection of AD. Feature extractions based on repeated permutation tests provided the desired features on the ventricular surface. The testing of the classifier was performed on a completely independent AD datasets, showing high sensitivity in AD detection.

3. Analysis of the correlation between the MMSE test and ventricular local shape variations. We investigated the relationship between the MMSE (Mini Mental State Examination) score assigned to a particular subject, and its ventricular shape changes (See Fig. 3). Covering the whole spectrum from normal cognition to AD (passing through Mild Cognitive Impairment), we could prove that MMSE score is correlated with the severity of atrophy in well-localized periventricular structures.

4. Analysis of ventricular shape changes across the whole spectrum of cognitive decline. In this study, we included normal subject, memory complainers (MC), patients with mild cognitive impairments (MCI), and subjects with moderate-to-severe AD. Within each category (MC, MCI, and AD), we investigated asymmetries in the way the neurodegenerative disorder is affective the right and left periventricular structures. We are currently investigating changes across the spectrum.

Status

The project is ongoing. We are currently improving our method by testing different similarity measurements, and including the possibility of simultaneous modeling of multiple shapes.

Gallery

Figure1
Fig 1. The GAME approach in two phases: Growing of the first mesh, and Adaptation to other instances.

Figure2
Fig 2. (top) Results of the permutation tests for two populations of healthy subjects and AD: the p values indicate which locations are found to be significantly different. (bottom) Displacement vectors showing the needed deformation to turn an average healthy ventricle into an average AD ventricle.

Figure3
Fig. 3. The p-value maps show areas of the ventricular surface that highly correlate with the MMSE score. Lower p-values (towards red) indicate higher correlation.

Contact

For further information, please contact:
Dr. JR Milles, PhD
Division of Image Processing
Department of Radiology, 1-C2S
Leiden University Medical Center
P.O. Box 9600
2300 RC Leiden
The Netherlands
Tel. +31 (0)71 526 5342
Fax. +31 (0)71 526 6801
mailto: J.R.Milles@lumc.nl

Publications

Journal/Conference Papers

  • Ferrarini L, Pievani M, Ganzola R, Reiber JHC, Frisoni G and Milles J. “Morphological Hippocampal Markers for Automated Detection of Alzheimer Disease and MCI Converters in MR Images”. Journal of Alzheimer's Disease, 17(3):643-659, 2009.
  • Ferrarini L, Palm WM, Olofsen H, van der Landen R, Blauw GJ, Westendorp RGJ, Bollen ELEM, Middelkoop HAM, Reiber JHC, van Buchem MA, Admiraal-Behloul F. “MMSE Scores Correlate with Local Ventricular Enlargement in the Spectrum from Cognitively Normal to Alzheimer Disease”. NeuroImage, 39(4):1832-1838, 2008.
  • Ferrarini L, Palm WM, Olofsen H, van der Landen R, van Buchem MA, Reiber JHC, Admiraal-Behloul F. “Ventricular Shape Biomarkers for Alzheimer's Disease in Clinical MR Images”. Magnetic Resonance in Medicine. 59(2):260-267, 2008.
  • Ferrarini L, Olofsen H, Palm WM, van Buchem MA, Reiber JHC, Admiraal-Behloul F. “GAMEs: Growing and adaptative meshes for fully automatic shape modeling and analysis”. Medical Image Analysis, 11(3):302-314, 2007.
  • Ferrarini L, Palm WM, Olofsen H, van Buchem MA, Reiber JHC, Admiraal-Behloul F. “Shape differences of the brain ventricles in Alzheimer's disease”. NeuroImage, 32(3):1060-1069, 2006.
  • Ferrarini L, Olofsen H, Palm WM, van Buchem MA, Reiber JHC, Admiraal-Behloul F. “Growing Cell Neural Networks for Fully Automatic Shape Modeling”. In: Proceedings MIUA 2006.
  • Ferrarini L, Olofsen H, Buchem MA van, Reiber JHC, Admiraal-Behloul F. “Fully automatic shape modelling using growing cell Neural Networks”. In: Proceedings MICCAI 2005.

Conference abstracts

  • van den Bogaard SJA, Dumas EM, Ferrarini L, Milles J, van der Grond J, Roos RAC and the TRACK-HD investigator group, “Shape analysis of subcortical structures in Huntington's disease”. In: Proceedings World Congress Huntington Disease, 2009.
  • Ferrarini L, Scheenstra AEH, Frisoni GB, Muskulus M, Pievani M, Ganzola R, Reiber JHC, Dijkstra J and Milles J, “Morphological changes in the hippocampus predict MCI conversion to AD: An MR-based comparison between Moore-Rayleigh and permutation tests”. In: Proceedings International Conference on Alzheimer's Disease, 2009.
  • de Jong LW, Ferrarini L, van der Grond J, Milles J, Bollen ELEM, Westendorp RJG, Middelkoop HAM, Reiber JHC and van Buchem MA, “Regional shape changes of striatum and thalamus in Alzheimer's disease: a morphometrical MRI study”. In: Proceedings International Society for Magnetic Resonance in Medicine, 2009.
  • Ferrarini L, Palm WM, Olofsen H, van der Grond J, van Buchem MA, Reiber JHC, Admiraal-Behloul F. “Significantly asymmetric patterns of periventricular atrophy in Alzheimer Disease, Mild Cognitive Impairment, and Memory Complainers detected on Clinical MR images”. In: Proceedings ISMRM 2008.
  • Ferrarini L, Olofsen H, Palm WM, van Buchem MA, Reiber JHC, Admiraal-Behloul F. Ventricular Shape Biomarkers for Alzheimer's disease in Clinical MR images. In: Proceedings ISMRM 2007
  • Olofsen H, Ferrarini L, Palm WM, van Buchem MA, Reiber JHC, Admiraal-Behloul F. Local Volumetric Analysis of the Brain Ventricles in Alzheimer's Disease using MRI. In: Proceedings ISMRM 2007
  • Ferrarini L, Palm WM, Olofsen H, van Buchem MA, Reiber JHC, Admiraal-Behloul F. Significant local shape differences of the brain ventricles in Alzheimer's patients compared to healthy elderly. In: Proceedings ISMRM 2006