Title

Detection of brain activated regions from functional MRI, fMRI, and fusion with structural MRI information.

Objectives

The goal of this work is to detect the brain activated regions and fuse this information with the structural MRI data. The work is developed in three distinct phases:

  • Formalization of the detection process of the brain active areas in a statistical inference basis [3], like, for instance, the Bayseian framework using the MAP or mutual information criteria. The goal is to obtain a robust segmentation algorithm as independent as possible of manual parameters defined by the MD [4]. However, the algorithm should not be totally automatic, allowing the MD to interact with the estimation process when needed.
  • Fusion of the fMRI and structural MRI data by aligning them in order to obtain a joint 3D anatomical and functional representation of the brain.
  • Finally the student should test the segmentation and reconstruction algorithms in a real experiment by creating, designing and testing a new paradigm.

Summary

The Statistical parametric mapping [1,2] - SPM - is generally used to identify functionally specialized brain responses and is the most prevalent approach to characterizing functional anatomy and disease-related changes. Functional MRI (fMRI) is a technique that allows to detect the activation of brain regions in response to external stimulus, in a controlled laboratorial environment.


The segmentation process of the active areas is a statistical inference problem where the uncertainty associated to the low contrast and noisy images must be taken into account and modeled. One of the most common techniques is the thresholding method where two brain MRI images, one taken at rest and the other taken after the stimulus, are compared. The activated areas are detected by image difference using a threshold parameter tuned manually by the medical doctor.


Furthermore, the segmented areas, obtained from the fMRI data, must be embebed and fused with the anatomical description of the brain that is obtained from the structural MRI data. To perform this task several difficulties must be overcome, like noise reduction, registering (alignment) and interpolation.

Work Plan

  • Bibliographic research
  • Activated region detection - Statistical Inference approach
  • Alignment and fusion of the functional and structural data
  • Development of new paradigms to test the algorithm using real data.
  • Thesis writing

Workplace

  • Instituto de Sistema e Robótica (ISR) / Instituto Superior Técnico (ISR)
  • Hospital da Cruz Vermelha (HCV)
  • Hospital de Santa Maria (HSM)

References

[1] Statistical Parametric Mapping: An Annotated Bibliography W.D. Penny, J. Ashburner, S. Kiebel, R. Henson, D.E. Glaser, C. Phillips and K. Friston. Wellcome Department of Cognitive Neurology, University College London, http://www.fil.ion.ucl.ac.uk/~wpenny/publications/methods_refs.html
[2] Karl J. Friston, Introduction and Overview, Experimental design and Statistical Parametric Mapping, Statistical Parametric Mapping Course 2007, May 17 (Thurs) - 19 (Sat), 2007.
[3] Bertand Thirion, fMRI data analysis: statistics, information and dynamics, PhD Thesis, Sophia-Antipolis, 2003.
[4] Hae Yong Kim and Javier Oscar Giacomantone, A NEW TECHNIQUE TO OBTAIN A CLEAR STATISTICAL PARAMETRIC MAP BY APPLYING ANISOTROPIC DIFFUSION TO FMRI, SIBGRAPI '02: Proceedings of the 15th Brazilian Symposium on Computer Graphics and Image Processing, 2002.

Initial results


3D reconstruction obtained from the estructural RMI data I'm initially working with.

    Initial results


    3D reconstruction obtained from the estructural RMI data I'm initially working with.