Find out how ICT can support biomedical and clinical researchFind out more. From Clever cars to clever farms... Embedded Systems
Automated Anatomical Structure Extraction for Diagnosis and Population Norms

The Automated Anatomical Structure Extraction for Diagnosis and Population Norms Project:

Study change in shapes of the hippocampus* as aid in diagnosis of (early) Alzheimer's Disease

  • Neuro-degenerative diseases (Alzheimer's Disease (AD), Parkinson's, schizophrenia, depression) are linked to changes in shape  of brain structures.
    • Worldwide incidence of AD is set to grow 4- to 5-fold by 2050 (AIHW 2007, Brookmeyer R. et al., Johns Hopkins Department of Biostatistics Working Paper, 2007)
    • Early intervention at onset of disease would impact mostly on number of persons needing high level of care (Broo07).

  • It is understood that for AD at least, hippocampal volume loss and  shape change occurs well before actual diagnosis of AD.
  • Goal: An effective mechanism for early diagnosis of AD by
    1. Developing discriminative ability for pathology on the basis of hippocampal shape alone;
    2. Developing an effective characterization of shape differences for  pathology.

(*The hippocampus is an area in the forebrain that regulates emotion and memory and plays a part in spatial navigation. Each person has two hippocampi. Changes in its shape and structure have been linked to Alzheimer’s and other diseases.)

 

Who will benefit?

This work will eventually lead to better diagnostic aids for clinicians, but initially the medical research community is the targeted end-user.

 

What are the key features?

The data:

  • The PATH data set:
    • from the Centre for Mental Health Research, Australian National University
    • a population representative study for the Canberra region
    • subset of interest comprises approx. 500 magnetic resonance image (MRI) scans of individuals in the 60s age band
  • The OASIS data set:
    • publicly available here
    • subset of interest comprises approx. 200 MRI scans of individuals in the 60-100 age range, 100 of which are diagnosed with Alzheimer's Disease (AD)
  • All MRI scans are hand-traced for the hippocampus by an expert neurologist.

The Research:

  • Develop mathematical representations of a 3D hippocampus shape ensuring that there is alignment. Hippocampi can be positioned differently in individuals; it is therefore important that the mathematical representation takes this fact into account to ensure alignment is uniform, enabling a valid comparison of shapes.
  • Extract statistics about shape variations.
  • Classify shapes according to characteristics like (i) healthy, (ii) diagnosed with AD, and/or (iii) any clinical measure of interest.
  • After classification of shapes,  determine the nature of the shape differences between the classes.

 

Results to Date (April 09)

  1. Able to show existence of hippocampal shape difference between normal and AD in the OASIS data set and to partially characterize its nature;
  2. Able to obtain similar results when discriminating for sex (female vs male) in the PATH data set;
  3. Able to show existence of correlation between one of our hippocampal shape descriptor and some clinical measures in the PATH data set;
  4. Able to show existence of correlation between same shape descriptor and occurence of AD in the OASIS data set.


Research team

Dr Nick Barnes, Dr Paulette Lieby, Professor Richard Hartley.
PhD Students Pengdong Xiao, Luping Zhou.


Collaborators

Centre for Mental Health Research, Australian National University.
Alfred Psychiatry Research Centre, Monash University. 

 

Publications

L. Zhou, R. Hartley, L. Wang, P. Lieby, and N. Barnes.
Identifying anatomical shape difference by regularized discriminative direction.
TMI,2009.

L. Zhou, L. Wang, R. Hartley, P. Lieby, N. Barnes.
Regularized discriminative direction for shape difference analysis.
MICCAI 2008 Conference Proceedings, 2008, 628--635.

L. Zhou, R. Hartley, P. Lieby, N. Barnes, K. Anstey,  N. Cherbuin, P. Sachdev.
A study of hippocampal shape difference between genders
by efficient hypothesis test and discriminative deformation.
MICCAI 2007 Conference Proceedings, 2007, 375--383.

P. Xiao, N. Barnes, T. Caetano, P. Lieby.
An MRF and Gaussian curvature based shape representation for shape matching.
IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR); IEEE International Workshop on Beyond Multiview Geometry: Robust Estimation and Organization of Shapes from Multiple Cues, Minneapolis, USA, 22 June 2007, (no pagination) on CD rom.

P.Lieby, N. Barnes, and B. D. McKay.
Topological Repair on Voxel-Based Quadrangular Meshes.
Mathematical Foundations of Computational Anatomy, MICCAI 2006 Workshop Proceedings, 2006, 146--155.