Small sample size learning for shape analysis of anatomical structures

P. Golland, W. E. Grimson, M. E. Shenton, R. Kikinis
MICCAI 2000: Third International Conference on Medical Robotics, Imaging and Computer Assisted Surgery
Pages 72-82
October 11-14, 2000

Download full paper

Abstract

We present a novel approach to statistical shape analysis of anatomical structures based on small sample size learning techniques. The high complexity of shape models used in medical image analysis, combined with a typically small number of training examples, places the problem outside the realm of classical statistics. This di culty is traditionally overcome by rst reducing dimensionality of the shape representation (e.g., using PCA) and then performing training and classi cation in the reduced space de ned by a few principal components. We propose to learn the shape di erences between the classes in the original high dimensional parameter space, while controlling the capacity (generalization error) of the classi er. This approach makes signi cantly fewer assumptions on the properties and the distribution of the underlying data, which can be advantageous in anatomical shape analysis where little is known about the true nature of the input data. Support Vector Machines with Radial Basis Function kernels are used as a training method and the VC dimension is used for the theoretical analysis of the classi er capacity. We demonstrate the method by applying it to shape classi cation of the hippocampus-amygdala complex in a data set of 15 schizophrenia patients and 15 normal controls. Using our technique, the separation between the classes and the con dence intervals are improved over a volume based analysis (63 small sample size learning theory provides us with a principled way of utilizing shape information in statistical analysis of the disorder e ects on the brain.

Reference

Golland P, Grimson WE, Shenton ME, Kikinis R. Small sample size learning for shape analysis of anatomical structures. In MICCAI 2000: Third International Conference on Medical Robotics, Imaging and Computer Assisted Surgery. Pittsburgh, PA, 2000;72-82.

Grants

NIH/NIMH 2R01 MH01110, NIH RO1 MH50747

Research area

shapeanalysis
© 2013 Psychiatry Neuroimaging Laboratory | Last updated 04.15.2013