The Robotics Institute
RI | Seminar | February 1, 2002

Robotics Institute Seminar, February 1, 2002
Time and Place | Seminar Abstract | Speaker Biography | Speaker Appointments

HAMMER: Hierarchical Attribute Matching Mechasim for Elastic Registration

Dinggang Shen
School of Medicine
Johns Hopkins University

Time and Place
1305 Newell-Simon Hall
Refreshments 3:15 pm
Talk 3:30 pm

In this talk, I will present a new approach for elastic registration of medical images, with applications in magnetic resonance images of the brain. Experimental results demonstrate remarkably high accuracy in superposition of images from different subjects, thus enabling very precise localization of morphological characteristics in population studies. There are two major novelties in the proposed algorithm. First, it uses an attribute vector, i.e. a set of geometric moment invariants (GMI's) that is defined on each voxel in an image, to reflect the underlying anatomy at different scales. The attribute vector, if rich enough, can distinguish between different parts of an image, which helps establish anatomical correspondences in the deformation procedure; it also helps reduce local minima, by reducing ambiguity in potential matches. This is a fundamental deviation of our method, referred to as HAMMER, from other volumetric deformation methods, which are typically based on maximizing image similarity. Second, in order to avoid being trapped by local minima, i.e. suboptimal poor matches, HAMMER uses a successive approximation of the energy function being optimized by lower dimensional energy functions, which are constructed to have significantly fewer local minima. This is achieved by hierarchically selecting features that have distinct attribute vectors, thus drastically reducing ambiguity in finding correspondence. A number of experiments demonstrate excellent performance, which allows superposition of image data from individuals with significant anatomical differences with accuracy comparable to the voxel dimensions.

Speaker Biography
Dinggang Shen is a tenure-track faculty, in the Johns Hopkins University School of Medicine. He received his BS, MS and PhD degrees in Electronics Engineering from Shanghai JiaoTong University in 1990, 1992 and 1995, respectively. He worked as a research fellow for three years in Hong Kong and Singapore, and as a post-doctor for two years in Hopkins. His research interests are in the areas of medical image analysis, computer vision, and pattern recognition. Dr. Shen is on the Editorial Board of Pattern Recognition Journal. He was recently awarded the best paper award at the workshop for Mathematical Methods in Biomedical Image Analysis, which was determined by vote of many of the leading researchers in the field.

Speaker Appointments
For appointments, please contact Yanxi Liu (

The Robotics Institute is part of the School of Computer Science, Carnegie Mellon University.