.

This research addresses two main technical challenges in Medical Imaging Analysis. The first is to improve the measurement accuracy of enhancing brain tissue by reducing the influence of large inter-slice distances in the low resolution scans in Magnetic Resonance (MR) images. The other main challenge is to accurately label the enhancing tissues by differentiating them from normal tissues in the brain.
We present a novel method to enhance image resolution by using a multi-resolution pyramid structure to combine orthogonally scanned Low Resolution Volumes (LRVs). Combination of different LRVs in this coarse-to-fine fashion solves the problem caused by large inter-slice spacing. The proposed method is applied to improving consistency and accuracy of tumor measurement. In the second part of this thesis, a new knowledge based framework is presented to determine enhancing brain tissue volume. While most previous methods require manual removal of non-tumor segmentations, our method provides an automatic and consistent way to measure enhancing brain tissue by utilizing the knowledge of anatomical structures provided by a probabilistic brain atlas and a priori knowledge of enhancing tissue intensity.
These methods are validated using both simulated and clinical brain MR images containing tumor and promising experimental results are demonstrated. The proposed method provides doctors a better way to accurately determine tumor size and can benefit the medical community.

Outline of Studies:

• Major: Computer Science
• Educational Career:
  
  • B.S.E.E., 2006, University of Central Florida, Orlando, Florida
• Committee in Charge:
  • Dr. Mubarak Shah,  Chair
  • Dr. Samuel Richie
  • Dr. Xun Gong
  • Dr. Xin Li

.