Improving
Clinical Diagnosis by Change Detection in Image Sequences (DAMD17-98-8045,
DAMD17-98-8044)
Due
to the highly variable behavior of cancerous masses, mammographic
examination of breast cancer requires an accurate comparison of
mammograms taken over a period of time and emphasizes on the change
of breast tissue associated with cancer development. Since the
lack of quantitative change detection scheme for mammograms has
caused too many call-backs and unnecessary biopsies, this work
aims to develop an automatic change detection method to quantitatively
extract the clinically important changes of suspicious lesions
for improved detection of breast cancer. We will apply our extensive
expertise in image registration, deformable tissue modeling, and
image segmentation, to a series of mammograms so that change information
can be highlighted and further analyzed by radiologist's closer
clinical inspection and by our well-established computer-aided
diagnosis (CAD) system.
We
will build a site model for each individual patient for monitoring
the breast tissue changes and extend our current research on image
registration, soft tissue modeling, and image segmentation, to
the early detection of breast cancer. Specific aims include: 1)
registration and segmentation of deformable breast tissue structures
across a series of mammograms; 2) construction of a site model
of the mammogram for individual patients showing the locations
of regions of interest and associated diagnostic information;
3) identification of clinically significant changes in both global
and local mass areas within the breast; and 4) integration and
evaluation of the developed techniques with existing CAD prototype.
At the conclusion of this project, we anticipate achieving the
following: 1) establish a reliable technique of monitoring breast
tissue changes associated with cancerous masses; 2) deliver a
CAD prototype that can incorporate tissue change information from
additional mammograms; 3) evaluate the merit of combining change
detection and CAD for improved clinical diagnosis using multiple
mammograms; and 4) acquire the experience necessary to explore
multimodality imaging for unified detection, diagnosis and treatment
assessment of breast cancer.
Copyright
©2004, Computational Bioinformatics and Bioimaging Laboratory
(CBIL), Alexandria Research Institute, Virginia Tech. Jointly
with The Catholic University of America.
Last
Updated: 03/22/2004. Suggestions/Comments
- Webmaster
