Principal Investigator:
Hanan Samet
Computer Science Department
University of Maryland
College Park, Maryland 20742
e-mail: hjs@cs.umd.edu
 
 

    Breast cancer is one of the leading causes of death in women.  Mammography is currently the most cost-effective method for early detection of breast cancer.  Between ten to thirty percent of women with breast cancer have negative mammograms, and in about two-thirds of these cases the cancer is evident upon review. Computer aided detection (CAD) and pre-screening can be used to increase the effectiveness of radiologists to avoid these missed diagnoses.  Alternative medical imaging approaches such as ultrasound or MRI could be more effective than mammography at detecting cancers or evaluating malignancy in certain types of women.  A large database of medical images with analysis is required to help train and test the CAD and pre-screening systems.  A database with images from multiple technologies like mammograms, MRI, and ultrasound will also enable research into the effectiveness and usefulness of each technique at cancer screening and the determination of malignancy.  We have developed a pictorial query specification system for this tool that enables users to specify queries by identifying the desired shapes or characteristics and specifying the spatial relationship between them using distance and direction.
 

Pictorial Query With Asymmetry

This example query is over both breasts in a mammogram, with the relevant areas highlighted by the user for a search for similar cases with asymmetry.