This application is in Displays and Workstations). Current CRT display technology may be adequate if software manipulation of images can overcome limitations of spatial resolution, dynamic range and luminance. Possible manipulations include: gray (tone) scale transformations, spatial filtering and compensation for thickness variation at the breast periphery. There are many variables involved and the program announcement listed the need from reliable computational models that predict human performance to reduce the parameter space that must be investigated by psychophysical studies. I have been working on human observer models for many years and recently began using hybrid images (with realistic lesions added to normal digital clinical images). For mammograms, I found that theoretically predicted variation of contrast thresholds for lesion detection as a function of lesion size was very surprising but was confirmed by experiment. For l lesions larger than 1 mm, the contrast threshold actually increased as lesion size increased due to the unusual statistical properties of normal breast parenchymal patterns (structure). I have also done experiments with simulated images and had good success in modeling all of my results. In this grant I propose to do three things. ( 1 ) Extend the 2AFC studies to compare human lesion detection and discrimination performance with observer model predictions for sub-classes of breast structure statistics. (2) Do observer experiments using more clinically realistic search and identification tasks to compare human results and model predictions. (3) Do observer experiments to evaluate performance using images modified by a variety of thickness compensation and adaptive contrast enhancement. I will also investigate a number of proposed gray scale transformations. All work will be done using a database of 800 General Electric full breast digital images.

Agency
National Institute of Health (NIH)
Institute
National Cancer Institute (NCI)
Type
Research Project (R01)
Project #
5R01CA087738-03
Application #
6633805
Study Section
Diagnostic Imaging Study Section (DMG)
Program Officer
Liu, Guoying
Project Start
2001-04-11
Project End
2006-03-31
Budget Start
2003-04-01
Budget End
2004-03-31
Support Year
3
Fiscal Year
2003
Total Cost
$314,811
Indirect Cost
Name
Brigham and Women's Hospital
Department
Type
DUNS #
030811269
City
Boston
State
MA
Country
United States
Zip Code
02115
Burgess, Arthur E; Judy, Philip F (2007) Signal detection in power-law noise: effect of spectrum exponents. J Opt Soc Am A Opt Image Sci Vis 24:B52-60
Burgess, Arthur (2004) On the noise variance of a digital mammography system. Med Phys 31:1987-95
Burgess, Arthur; Jacobson, Francine; Judy, Philip (2003) Mass discrimination in mammography: experiments using hybrid images. Acad Radiol 10:1247-56