Share Email Print
cover

Proceedings Paper

Can the evaluation of a simple test object be used to predict the performance of a contrast-detail analysis in digital mammography?
Author(s): H. Bosmans; K. Lemmens; J. Jacobs; D. Vandenbroucke; F. Zanca; K. Michielsen; B. Verbrugge; K. Smans; G. Marchal
Format Member Price Non-Member Price
PDF $14.40 $18.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

The purpose was to find the correlation between a Figure of Merit (FoM) calculated from a new (simple) test object for Quality Control in digital mammography and CDMAM threshold thicknesses. The FoM included the signal difference to noise ratio, modulation transfer function of the complete system (including scatter and grid) and normalized noise power spectrum. The pre-programmed exposure settings for clinical work were used, as was done for the CDMAM acquisitions. The FoM is calculated from 2 images only (an image from the QC test object and an image of a corresponding homogeneous plate imaged with the same exposure settings). This FoM was evaluated in frequencies that match with the diameters of the gold disks in the CDMAM phantom. Computerized CDMAM analysis uses 16 images per system. The software program "cdcom" (www.euref.org) was used for the 4-AFC experiment. All matrices were averaged, smoothed with a Gaussian filter and psychometric curves were fitted through the correctly detected fractions to obtain the threshold thickness with a detectability of 62.5% for all diameters. Images have been acquired on 10 different systems (2 computed radiography (CR) systems, 6 direct radiology (DR) systems and 2 photon counting systems). The reproducibility of the QC metrics from images of the new phantom was assessed. The standard error on the mean of the FoM was for the highest frequency 8.1% for a CR system and 5.6% for a DR system. The main component in this error is due to the NNPS and the limited number of independent pixels used in this analysis. Parameters calculated from both phantoms are sensitive to variation in mean glandular dose levels. Present results show a weak correlation (R2=0.60) between the FoM at 5lp/mm and CDMAM threshold values for the 0.1mm objects when all system data are pooled. If evaluated for separate systems, the correlation holds promise for automated, periodic performance evaluations of digital mammography systems with the simplified phantom.

Paper Details

Date Published: 12 March 2009
PDF: 9 pages
Proc. SPIE 7263, Medical Imaging 2009: Image Perception, Observer Performance, and Technology Assessment, 726319 (12 March 2009); doi: 10.1117/12.811752
Show Author Affiliations
H. Bosmans, Univ. Hospitals of the KU Leuven (Belgium)
K. Lemmens, Univ. Hospitals of the KU Leuven (Belgium)
J. Jacobs, Univ. Hospitals of the KU Leuven (Belgium)
D. Vandenbroucke, Agfa Healthcare (Belgium)
F. Zanca, Univ. Hospitals of the KU Leuven (Belgium)
K. Michielsen, Univ. Hospitals of the KU Leuven (Belgium)
B. Verbrugge, Univ. Hospitals of the KU Leuven (Belgium)
K. Smans, Univ. Hospitals of the KU Leuven (Belgium)
G. Marchal, Univ. Hospitals of the KU Leuven (Belgium)


Published in SPIE Proceedings Vol. 7263:
Medical Imaging 2009: Image Perception, Observer Performance, and Technology Assessment
Berkman Sahiner; David J. Manning, Editor(s)

© SPIE. Terms of Use
Back to Top