Figures of merit for detectors in digital radiography. II. Finite number of secondaries and structured backgrounds

The current paradigm for evaluating detectors in digital radiography relies on Fourier methods. Fourier methods rely on a shift-invariant and statistically stationary description of the imaging system. The theoretical justification for the use of Fourier methods is based on a uniform background flue...

Full description

Saved in:
Bibliographic Details
Published in:Medical physics (Lancaster) 2004-02, Vol.31 (2), p.359-367
Main Authors: Pineda, Angel R., Barrett, Harrison H.
Format: Article
Language:English
Subjects:
Algorithms
biomedical equipment
Computer Simulation
covariance matrices
diagnostic radiography
digital radiography
DQE
Fourier Analysis
Fourier transforms
image intensifiers
Image Processing, Computer-Assisted
image quality
image resolution
image sampling
lumpy background
medical image processing
Medical imaging
Models, Statistical
NEQ
Non‐ionizing radiation equipment and techniques
Normal Distribution
Physicists
Radiographic Image Enhancement - methods
Random Allocation
Stochastic Processes
Systems analysis
X‐ and γ‐ray sources, mirrors, gratings, and detectors
X‐ray detection
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The current paradigm for evaluating detectors in digital radiography relies on Fourier methods. Fourier methods rely on a shift-invariant and statistically stationary description of the imaging system. The theoretical justification for the use of Fourier methods is based on a uniform background fluence and an infinite detector. In practice, the background fluence is not uniform and detector size is finite. We study the effect of stochastic blurring and structured backgrounds on the correlation between Fourier-based figures of merit and Hotelling detectability. A stochastic model of the blurring leads to behavior similar to what is observed by adding electronic noise to the deterministic blurring model. Background structure does away with the shift invariance. Anatomical variation makes the covariance matrix of the data less amenable to Fourier methods by introducing long-range correlations. It is desirable to have figures of merit that can account for all the sources of variation, some of which are not stationary. For such cases, we show that the commonly used figures of merit based on the discrete Fourier transform can provide an inaccurate estimate of Hotelling detectability.
ISSN:0094-2405
2473-4209
DOI:10.1118/1.1631427