2166-Handout

College on Medical Physics. Digital Imaging Science and Technology to

Enhance Healthcare in the Developing Countries

13 September - 1 October, 2010

Digital Radiography Image Parameters

SNR, MTF, NPS, NEQ, DQE

Slavik Tabakov
King's College London
United Kingdom
 Digital Radiography Image Parameters
SNR, MTF, NPS, NEQ, DQE

Dr Slavik Tabakov

Dept. Medical Eng. & Physics,

King's College London

E-mail : slavik.tabakov@kcl.ac.uk

V
D2 Signal-to-Noise Ratio: SNR

'C = [D2 – D1]/ V

D1 SNR
I –– Intensity
D –– Density
E - Exposure
Subject Contrast Limiting Contrast
'C =I2 – I1 (Signal) and Noise

Visual contrast
'C = logI2 – logI1
Radiographic contrast
'C = [D2 – D1]/D1 high contrast + noise ; low contrast + noise
 Ideal

real

Contrast Transfer Function

Concept

10% - cut-off frequency (lim. sp. res.)

 Spatial frequency
(Line pair per mm)
and
Blur measurement
(the problem in medical
imaging is larger as the
objects are not with sharp
edges, what changes the
peripheral absorption µd )

Example from the image >>>

At high blur limiting Sp. Resolution is 4.5 lp/mm
In this case the smallest object to be seen is =
= 1/(2 x 4.5) = 0.11mm [ 1/(2 x sp.res) ]
Modulation Transfer Function
MTF=(recorded signal f)/(origin. signal f);
also MTF(f) = | FT{LSF(x)} |

MTF~m = sin ʌ.uf / ʌ.uf ,where

uf = f/L * (M-1)/2M ,where
M-magnif.; f –– focal spot; L –– period of the
structure (~ to spatial frequency)
Assessment of
Spatial Resolution (Lp/mm)

Assessment of
Contrast Resolution
(C/D)
 Typical Contrast/Detail diagram for the Leeds TO 10 phantom for
various II filed sizes (old and new II)

Contrast and Resolution of various X-ray detectors and methods

Contrast:
1. CT
2. Film
3. Fluo

Resolution:
1.Film
2.Fluo
3.CT Visible
space

Ideal Contrast-
Detail curve
 Image Quality in Digital Imaging
* Noise in the image:
superposition of a meaningless set of signals
over meaningful signals

There are two primary contributions to noise;

the quantum noise from statistical fluctuations
in the number of X-ray quanta detected per
unit area (quantum mottle) and the noise
arising from variations in the imaging system

CT noise (on image) is measured as standard

deviation of mean CT values (it varies most
often from 1 to 10 CT)
- Standard deviation (noise):

V = [6(CTi - CTmean)2 / (n-1)]1/2 Noise is the main limitation of contrast

resolution
- Min. contrast > noise level (HU)

Noise Power Spectrum

(Wiener noise spectrum)
••NPS takes both the magnitude and
texture of noise into account

The main problem in using the NPS

to determine noise properties in an
imaging system is that the method
requires averages (integrals) over an
infinite data set to obtain the true NPS
but we have finite set of data.
 Different elements of the Imaging chain
contribute to the overall NPS
The Normalised NPS (NNPS), is related to the
large area signal (LAS) i.e. the average pixel
value in the image.

NPS
NNPS
LAS 2
N
The Noise Equivalent Quanta (NEQ) is a measure of the
signal to noise ratio (SNR) of an imaging system.
SNR
NEQ is the number of quanta N incident on an ideal
N

SNR 
detector that would give the same output SNR as a non-
ideal detector ( noise ı ~ ¥N , N- incident photons ) 2
An ideal detector will detect all incident quanta, will add
NEQ out
no noise and has no blur.
NEQ can be considered as the number of quanta used in MTF ( f )
acquiring an image at a particular dose level, as a SNRout
function of spatial frequency. NPS ( f )

NEQ drops off rapidly with increasing spatial frequency because both MTF and
NNPS drop off with increasing spatial frequency.

NEQ is often plotted against quantum fluence. This shows well the relationship
between x-ray dose and the image quality of a system.

The graph on the right is for a screen-film system and it is clear that a specific dose
is required to achieve the best NEQ.
 NEQ as a function of detector exposure taken at 2 lp/mm

Selenium detector (blue) is compared to (1), (2) -

Conventional Screen-film combinations
and (3) - ST III storage phosphor screen

Selenium detector has smaller blur

due to the specific technology

NEQ is useful for measuring how efficient the imaging system is with respect to the
incident X-ray quanta used in image formation and can directly describe the potential of
the actual image. However when comparing the ideal and non-ideal detector there are
losses arising from X-ray quanta not being absorbed, increased noise being introduced by
the system during the conversion process, and additional noise independent of the X-ray
quanta being introduced by the system. The efficiency can be calculated by comparing
the NEQ to the quantum fluence of the incident radiation. This leads to DQE.

The Detective Quantum Efficiency (DQE) is effectively a measure of how well the x-
rays are used in an imaging system (the efficiency of converting input quanta to signal
or the efficiency of preserving the SNR).
DQE is influenced by the MTF, readout and quantum noise, and detection efficiency in
an imaging system.

DQE is defined as the ratio of the DQE can also be expressed in terms
squared SNR at the output of the of measurable quantities, including
detector to the SNR of the input. MTF and NPS, where
DQE=1 means that all produced
qo is the mean incident fluence and
quanta are used to make the image
G is the system gain.
without any noise

DQE
SNR  2
out NEQ q0G 2 MTF 2
SNR  2
in N
DQE
NPS
DQE is independent of the detector technology and focuses only on its input and output
signals. This way it can be used as a method of comparison of different imaging systems
(a quantitative figure of merit).
DQE is a measure of image quality (high DQE = better the resolution of the img. system).
As with NEQ the DQE is often plotted against x-ray fluence;

The figure below (left) is for the same screen-film system (shown in NEQ).
Obviously x-ray dose for this system is important for the image quality (hence under or over-
exposed film provides a useless image). Also DQE drops quickly with increasing spatial
frequency (most noise is at high spatial frequency and affects more higher frequency signals).

The figure below (right) shows approximate DQE for various detectors.
The CsI detector has better DQE compared with a-Se detector (the worst being screen-film).

A detector that has a DQE value of twice that of another is said to be twice as efficient
and therefore only requires half the amount of X-ray dose to produce an image with the
same SNR.
Hence, in theory, the higher the DQE of the detector the lower the patient exposure dose.

Another general property of the DQE is that it increases with decreasing X-ray energy due
to more efficient X-ray absorption at low kVp values. This reduction becomes less
prominent for higher spatial frequencies. Contrast resolution for low-contrast details also
improves with the DQE.

A problem of the DQE is that it describes only the detector (not the whole imaging system)

Current DQE values of

direct and indirect
digital systems and the
current research areas
References:
M. Spahn. Flat Detectors and their Clinical Applications. European Journal of Radiology 15 1934-1947 (2005)
Varian medical systems. www.varian.com. Online
Steven Evans Digital Imaging. Course notes
W Zhao, J Law, D Waechter, Z Huang and J Rowlands. Digital radiology using active matrix readout of amorphous selenium: Detectors with high voltage
protection. Medical Physics 25 (4) (1998)
S Webb. The Physics of Medical Imaging. IOP Publishing Ltd
D Darambara. Introduction to MTF, NPS, DQE. Course notes
J Dobbins, E Sarnei, N Ranger and Y Chen. Intercomparison of methods for image quality characterization. II. Noise Power Spectrum. Medical Physics 33
(5) (2006)
K Faulkner and B Moores. Analysis of x-ray computed tomography images using the noise power spectrum and autocorrelation function. Physics in
Medicine and Biology 29 (11) (1984)
K Hanson. A simplified method of estimating noise power spectra. Physics of Medical Imaging 1998.
P Monnin, D Gutierrez, S Bulling, D Lepori, J-F Valley and F Verdun. A comparison of the performance of modern screen-film and digital mammography
systems. Physics in Medicine and Biology 50 2617-2631 (2005)
A Workman and D Brettle. Physical performance measures of radiographic imaging systems. Stockton Press 1997
S-A Starck, M Bath and S Carlsson. The use of detective quantum efficiency (DQE) in evaluating the performance of gamma camera systems. Physics in
Medicine and Biology 50 1601-1609 (2005)
R Schaetzing. A Comparison of Screen/Film and Digital Imaging: Image Processing, Image Quality and Dose. AGFA Coorporation. Online presentation
E Samei and M Flynn. An experimental comparison of detector performance for computed radiography systems. Medical Physics 29 (4) 447-452 (2002)
C Schaefer-Prokop and M Prokop. Digital radiography of the chest: Comparison of the selenium detector with other imaging systems. Medicamundi 41 (1)
1997.
J Seibert. Tradeoffs between image quality and dose. Pediatric Radiography 34 183-195 (2004)
http://www.kcare.co.uk/Education/glossary.htm#glossaryn
http://www.aapm.org/meetings/99AM/pdf/2798-87374.pdf
http://www.gehealthcare.com/inen/rad/xr/education/dig_xray_iq.html