Scanner comparisons Comments given to the physicist The image - - PDF document

WC2003 August 2003

Methods for Comparison of Image Quality and Dose in Computed Tomography Sue Edyvean, Nicholas Keat, Maria Lewis, Julia Barrett, David Platten ImPACT (Imaging Performance assessment of CT Scanners) London UK

www.impactscan.org

WC2003 August 2003

Scanner comparisons

• Comments given to the physicist

– “The image quality on the old scanner is better than on the new scanner” – “The scanner at the other clinic I work in gives better images” – “I want to buy a low dose scanner”

• Differences in image quality and dose

– Intrinsic differences ? – Due to scanning protocols alone ?

WC2003 August 2003

Image quality and dose dependence

• Intrinsic factors

– Detectors

• material
• detector configuration
• numbers of detectors

– Data acquisition rates – Software corrections – Filtration – Focal spot – Geometry

• ie focus-axis,

focus-detector distances

x-ray tube x-ray detectors slip rings

WC2003 August 2003

Image quality and dose dependence

• Scan protocols

– Tube current – Tube voltage – Reconstruction algorithms – Collimation width – Helical pitch – Interpolation algorithms – Image slice thickness

• Clinical application

Image Plane

WC2003 August 2003

Image quality and dose parameters

• Comparison of image quality

– Measured data not perception based information

• Image quality

– Image noise – Spatial resolution (scan plane) – Image thickness (spatial resolution in z-axis)

• Radiation dose

– CTDIvol

WC2003 August 2003

Image Noise

image appropriate phantom region of interest (roi):

noise = standard deviation σ

WC2003 August 2003

Spatial resolution (scan plane)

CTw CTPTFE

Position Along Edge

100 50 10

MTF (%) spatial frequency (cm-1)

high contrast edge ESF → LSF → MTF

sharp image smooth image

Spatial resolution (scan plane)

WC2003 August 2003

Image thickness (z-axis spatial resolution)

• Inclined Al ramps (axial): FWHM of z-sens profile

WC2003 August 2003

Helical image thickness

• images reconstructed at

intervals of 1/10 of the slice thickness

• mean CT number at centre

used to create the profile

50 100 150 200

• 10
• 8
• 6
• 4
• 2

2 4 6 8 10

reconstructed image position mm CT Numbers

2.5mm

tungsten or gold disc, 0.05 mm

WC2003 August 2003

Radiation Dose (CTDIvol)

Pitch 1 Pitch 2

• CTDIw = CTDI100 averaged in scan plane

= 1/3.CTDIC + 2/3.CTDIp

• CTDIvol

= CTDIw averaged along z-axis = CTDIw / Pitch

WC2003 August 2003

Comparison of image quality and dose

• Start with typical clinical protocol

– (eg abdomen) minimise built in clinical application factors

• Standardise scan parameters where possible

– minimise effect from these variables – eg kV, pitch, collimation, nominal image width

• Correct for known parameter inter-dependencies

– eg. noise and dose, noise and image width

• Obtain trend data for other inter-dependencies

– eg. noise and resolution

WC2003 August 2003

Scanning dependent factors

• Noise with kV

Noise Versus kV

0.50 1.00 1.50 2.00

80 100 120 140

kV

Hard Beam Soft Beam

Values normalised to 120 kV

Scanner A Scanner B

WC2003 August 2003

Relative CTDI

0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 5 10 15 20 25 30 35

collimation relative CTDI

four and sixteen slice poor single slice good single slice

Scanning dependent factors

WC2003 August 2003

• Single slice scanner
• Nominal image width = collimation width
• Multi slice scanner

– Axial

• Collimation width (eg. 10 mm)
• Detector acquisition width (eg 2.5 mm) = nominal image width

– Spiral

• Collimation width (e.g 10 mm)
• Detector acquisition width (eg 2.5 mm)
• Nominal reconstructed Image width

z-axis eg 10 mm collimation: 4 x 2.5 mm

Standardise scan parameters

WC2003 August 2003

Correct for known parameter inter-dependencies

• Noise with dose (tube current), slice thickness

1 mA

α α α α

noise

( ) ( )

2 1 slice thickness noise 2

( ) ( )

α α α α

WC2003 August 2003

• Noise, resolution ← reconstruction filter, focal spot
• Noise and resolution with reconstruction filter

– GE Lightspeed

h/b: soft, standard, lung, detail, bone, edge

– Siemens Volume Zoom

h/b: AH/AB..10,20,30,40,50,60,70

– Toshiba Aquilion

h: FC20,21,22,23,24,25,26,27,28,30,80: b: FC….10,11,12,13,14,30

– Marconi (Philips) MX8000

h: A,EB,EC,B,C,D; b: A,EC,B,C,D

Identify other inter-dependent relationships

WC2003 August 2003

Noise and reconstruction filter

0.1 1.0 10.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 Average (MTF50, MTF10) % noise for 40 mGy Siemens S4

noise and resolution images reconstructed with different algorithms

%noise for 40 mGy and 5 mm image width

WC2003 August 2003 0.1 1.0 10.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 Average (MTF50, MTF10) % noise for 40 mGy GE LightSpeed Plus Philips Mx8000 Siemens S4 Toshiba Aquilion

A B C D

Noise and reconstruction filter

%noise for 40 mGy and 5 mm image width

WC2003 August 2003

R2 = 0.9491 R2 = 0.9969 Power fits R2 = 0.9998 R2 = 0.2977 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 Average (MTF50, MTF10) % noise for 40 mGy GE LightSpeed Plus Philips Mx8000 Siemens S4 Toshiba Aquilion

Noise and reconstruction filter

A B C D

WC2003 August 2003

R2 = 0.9585 R2 = 0.9823 Linear fits R2 = 0.9996 R2 = 0.2833 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 Average (MTF50, MTF10) % noise for 40 mGy GE LightSpeed Plus Philips Mx8000 Siemens S4 Toshiba Aquilion

Noise and reconstruction filter

A B C D

WC2003 August 2003

Noise Resolution Image width Dose A 3 HU 3.4 c/cm 5 mm 20 mGy B 4.5 HU 3.8 c/cm 4.5 mm 25 mGy A 3 HU 3.4 c/cm 5 mm 20 mGy B 3 HU 3.4 c/cm 5 mm 36 mGy

Image quality and dose: normalise data

WC2003 August 2003

Noise Resolution Slice width Dose A 3 HU 3.4 c/cm 5 mm 20 mGy B 4.5 HU 3.8 c/cm 4.5 mm 25 mGy A 3 HU 3.4 c/cm 5 mm 20 mGy B 4 HU 3.4 c/cm 5 mm 20 mGy

Image quality and dose: normalise data

WC2003 August 2003 0.1 1.0 10.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 Average (MTF50, MTF10) % noise for 40 mGy GE LightSpeed Plus Philips Mx8000 Siemens S4 Toshiba Aquilion

%noise for 40 mGy and 5 mm image width

A B C D

Image quality and dose: display data

WC2003 August 2003

• Noise, dose and slice thickness relationship

– established

• Noise and resolution relationship

– from graph (best fit) – theoretical

Image quality and dose: single numerical factor z 1

2 α

σ D 1

2 α

σ

WC2003 August 2003

Theoretical : imaging theory

• Rodney Brookes and Giovanni di-Chiro (1976)

– Statistical limitations in x-ray reconstructive tomography

• Medical Physics Vol 3, No 4 July 1976
• Riederer S.J., Pelc N.J. and Chesler D.A. (1978)

– The Noise Power Spectrum in Computed Tomography

• Physics in Medicine and Biology 1978 23(3), 446-454

D z f 3

2 α

σ

z 1

2 α

σ D 1

2 α

σ

2 / 3

f α σ

limiting resolution

WC2003 August 2003

• Single numerical quality factor
• High Q

– High image quality

• high resolution, low noise and thin slice

– Low dose

Image quality and dose

D z f 3

2 α

σ

D z f const Q σ

2 / 3

. =

WC2003 August 2003

Noise Resolution Slice width Dose A 3 HU 3.4 c/cm 5 mm 20 mGy B 4.5 HU 3.8 c/cm 4.5 mm 25 mGy A 3 HU 3.4 c/cm 5 mm 20 mGy B 3 HU 3.4 c/cm 5 mm 36 mGy Q = 1.0 Q = 0.7

Image quality and dose: single numerical factor

WC2003 August 2003

Comparison of image quality and dose

• Start with typical clinical protocol
• Standardise scan parameters where possible

– eg collimation, kV, pitch, nominal image width

• Correct for known parameter inter-dependencies

– eg. noise and dose, image width

• Obtain trend data for other inter-dependencies

– eg. noise and resolution

• Review performance data

– Normalise values to give single value of dose or noise – Display normalised noise against resolution – Calculate a single quality performance parameter

WC2003 August 2003

Methods for Comparison of Image Quality and Dose in Computed Tomography Sue Edyvean, Nicholas Keat, Maria Lewis, Julia Barrett, David Platten ImPACT (Imaging Performance assessment of CT Scanners) London UK

www.impactscan.org