Comparison of CT Scanner Image Noise, Image Width, Dose and Spatial Resolution Using Standard Test Methods Sue Edyvean, Nicholas Keat ImPACT (Imaging Performance Assessment of CT Scanners) London UK www.impactscan.org ImPACT 1 AAPM 2004 ImPACT • An evaluation group of the UK Department of Health – MHRA (Medicines and Healthcare products Regulatory Agency) • St George’s Hospital, London • Perform independent evaluations of CT scanners 2 AAPM 2004 Comparison of image quality and dose • • • • Issues of comparison Test methods Scanning approach Dealing with the data 3 AAPM 2004 1 How can we compare image quality? ? 4 AAPM 2004 Image quality comparison • Defining image quality – perception or numbers • Image quality descriptors – Image noise – Spatial resolution (scan plane) speckle sharpness – Image thickness (spatial resolution in z-axis) – Contrast – Artefacts 5 AAPM 2004 Image quality and radiation dose • Differences in image quality and dose – Intrinsic differences in scanner design – Due to scan protocol and scan parameters 6 AAPM 2004 2 Intrinsic factors • Scanner design – Detectors • material • numbers, rows • configuration Whizzo CT x-ray tube filtration FAD – Data acquisition rates – Software corrections FDD • e.g. beam hardening – X-ray tube • filtration • focal spot detectors – Geometry • focus-axis, focus-detector distances 7 AAPM 2004 Scan protocols • Clinical application – X-ray filter, software corrections (eg beam hardening) • Scan and reconstruction parameters – X-ray tube • current, voltage, focal spot – – – – – – – Scan time Field of view Reconstruction algorithm Image slice thickness Collimation width No. slices (→ detector acquisition width) Helical pitch and interpolation algorithm 8 AAPM 2004 Comparison of image quality and dose • • • • Issues of comparison Test methods Scanning approach Dealing with the data 9 AAPM 2004 3 Image quality descriptors • Standard test methods – Noise – Resolution – Image slice width imaged slice sharp image MTF (%) 100 smooth image 50 fwhm 10 aluminium ramps spatial frequency (cm-1) Std. Dev. σ fwhm of image profile Average MTF50 and MTF10 10 AAPM 2004 Image quality – helical scanning • Noise – Same test • Spatial resolution – Same test • Image width tungsten or gold disc, 0.05 mm – Use helical test tool → fwhm CT Numbers 200 2.5mm 150 100 z-axis fwhm 50 0 -10 -8 -6 -4 -2 0 2 4 6 8 10 Z-axis position of image (mm) 11 AAPM 2004 Radiation Dose (CTDIw, CTDIvol) • CTDIw = dose averaged in scan plane • CTDIvol = CTDIw / Pitch = dose averaged in scan plane and z-axis ie volume averaged CTDI Pitch 2 12 AAPM 2004 4 Comparison of image quality and dose • • • • Issues of comparison Test methods Scanning approach Dealing with the data 13 AAPM 2004 Scanning approach • Start with same clinical protocol on each scanner (e.g. head, abdomen etc) – minimise built in factors such as beam hardening • Standardise scan parameters – to minimise their effect on image quality and dose 14 AAPM 2004 Dependence on scan settings Image width Noise Scan plane resolution Dose mA kV Focal spot selection Scan time ^ Nominal image width Beam collimation * Detector group width Convolution kernel Pitch ^^ ^^ ^^ Interpolation algorithm ^ If scan time affects no. samples, ^^ In some circumstances , * In almost all cases15 AAPM 2004 5 Standardise scan parameters • Noise and dose dependent on kV • Extent depends on filtration Noise Versus kV 2.00 Hard Beam Scanner A to 120 kV for each scanner Noise normalised Scanner Soft BeamB 1.50 1.00 0.50 80 100 120 140 kV 16 AAPM 2004 Standardise scan parameters • Dose dependent on collimation (beam width) – Different doses (eg 4 mm, 16 mm) – Same image width and noise (eg 1 x 4 mm, 4 x 4 mm) • Scan with same collimation Relative CTDI (4 and 16 slice) 2.4 same image width, same noise relative CTDI 2.2 2.0 1.8 1.6 1.4 1.2 1.0 0.8 0 5 10 15 20 collimation 25 30 3517 AAPM 2004 Comparison of image quality and dose • • • • Issues of comparison Test methods Scanning approach Dealing with the data 18 AAPM 2004 6 Comparison of image quality and dose • Scanner A and Scanner B – Head protocol – Standardised scan parameters • kV, collimation, nominal image width, focal spot, scan time 19 AAPM 2004 Comparison of image quality and dose • Image quality and dose parameters all different Noise Resolution^ A 7.6 HU 5.4 c/cm 4.5 mm 28 mGy B 5.4 HU 5.9 c/cm 5.0 mm 40 mGy AAPM 2004 Image width Dose 20 ^resolution = average 50% and 10% MTF Dealing with the data • Correct for known parameter relationships – ie noise and image width, noise and dose (mAs) noise ∝ 1 image width noise ∝ 1 mAs • Obtain trend data for other relationships – noise and resolution 21 AAPM 2004 7 Correct for known parameter relationships • Correct noise for same image width and dose Noise Resolution Image width A 7.6 HU 5.4 c/cm 4.5 mm Dose 28 mGy B 5.4 HU 5.9 c/cm 5.0 mm 40 mGy 22 AAPM 2004 ^resolution average 50% and 10% MTF Correct for known parameter relationships • Correct noise for same image width and dose Noise Resolution Image width A 7.6 HU 5.4 c/cm 4.5 mm Dose 28 mGy B 5.4 HU 5.9 c/cm 5.0 mm 40 mGy A 6.0 HU 5.4 c/cm 5.0 mm 40 mGy B 5.4 HU 5.9 c/cm 5.0 mm 40 mGy 23 AAPM 2004 ^resolution average 50% and 10% MTF Obtain trend data for other relationships • Noise and resolution both dependent on recon. filter • Scan noise and resolution test objects • Reconstruct with the range of filters – – – – GE LightSpeed: soft, standard, lung, detail, bone, edge Siemens Sensation: AH/AB..10,20,30,40,50,60,70 Toshiba Aquilion: FC 20,21….,FC 10,11…….. Philips Mx8000/Brilliance: A,EB,EC,B,C,D 24 AAPM 2004 8 Noise and resolution with reconstruction filter noise ( %σ ) at 40 mGy and 5 mm image % noise for 40 width mGy 10.0 noise and resolution Siemens S4 images reconstructed with different filters 1.0 0.1 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 Average (MTF50, MTF10) Average of MTF 50 and MTF10 (cycles/cm) 25 AAPM 2004 Noise and resolution with reconstruction filter noise ( %σ ) at 40 mGy Noise at 40 mGy (% SD) and 5 mm image width 10.0 noise and resolution images reconstructed with different filters noise = resolution 1.0 Scanner A Scanner B 0.1 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 Average MTF and10MTF / cm) MeanofMTF , MTF (cycles per cm) 5050 10 (cycles 26 AAPM 2004 Comparison of image quality and dose • Standard resolution, image width and dose Noise Resolution Image width Dose A 7.6 HU 5.4 c/cm 4.5 mm 28 mGy B 5.4 HU 5.9 c/cm 5.0 mm 40 mGy A 7.8 HU 5.9 c/cm 5.0 mm 40 mGy B 5.4 HU 5.9 c/cm 5.0 mm 40 mGy 27 AAPM 2004 8 ^resolution average 50% and 10% MTF 9 Comparison of image quality and dose • Standard noise, resolution and image width Noise Resolution Image width A 7.6 HU 5.4 c/cm 4.5 mm Dose B 5.4 HU 5.9 c/cm 5.0 mm 40 mGy A 5.4 HU 5.9 c/cm 5.0 mm 83 mGy B 5.4 HU 5.9 c/cm 5.0 mm 40 mGy 28 mGy 28 8 ^resolution average 50% and 10% MTF AAPM 2004 Is it always so easy ? 10.0 A B noise ( %σ ) at 40 mGy and 5 mm Noise at image 40 mGywidth (% SD) 5 C D 1.0 sometimes there is no localised trend 0.25 0.1 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 Mean MTF50, MTF10 (cycles per cm) 29 AAPM 2004 Defining the trend 10.0 Siemens power S4 = 2 2.9 R = 0.9942 2 R = 0.9755 2 R = 0.9411 2 R = 0.9972 Philips Mx8000 power = 2.6 powerAquilion = 2.3 Toshiba % noise for 40 mGy noise ( %σ ) at 40 mGy and 5 mm image width 5 noise ∝ power = 2.7 GE LightSpeed Plus 1 dose.imagewidth 1.0 a a noise noise α∝ resolution resolution dose × image width where a is ~ 2.7 for this protocol 0.25 where a is a ~ 2.7 for this protocol 0.1 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 Average (MTF50, MTF10) 30 AAPM 2004 10 Defining the trend • 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 limiting resolution σ ∝ f 1.5 31 AAPM 2004 Can we use a single number ? • Noise, image width, dose and resolution σ∝ 1 zD σ ∝ f a σ∝f zD a • Q factor – Image quality relative to dose Q= f σ 1.5 zD 32 AAPM 2004 Comparison of image quality and dose • Q value Noise Resolution Image width Dose Q A 7.6 HU 5.4 c/cm 4.5 mm 28 mGy 4.7 B 5.4 HU 5.9 c/cm 5.0 mm 40 mGy 5.9 A 7.8 HU 5.9 c/cm 5.0 mm 40 mGy B 5.4 HU 5.9 c/cm 5.0 mm 40 mGy 33 AAPM 2004 ^resolution average 50% and 10% MTF 11 Comparison of image quality and dose • Start with typical clinical protocol • Standardise scan parameters where possible • kV, collimation, nominal image width, focal spot, scan time • Dealing with the data – Correct noise for known parameter relationships • Noise and dose, noise and image width – Obtain trend data for other relationships • Noise and resolution – Use graph as the comparison – Correct values to give single value of noise – Calculate a single quality performance parameter 34 AAPM 2004 Limitations of approach • Numerical, not perceived image quality • Standard image quality parameters – Single number does not represent complete IQ – Full descriptors of image quality – Noise power spectrum – Complete MTF curve – Complete z-sensitivity profile • Helical pitch and reconstruction techniques – Noise and z-sensitivity dependence on pitch not always easy to define – Shape of z-sensitivity profile affected • Artefacts 35 AAPM 2004 Comparison of CT Scanner Image Noise, Image Width, Dose and Spatial Resolution Using Standard Test Methods Sue Edyvean, Nicholas Keat ImPACT (Imaging Performance assessment of CT Scanners) London UK www.impactscan.org ImPACT 36 AAPM 2004 12