CT Protocol Optimisation: Balancing Image Quality and Dose
Transcription
CT Protocol Optimisation: Balancing Image Quality and Dose
CT Protocol Optimisation: Balancing Image Quality and Dose Maria Lewis Guy’s & St.Thomas’ NHS Foundation Trust Overview • • • • • Introduction Challenges of CT protocol optimisation Radiation dose optimisation features Dose audit Approaches to balancing image quality and dose Trends in CT doses • Rapid technical developments and expanding list of applications have led to a dramatic increase in the use of CT Hall & Brenner, BJR 2008 1997/8 80 2008 • CT contribution to collective dose • 1988 ≈ 40% • 2008 ≈ 70% % Dose contribution 70 60 50 40 30 20 10 0 Conventional CT Angiography Interventional UK data, HPA, 2008 Trends in CT doses • In a well-optimised practice* doses have been decreasing Europe 1999 Europe 2004 UK 2003 *Data from Mayo Clinic Rochester - Routine abdomen exam Trends in CT doses From Mahesh, AAPM 2010 - New Technologies for image quality improvement and dose reduction Trends in CT dose The equipment… …and how it is used Optimising dose and image quality • Aim • achieve desired image quality • lowest radiation dose possible • Requirement • fully utilise the capabilities of the equipment Optimising dose and image quality Imaging challenges in CT Patient motion Small structures Low contrast structures Scan time Spatial resolution Image noise Artefacts Radiation dose Protocol design kV Slice width Rota2on 2me Beam width Pitch Flying focal spot Focal spot Recon. interval FBP/ Itera2ve Acquired width Scan FOV mA Recon. kernel Scan Recon. length FOV Protocol design • Spatial detail • Contrast resolution Protocol design: example • X-ray beam width selection 4-slice scanner 16 x 1.25 mm detector banks Protocol for CTPA exam Protocol dessign example • How long does it take to cover 300 mm length? 300 mm Rot. time (s) 1 1 1 Pitch 1 1 1 Beam width (mm) 5 10 20 1.25 2.5 5 60 30 15 Z-axis resolution (mm) Total scan time (s) Protocol design example • Is spatial resolution adequate? 300 mm Rot. time (s) 1 Pitch 1 Beam width (mm) 20 Z-axis resolution (mm) 5 Total scan time (s) 15 Protocol design example • Is spatial resolution adequate? 300 mm Rot. time (s) 1 0.5 Pitch 1 1 Beam width (mm) 5 5 1.25 1.25 60 30 Z-axis resolution (mm) Total scan time (s) Protocol design example • What about dose? Beam width (mm) z-axis GE (%) Relative dose 20 (4 x 5) 97 1.00 10 (4 x 2.5) 83 1.17 67 1.45 5 (4 x 1.25) z-axis 4 x 5 mm 4 x 2.5 mm 4 x 1.25 mm Protocol design example • Helical scanning includes extra rotations at either end of imaged volume:‘over-ranging’ • Additional dose more significant for wider beams particularly for short scan ranges Tzedakis et al, Med. Phys. (2005) 32 (6) Protocol design example • On modern scanners dose from ‘over-ranging’ is reduced with dynamic collimators • Collimator blades open and close asymetrically at start and end of scan Scan range Conventional technology without Dose Shield Courtesy Siemens Medical Systems Scan range SOMATOM Definition AS+ with Adaptive Dose Shield Protocol design example • Any other implications? • Cone beam artefacts with wider beams? • E.g. For head scans it may be better to use 20 mm beam width instead of 40 mm Protocol review and optimisation Focal kV mA/ Scan Recon. Recon. spot Beam kV AEC length interval interval width Protocol review is a FBP/ complex task Focal and Slice FBP/ Slice Acquired Itera2ve Acquired width Itera2ve spot width should undertaken width be width Pitch with caution and as a Flying team Scan Scan Recon. Rota2on Rota2on focal FOV FOV FOV 2me Flying 2me spot Recon. focal FOV spot mA/ Pitch AEC Recon. Recon. kernel kernel Beam Scan width length Optimisation – Working as a team • Team members: • Medical Physicist – technical • CT Radiologist – clinical • CT Technologist - implementation • Cultivate good inter-profession relationships CT Protocol management and review • How team members may work together and in parallel: AAPM Guidelines: J Applied Clin Med Phys. 2013;14(5):3-12 Protocol review & optimisation • Scanner arrives with default protocols – good starting point • Applications training • Protocols adapted to local practice (?) with input from: • Lead CT technologist • CT radiologists (from different specialists) • Medical physicist • Use application training and acceptance to learn about scanner capabilities Protocol review & optimisation • Once in use, review any obvious weaknesses in image quality? • Too noisy, poor contrast, artefacts, high dose… • When making protocol changes MUST understand your scanner • Consult colleagues with same scanner • Review literature and web resources • www.aapm/pubs/CTprotocols • AAPM CT Dose summit meetings • www.aapm.org/meetings/2013CTS/presentations.asp • Perform your own phantom studies if necessary Trends in CT dose The equipment… CT dose optimisation features • Radiation dose control is now a priority for manufacturers • Main dose optimisation features: • • • • • Automatic tube current adjustment (CT AEC) Automatic kV selection Adaptive collimation Organ specific dose modulation Iterative reconstruction AEC in CT • Removes guesswork from manual adjustment for patient size Longitudinally Rotationally high current attenuation Patient size low current AEC in CT • In practice different levels of modulation are usually combined • • • Patient size Longitudinal Rotational Implementation of AEC in CT Implementations of AEC in CT • Reference level of image quality must be set Manufacturer Image Quality setting For reduced dose Philips mAs/slice Decrease mAs/slice Siemens Quality reference mAs Decrease Qual. ref. mAs GE Noise index Increase NI setting (standard deviation) Increase S.D. setting AECS.D.can increase Tube current as well as Attenuation decrease N.I. = 10 the dose! Toshiba 400 350 tube current 300 250 200 150 100 50 N.I. = 15 Automatic kV selection • Traditionally tube potential of 120 kV used • Decreasing tube potential • Increases noise • Increases contrast between high & low z materials 120 kV Automatic kV selection Automatic kV selection Dynamic collimation • Dose from ‘over-ranging’ reduced • Collimator blades open and close asymetrically at start and end of scan Scan range Conventional technology without Dose Shield Courtesy Siemens Medical Systems Scan range SOMATOM Definition AS+ with Adaptive Dose Shield Organ-based tube current modulation • Siemens X-CARE: dose reduction to sensitive anterior organs X-rays OFF 120° Lungren, AJR, 2012 In-plane bismuth shields • Or use bismuth shields? • Use controversial Kim et al Pediatr Radiol 2010; 40:1739 In plane bismuth shields • AAPM statement on use of in-plane bismuth shields AAPM Position Statement on the Use of Bismuth Shielding for the Purpose of Dose Reduction in CT scanning Policy Text: Bismuth shields are easy to use and have been shown to reduce dose to anterior organs in CT scanning. However, there are several disadvantages associated with the use of bismuth shields, especially when used with automatic exposure control or tube current modulation. Other techniques exist that can provide the same level of anterior dose reduction at equivalent or superior image quality that do not have these disadvantages. The AAPM recommends that these alternatives to bismuth shielding be carefully considered, and implemented when possible. www.aapm.org/publicgeneral/BismuthShielding.pdf Iterative reconstruction • Process of repeatedly improving an image by comparison to the measured data From: Iterative reconstruction methods in x-ray CT Beister, Kolditz, Kalender, Physica Medica (2012) 28 Iterative reconstruction • Can offer improvements compared to FBP methods • Noise reduction without degraded resolution • Artefact reduction • Improved spatial resolution • What’s the downside? Measured data • Computational cost: clinically practical? • Change in image texture Compare Update Image 2 n 3 Iterative Reconstruction Techniques, UKRC 2012 Adapted from: Keat, Iterative Reconstruction Techniques, UKRC 2012 Iterative reconstruction • Change in image texture can have significant effect on visualisation of structures (low contrast detectability) • Clinical acceptability of images must be considered S Singh, AAPM 3rd CT Dose Summit, 2013. Iterative Reconstruction: Dose it work to reduce noise... Iterative Reconstruction • Vendor specific implementations Vendor System Approach GE ASIR – Adaptive Statistical Image Reconstruction Statistical method Veo (MBIR) Statistical + Geometric iDose4 Statistical IMR – Iterative Model Reconstruction (WIP) Statistical + Geometric IRIS – Iterative Reconstruction in Image Space Image based Philips Siemens SAFIRE – Sinogram AFfirmed Statistical Iterative Reconstruction Toshiba AIDR – Adaptive Iterative Dose Reduction Statistical Dose audit The equipment… …and how it’s used Dose audit • Collect, for common scans • patient weight, sex • CTDIvol & DLP for each series • Select standard patients • 60 - 80 kg adults • 3.5, 9, 19, 32, 35 ± 15% for children • Compare to • National DRLs • Local DRLs • Scientific literature CT TAP Hospital A NDRL CTDIvol (mGy) DLP (mGy) 12 840 12, 14 940 Courtesy Elly Castellano, RMH Dose audit example • District General Hospital with 2 CT scanners • GE LightSpeed 32 • Siemens Somatom Sensation 64 Background • Initial review of doses by radiologist • Small numbers (8 patients per scanner) • No info on patient size The use of this scanner must be suspended Scan type Routine Abdo-Pelvis DLP (mGy.cm) GE Siemens 1707 753 National DRL (mGy.cm) 560 Method • Protocols • Routine abdo-pelvis: helical, contrast-enhanced Scan type AEC Beam width (mm) Recon slice (mm) Pitch kV Rotation time (s) Noise index/ Qual. ref. mAs Max/min mA Recon filter GE Helical Smart mA 32 x 1.25 (40 mm) 5 0.969 120 0.8 24.6 750/100 Standar d Siemens Helical CAREDose 4D 24 x 1.2 (28.8 mm) 5 1.4 120 0.5 200 - B31f Dose review methodology • From RIS system ~50 consecutive patients selected from each scanner • Patient images reviewed on PACS • Measurements of patient size • Noise values in ROIs • CTDIvol and DLP from dose report Dose review methodology • Patient dimensions – effective diameter • Geometric mean (GM) of AP and lat dimensions • GM = √36.2*28.2 = 32 cm 36.2 cm 28.2 cm Dose review methodology • Measurement of noise: ROI Level 1 • Liver and aorta Dose review results Scanner Patient dimension (cm) CT no: Liver ROI Noise: Liver ROI CTDIvol (mGy) DLP (mGy.cm) GE 28.9 ± 3.7 97± 20 13.8 ± 2.9 14.4 ± 11.0 689 ± 554 Siemens 28.1 ± 3.0 91 ± 15 12.3 ± 2.3 11.5 ± 2.7 552 ± 141 Mean values ± S.D • National DRL = 560 mGy.cm • Mean GE doses ~ 25% higher than Siemens doses for Siemens routine abdo-pelvis protocol Results:Doses CTDI vol versus patient size y = 0.7487x - 9.2314 R2 = 0.7606 Siemens 70.0 CTDIvol (mGy) 60.0 50.0 40.0 30.0 20.0 10.0 0.0 20.0 25.0 30.0 35.0 40.0 Doses for GE routine abdo-pelvis protocol Mean water-equivalent patient diameter (cm) 70.0 CTDIvol (mGy) 60.0 y = 0.2468e0.1354x 2 R = 0.8419 GE 50.0 40.0 30.0 20.0 10.0 0.0 20.0 25.0 30.0 35.0 Mean water-equivalent patient diameter (cm) 40.0 Results • Dose variation with patient size sub-group DLP (mGy.cm) GE 3000 2500 2000 1500 1000 500 0 Siemens DRL 20.1 - 25.0 25.1 - 30.0 30.1 - 35.0 Mean patient dimension (cm) 35.1 - 40.0 Tube current modulation approach 1.6 1.0 0.6 se a e cr 0.4 e d ag str e d on e g d cre ec as re e as e av er im Sl k e 1.2 se crea n i se ng stro ge increa avera ak increase we 0.2 0.0 0.5 1.0 Reference attenuation 1.5 2.0 2.5 3.0 Relative attenuation Image Quality reference tube current 1.4 0.8 a we co Tub ns e tan cu t im rren a g t fo en r ois e 1.8 es Ob Relative tube current 2.0 Dose audit example: Outcome • We suggested setting higher Noise Index values for large patients (‘Large’ protocol) • problems if patient size not assessed correctly • GE recommend controlling dose to large patients by reducing ‘Max’ mA setting • non-uniform IQ along patient • Hospital decided that all large patients scanned on Siemens scanner Balancing image quality and dose • CT dose – how low can we go? • What is diagnostic threshold? • Dependent on diagnostic task Karmazyn B et al, AJR Jan 2009 Balancing image quality and dose Various approaches to determining diagnostic threshold • Gold standard: Blinded human observer studies at different dose levels • Progressive reduction of mAs in small increments • Useful tool: simulation of reduced dose scans by addition of noise Balancing image quality and dose Aquilion 16 120 kV, 200 mAs, 5 mm Scanned dose: 1 Noise = 7.6 HU 0.8 0.3 0.9 0.7 0.4 0.15 0.1 Simulated dose: 0.2 0.5 0.6 Noise = 24 HU Images courtesy Y. Muramatsu, NCC Tokyo Conclusions • Modern scanners are complex • Every scanner is different – essential to understand operation of your scanner • A team approach is essential in CT protocol review • Make use of information resources • Consult: user manuals, manufacturers, websites, scientific literature and colleagues • Dose audit is important and is a good first step in highlighting optimisation issues Thank you for listening [email protected] Developments in technology CT scanner ~1971 … • • • • 4 min rotation time Recon overnight 2 slices/rot (8 mm) 80 x 80 matrix …40 years later • • • • < 0.3 s rotation Real time recon up to 320 slices (0.5 mm) 512 x 512 matrix Implementation of AEC in CT • Implementation manufacturer specific Longitudinal Patient size GE Philips Siemens Toshiba Auto mA DoseRight ACS DoseRight Z-DOM Rotational Smart mA DoseRight D-DOM CARE Dose 4D SUREExposure 3D Iterative loops • Steps 1. 2. 3. 4. 5. Acquire raw data-> sinogram Generate initial image (FBP) Forward project Compare FBP for correction image (to aid convergence) 6. Apply image regularization 7. Update image 8. Repeat 3-7 as necessary 5 6 4 1 7 2 • Iterate until suitable convergence reached Iterative Reconstruction Techniques, UKRC 2012 N Keat, Iterative Reconstruction Techniques, UKRC 2012 3 Highlighting radiation risk • CT – An increasing source of radiation exposure. Brenner & Hall, NEJM, Nov 2007 28th November 2007 Overuse of diagnostic CT scans may cause as many as 3 million excess cancers in the USA over the next two decades, doctors report today... “normal” dose → hair loss, skin burns, cataractogenesis? • CT over-exposure incidents Mad River Incident: repeated scans in single location on baby 2008 Mad River incident •• Class action lawsuits against multiple hospitals and vendors • repeated head scans on child • Media attention & requests for CT • Skin dose > 7 Gy experts/ opinions • 2009 – 2010 brain perfusion overdose incidents Cagnon, CT Protocols, AAPM 2012 • multiple hospitals and vendors • ~ x 8 expected dose (3 – 4 Gy) • Lack of standardisation and/or poor understanding of protocol and equipment capabilities 8 Scan & reconstruction parameters Focal kV mA/ Scan Recon. Recon. spot Beam kV AEC length interval interval width FBP/ Slice FBP/ Focal Slice Acquired Itera2ve Acquired width spot width width Itera2ve width Pitch Flying Scan Scan Recon. Rota2on Rota2on focal FOV FOV FOV 2me Flying 2me spot Recon. focal FOV spot mA/ Pitch AEC Recon. Recon. kernel kernel Beam Scan width length Defining imaging task • Spatial detail • Contrast resolution Defining imaging task • Standard-sized patients • Non-standard patients Case study 2: CT screening clinics • Dose audit at CT screening provider • 13 clinics • 4 scanner models: GE, Siemens, Toshiba • Screening examinations • • • • • abdo/pelvis calcium scoring lung virtual colonoscopy + combinations of the above Case study 2: CT screening clinics • Protocols • Weight-based mAs tables used on GE & Toshiba scanners • CARE Dose 4D used on Siemens scanners Case study 2: CT screening clinics • Dose audit • Only total DLP values for whole exam available • e.g. calcium sore + virtual colonoscopy • Patient weight also documented Case study 2: CT screening clinics • DLPs for virtual colonoscopy + calcium score Mean DLPALL = 689 Mean DLPSTD = 523 Mean DLPALL = 513 Mean DLPSTD = 502 Mean DLPALL = 624 Mean DLPSTD = 593 Mean DLPALL = 668 Mean DLPSTD = 558 Case study 2: CT screening clinics • All sites should adopt automatic exposure control (AEC). • Sites should document the CTDIvol value as well as the DLP. • Sites should document CTDIvol and DLP for each exam series. • CT technologists should receive further training on CT dose issues and in particular on operation of AEC systems. Implementations of AEC in CT • Using AEC ≠ Dose reduction • dose can increase or decrease depending on patient size • dose can increase or decrease depending on reference ‘image quality setting’ AEC in CT • Tips • Centre patient in FOV • Consider order of SPRs - final one usually used in AEC Elliptical phantom 16 x 30 cm SPR AEC mode CTDIvol (mGy) Lat then AP Auto mA 5.7 AP then lat Auto mA 10.1 Courtesy Elly Castellano, RMH Courtesy Siemens AEC in CT • Removes guesswork from manual adjustment for patient size Automatic kV selection Nelson, Optimal kV selection. AAPM 3rd CT Dose Summit 2013 Organ-based tube current modulation • May not always be effective • Another approach: in-plane bismuth shields Iterative reconstruction • Clinicians initially reported ‘waxy’nature to images • Change in noise structure From: Iterative reconstruction methods in x-ray CT Beister, Kolditz, Kalender, Physica Medica (2012) 28, 94-108 Iterative Reconstruction Techniques, UKRC 2012 used tential straint straint –5!d", suggest that the CNR be no less than and the noise be no higher than those obtained at the reference tube potential. With this very tight constraint on image noise, the RDFs at 80 kV were 0.780, 1.005, 1.230, 1.897, and 2.905 for XS, S, M, L, and XL phantoms, respectively. For the requirement of equal noise, there is a dose reduction at 80 kV for only the XS phantom size and there is a big dose increase at 80 kV mA at each kV adjusted to give same CNR for L and XL phantom sizes. When ! = 1.25, the desired imagefor quality at other tube potentials images satisfies twoat conditions: some applications low kV too Automatic kV selection • Relative dose factor (RDF) all • • noisy RDF with no noise constraint (for iodine contrast) 2.5 ge 2 RDF Phantom dimensions (mm): XS = 150 x 150 S = 300 x 200 M = 350 x 250 L = 400 x 300 XL = 480 x 380 Extra Small Small Medium Large Extra Large 1.5 1 0.5 140 0 80 Yu et al. Med Phys. 2010;37(1) 100 120 kV 140 Protocol design example • Choice of settings is scanner dependent • On a 16 - slice scanner can achieve 1.25 mm slices with 20 mm beam width z-axis 16 x 1.25 mm