Shane Foley: IPR, Chicago 2016 Iterative Reconstructions: The Impact on Dose and Image Quality Shane Foley, PhD Radiography and Diagnostic Imaging School of Medicine University College Dublin. Radagrafaíocht agus Iomhánna Fáthmheasach Scoil an Leighis Colaiste na hollscoile, BAC.
LEARNING OUTCOMES To understand the basic physical principles of iterative reconstruction techniques To become familiar with the various types of iterative reconstruction To review the published literature on the effects of IR use on both patient dose and image quality Shane Foley: IPR, Chicago 2016
OUTLINE CT image reconstructions The evolution of iterative reconstruction (hybrid model based) Clinical applications Potential disadvantages Shane Foley: IPR, Chicago 2016
C.T. CONTEXT Use has proliferated >80 million per year in USA 10% per year increase in most countries Radiation dose concerns Esp. for pediatrics Justification, optimisation, dose limitation Shane Foley: IPR, Chicago 2016
C.T. DOSE SURVEYS Large inter hospital discrepancies UK: 10-40 fold 1 Norway: 8-20 fold 2 Ireland: 2-24 fold 3 Local site preferences and protocols largest influence on patient dose Goal of <1mSv imaging! 1. Shrimpton et al (1991). Survey of CT practice in the UK. Part 2: Dosimetric aspects. Chilton, NRPB-R249 (London, TSO) 2. Olerud HM. Analysis of factors influencing patient doses from CT in Norway. Radiation Protection Dosimetry, 71 (2), 123-133 (1997) 3. Foley et al (2012). Establishment of CT diagnostic reference levels in Ireland. British Journal of Radiology.
RADIATION DOSE OPTIMISATION Operator dependent: Patient positioning / Protocol selection / Range limitation Technological features: 1994: AEC 2007: Dynamic beam collimators 2010: Tube voltage modulation 1999: ECG pulsing 2009: Iterative reconstruction 2010: Model based IR
C.T. IMAGE RECONSTRUCTION Filtered Back Projection (FBP) Used since CT was first developed Still primary method in use Main advantage = simple & quick Shane Foley: IPR, Chicago 2016
FILTERED BACK PROJECTION (FBP) Reconstructs an image by: a) Obtaining projection profiles b) Filtering each view Removes blurring seen in simple back projection Different filters for different tasks (sharp v smooth) c) Back projecting views Shane Foley: IPR, Chicago 2016
LIMITATIONS OF F.B.P. Any noise in raw data is propagated in image Traditionally countered by increasing exposure / thicker slices Higher spatial resolution comes with higher noise Based on many assumptions: focus spot is infinitely small that every detector element is small pencil beam without polychromatic spectrum. Each voxel has no shape or size Shane Foley: IPR, Chicago 2016
NOISE IN C.T. = principal limiting feature for low dose CT Degrades image quality: LCD But in CT: Quantum noise (detection of discrete photons) can be modelled by Poisson process Noise from other parts of the imaging chain can be statistically modelled Mathematic methods can be used to determine the most likely true signal from noisy projection data Edyvean S (2005). Impactscan.org
I.R. TECHNIQUES Not new available for 20+ years Used in SPECT and PET Introduction to CT slow - extensive computer power needed I.R. allows decoupling of spatial resolution & noise Correction loop introduced to reduce noise Iteration = the act of repeating a process Shane Foley: IPR, Chicago 2016
I.R. BASICS 1. Starts with an estimate of truth! (often FBP image) 2. Compares initial projection to a model projection 3. Initial projection is updated and process repeated (multiple iterations) until differences become acceptably small 4. By modelling noise can generate images with lower noise IR: does NOT decrease radiation dose Allows lower exposure settings to produce equivalent noise OR improved image quality Shane Foley: IPR, Chicago 2016
I.R. TECHNIQUES McCollough et al (2012) Radiology. Aug; 264(2): 567 580
I.R. BENEFITS Reduce noise within CT images Improved CNRs, or Reduce artefacts Beam hardening / Metallic implants / scatter Improved spatial resolution Avoiding data filtration before back projection Shane Foley: IPR, Chicago 2016
I.R. OPTIONS Full (pure) IR Physical properties of the acquisition system taken into account Computationally very expensive: reconstruction = cumbersome & time consuming Hybrid IR modified (faster) IR technique Blends FBP with IR, performing majority of noise reduction in image space Does not improve spatial resolution Does not model the optics Shane Foley: IPR, Chicago 2016
I.R. TECHNQIUES Currently three different options IR performed using: Image (slice) data Projection (raw) & image data Projection data only Image space: noise reduction primarily performed by statistical process Projection space: artefact reduction and increased spatial resolution Shane Foley: IPR, Chicago 2016
I.R: IMAGE DATA Eg: IRIS Iterative reconstruction in image space (Siemens) Process: Raw data first reconstructed using FBP Then forward projected with multiple iterations according to modelling of the noise data Pro: Recon time only slightly longer than FBP Con: assumes ideal system (LCD / streak artefacts not sign. improved) 1 1. Bittencourt MS, Schmidt B, Seltmann M et al. Iterative reconstruction in image space (IRIS) in cardiac computed tomography: initial experience. Int J Cardiovasc Imaging 2011;27:1081-7
I.R: PROJECTION & IMAGE DATA Eg: ASIR / AIDR 3D / SAFIRE / idose 4 Process: Projection data first reconstructed with FBP Compared with ideal noise model based on statistics (accounting for photon and electronic noise) Multiple iterations performed that compare each updated voxel with the ideal noise model until the algorithm converges Pro: recon times only slightly longer than FBP Con: assumes ideal system / limited reduction of streak artefact and spatial resolution improvement 1 1. Katsura M, Sato J, Akahane M et al. Comparison of pure and hybrid iterative reconstruction techniques with conventional filtered back projection: image quality assessment in the cervicothoracic region. Eur J Radiol. 2013;82:356-60
I.R: PROJECTION DATA ONLY Eg: MBIR / FIRST / IMR Models entire x-ray beam and system optics Pro: less noise 1 / improved spatial resolution and low contrast detectability, fewer streak artefacts 2 Con: Recon time significantly longer than FBP / Different look & feel to FBP, users need to adapt 1. Katsura M, Matsuda I, Akahane M et al. Model-based iterative reconstruction technique for radiation dose reduction in chest CT: comparison with the adaptive statistical iterative reconstruction technique. Eur Radiol. 2012;22:1613-23. 2. Thibault JB, Sauer KD, Bouman CA et al. A three dimensional statistical approach to improved image quality for multislice helical CT. Med Phys 2007;34:4526 4544
MODEL BASED I.R. Includes details of the whole CT system & optics Physics modelling (scatter / cross talk) Focal spot size Detector element size The patient System geometry Beam energy Cone angle Electronic noise from the system Goal is to improve noise & spatial resolution at the same time Shane Foley: IPR, Chicago 2016
WHAT THE MANUFACTURERS SAY I.R. DOES Siemens: IRIS and SAFIRE* allows for up to 60% radiation dose reduction in routine clinical use. 1 Philips: 60-80% lower dose with 43-80% increased low contrast detectability and 70-83% lower noise 2 Toshiba: >75% dose reduction 3 GE: <1mSv scanning with higher resolution 4 1. Siemens Guide to Low dose. Available at http://www.siemens.com.au/files/healthcare/education/lowdose/hc_guidetolowdose.pdf 2. Philips (2016). Iterative Model based reconstruction. Available at http://www.usa.philips.com/healthcare/product/hcnctd449/imr-reconstruction-technology 3. Toshiba (2014). White paper on AIDR 3D 4. GE (2015). Veo product
WHAT THE LITERATURE SAYS I.R. DOES 45% dose reduction for trunk CT Less noise (better lesion detection) & better spatial resolution than ASIR (30%) Ideally suited to pediatrics (better lesion detection & delineation) 92% dose reduction (depending on clinical indication) Improved visibility of small structures on coronal plane 1. Smith et al (2014). Model-based Iterative Reconstruction: Effect on Patient Radiation Dose and Image Quality in Pediatric Body CT. Radiology270: Number 2 February 2014 2. Miéville et al (2013). Model-based iterative reconstruction in pediatric chest CT: assessment of image quality in a prospective study of children with cystic fibrosis. Pediatr Radiol 43:558 567
CLINICAL APPLICATIONS: CHEST Most widely investigated body part for pediatric IR High inherent contrast & low attenuation enable tolerance of image noise Studies to date all agree: IR allows dose reductions (31-92%) with improved image quality Dose reductions esp. useful due to radiosensitive organs in chest 1. Miéville et al. Paediatric cardiac CT examinations: impact of the iterative reconstruction method ASIR on image quality preliminary findings. Pediatr Radiol 2011; 41:1154 1164 2. Miéville et al (2013). Model-based iterative reconstruction in pediatric chest CT: assessment of image quality in a prospective study of children with cystic fibrosis. Pediatr Radiol 43:558 567
Padole et al (2015). 2015;AJR 204: W384-W392. FBP: 7mGy SafeCT: 1.6mGy ASIR: 1.6mGy MBIR: 1.6mGy
den Harder et al (2015). AJR 204: 645-653. CHEST: 11yo 80kVp / 70mAs / 38 DLP FBP: Noise 113 CNR 0.6 idose4: Noise 79 CNR 0.7 IMR 3: Noise 15 CNR 3.0, - Streak artefact
CHEST Smoothing effect of IR can potentially result in small structures not being visible Studies to date reported MBIR improved visualisation of small structures (lung fissures & small vessels) 1,2 Further studies needed 1. Koc et al (2014). Computed tomography depiction of small pediatric vessels with model-based iterative reconstruction. Pediatr Radiol 2014; 44:787 794 [ 2. Miéville et al (2013). Model-based iterative reconstruction in pediatric chest CT: assessment of image quality in a prospective study of children with cystic fibrosis. Pediatr Radiol 43:558 567
Han et al (2011). JCCT May CARDIAC CT: 80 kv, DLP 12 SAFIRE: reduced noise 35%, improved SNR/CNR 50%
1. Vorona et al (2011). Reducing abdominal CT radiation dose with the adaptive statistical iterative reconstruction technique in children: a feasibility study. Pediatr Radiol 2011; 41:1174 1182 2. Singh et al (2012). Radiation Dose Reduction with Hybrid Iterative Reconstruction for Pediatric CT. Radiology 12 263 3. Young et al (2016). Pediatr Radiol (2016) 46:303 315 ABDOMEN More challenging due to lower organ contrast Studies show varying % dose reductions according to vendor and application ASIR: 33% dose reduction (40% ASIR) while maintaining IQ 1 38-46% dose reduction 2 MBIR: 76% reduction vs FBP 3
Padole et al (2015). 2015;AJR 204: W384-W392. ABDOMEN: 7y.o. weight loss, diarrhea FBP: 4.3mGy FBP: 1.3mGy ASIR : 1.3mGy MBIR: 1.3mGy
LESION DETECTION (14 yo lymphoma) MBIR images were rated as superior to 100% ASIR images MBIR 100% ASIR FBP 1. Smith et al (2014). Model-based Iterative Reconstruction: Effect on Patient Radiation Dose and Image Quality in Pediatric Body CT. Radiology270: Number 2 February 2014
HEAD C.T. Most data from adult studies Improved SNR, CNR in ASIR vs FBP ASIR: 30% dose reductions vs FBP with similar IQ + lesion conspicuity Intracranial haemorrhage better seen on SAFIRE at reduced dose vs FBP 0.9mSv MBIR: improves IQ vs ASIR with dose reduction Paediatric studies: ASIR: 28% dose reduction for 3- to 12-year-old patients and 48% in reduction at 30 mgy >12 years vs FBP 1 MBIR: 50% dose reduction with similar IQ vs ASIR 2 1. McKnight et al Pediatr Radiol. 2014 Aug;44(8):997-1003 2. Smith et al (2014). Model-based Iterative Reconstruction: Effect on Patient Radiation Dose and Image Quality in Pediatric Body CT. Radiology270: Number 2 February 2014
den Harder et al (2015). AJR 204: 645-653. HEAD C.T: 8yo 120kV, 198mAs, DLP 527 Better grey white matter discrimination with IMR FBP: Noise 7.7 CNR 0.3 IMR 1: Noise 3.2 CNR 3.4 IMR 2: Noise 2.8 CNR 3.9 IMR 3: Noise 2.4 CNR 4.5
SPINAL C.T. No studies to date on paediatrics Adults: 40% dose reductions (Safire) with better image quality for IV disks, neural foramina & ligaments BUT IQ for non spinal soft tissues declined slightly Need more studies Shane Foley: IPR, Chicago 2016
IMPACT ON IMAGE QUALITY Noise Spatial resolution (MTF Smith et al Radiology) Texture Artefacts Shane Foley: IPR, Chicago 2016
IMAGE QUALITY: NOISE Obvious reduction in image noise w/o affecting signal BUT conventional relationship between noise and dose may not be valid anymore. Noise is much less sensitive to the dose change. 50% dose reduction results in only 15-19% increase in noise. (vs 40% in FBP) 1 This relationship needs re-assessment under different dose levels 1. Dong (2014). AAPM Shane Foley: IPR, Chicago 2016
IMAGE QUALITY: SPATIAL RESOLUTION 1. Smith et al (2014). Model-based Iterative Reconstruction: Effect on Patient Radiation Dose and Image Quality in Pediatric Body CT. Radiology270: Number 2 February 2014
IMAGE QUALITY: TEXTURE MBIR: unusually smooth Less conspicuous on MPR images Radiologists need to adapt Blotchy / pixillated appearance reported in adults NOT paediatrics Lack of data / radiologists higher acceptance of noise/artifacts in paeds Earlier versions of IR algorithms Shane Foley: IPR, Chicago 2016
ARTEFACTS IR: potential to reduce beam hardening & photon starvation MBIR: subtle staircase effect on bony interfaces 1. Deak et al (2013). Radiology 266 1
ARTEFACTS MBIR: skin surfaces Small blacked out pixels on axial? Clinically significant 1. Deak et al (2013). Radiology 266 1
RING ARTEFACTS Wagner-Bartak et al (2013). CT model based iterative reconstruction technique: Pearls, Pitfalls and Practical solutions. RSNA
RING ARTEFACTS Ali Khawaja et al (2015). European Journal of Radiology, Volume 84, Issue 1, 2015, 2 10
BREAST SHIELD ARTEFACT Some evidence that shields may cause further artefacts Wagner-Bartak et al (2013). CT model based iterative reconstruction technique: Pearls, Pitfalls and Practical solutions. RSNA
DISADVANTAGES I.R. Computationally time consuming Expensive Compatible with only latest scanners? Need for vendor neutral options: e.g. MedicVision, Clarity, Saphenia Vendor specific algorithms may not be applicable to those of others Artefacts Shane Foley: IPR, Chicago 2016
1. Available from http://www.usa.philips.com/healthcare/product/hcnctd449/imr-reconstruction-technology 2. Jensen et al (2013). Model-based Iterative Reconstruction (MBIR) Implementation: Process Description and Lessons Learned. RSNA 3. Vardhanabhuti et al (2013). Image quality assessment of standard and low dose CT using FBP, ASIR and MBIR, AJR 200, 545-552 TIME CONSUMING Pure IR much longer FBP: 26 images per second (ips) SAFIRE: 20 ips idose4: 18 ips VEO: 0.09 ips (35-40mins per abdomen/pelvis) Mieville (2013): 30-60 mins chest CT depending on patient size & DFOV? Suitable for emergency dept patients or in patients?
SUMMARY Excellent potential for dose reduction Standard IR: reduces noise + some artefacts Hybrid IR: significant dose reductions in acceptable time Model based IR: potential for further dose reductions with improvements in image quality Still testing limits and practical opportunities Shane Foley: IPR, Chicago 2016
Questions.? shane.foley@ucd.ie shane.foley@ucd.ie
REFERENCES den Harder et al. (2015) Hybrid and Model-Based Iterative Reconstruction Techniques for Pediatric CT. AJR; 204:645 653 Karmazyn et al (2013). Optimization of Hybrid Iterative Reconstruction Level in Pediatric Body CT. AJR 2014; 202:426 431 Mehta et al. (2013). ITERATIVE MODEL RECONSTRUCTION: SIMULTANEOUSLY LOWERED COMPUTED TOMOGRAPHY RADIATION DOSE AND IMPROVED IMAGE QUALITY. MEDICAL PHYSICS INTERNATIONAL Journal Miéville et al (2013). Model-based iterative reconstruction in pediatric chest CT: assessment of image quality in a prospective study of children with cystic fibrosis. Pediatr Radiol 43:558 567 Liu (2014). Model based iterative reconstruction: A promising algorithm for Todays Computed Tomography Imaging. Journal of Medical Imaging Sciences 45: 131-136 Padole et al (2015). CT Radiation Dose and Iterative Reconstruction Techniques. AJR 204 4 384-392 Smith et al (2013). Model-based iterative reconstruction: Effect on Patient Radiation Dose and Image Quality in Pediatric Body CT. Radiology 270 2 526-534. Tricarico et al (2013). Cardiovascular CT angiography in neonates and children. Image quality and potential for radiation dose reduction with iterative image reconstruction techniques. Eur Radiol 23, 1306-1315 Wagner-Bartak et al (2013). CT model based iterative reconstruction technique: Pearls, Pitfalls and Practical solutions. RSNA Young et al (2016) Knowledge-based iterative model reconstruction: comparative image quality and radiation dose with a pediatric computed tomography phantom. Pediatr Radiol 46:303 315 Shane Foley: IPR, Chicago 2016