Ehsan Samei, PhD. Duke University

Transcription

1

2 Ehsan Samei, PhD Duke University

3 Duke CIPG Clinical Imaging Physics Group

4 Medical Physics 2.0? A bold vision for an existential transition of clinical imaging physics 12 hrs didactic lectures on CT, MRI, NM, Fluoroscopy, Rad, Mammo, US, IT RSNA 2013 RSNA 2014 RSNA 2015 AAPM 2014 AAPM 2015 Clinical Imaging Physics. Wiley and Sons, 2015

5 Overarching presuppositions Imaging should render relevant state of health/disease accurately and precisely enough so that it can lead to definitive clinical decisions Images should be more reflective of the state of the patient than the technique used

6 Reality checks Clinical activities are heterogeneous/compounded/complex operations Increased scrutiny Limited resources How imaging physics can add value in the quality of imaging operation?

7 4 Key formative questions

8 1. Where is imaging in medicine enterprise? Transformative technology The face of modern medicine

9 2. Where medical imaging is going? Evidence-based medicine Practice informed by science Precision medicine Quantification and personalization of care Value-based medicine Scrutiny on safety, performance, consistency, stewardship, ethics Comparative effectiveness and meaningful use Enhanced focus on actual utility

10 3. What has been the role of imaging physics in medicine? Remember Roentgen! Medical physics is the foundational discipline behind Radiology and Radiation Oncology

11 4. What has been the clinical role of imaging physics? Ensuing quality and safety of clinical imaging systems We have done a GREAT job using engineering and physics concepts to Design systems with superior performance Ensure minimum intrinsic performance Claim compliance But

12 Imaging physics enterprise Education Research Clinical Practice

13 Imaging physics enterprise Education Clinical Practice Research

14 Why 1.0 is not enough Not translatable, historically, to clinical care, mostly equipment-focused Relevance? Outdated science and technology Compliance lags behind clinical needs and innovations

15 Why 1.0 is not enough Clinical performance? Optimization of use? Consistency of quality? Changing technology? Value-based healthcare?

16 1.0 to 2.0 Clinical imaging physics extending from compliance to excellence intrinsic to extrinsic Equipment to operation Specifications to performance Quality to consistency Assumption to actual utility Promises to implementation

17 Inspiring practices of Medical Physics Science: Practices that are scientifically-informed by latest findings and methods 2. Relevance: Objectives that are oriented towards the actual clinical practice 3. Stewardship: Approaches that are pragmatic in the meaningful and efficient use of resources 4. Integration: Solutions that are comprehensive and holistic in view of clinical operation 5. Empowerment: Education enabling others to be their best in their domain of care

18 Clinical role of medical physics Applying the science of imaging physics to ensure quality, safety, consistency, and optimum utilization in the clinical practice of medical imaging Ensuring quality, safety, consistency, optimality 18

19 Ensuring quality, safety, consistency, optimality Prospective performance assessment Quality by inference Prospective protocol definition Quality by prescription Retrospective quality assessment Quality by outcome

20 Ensuring quality, safety, consistency, optimality Prospective performance assessment Quality by inference Devising better metrics Automated characterization Prospective protocol definition Quality by prescription Retrospective quality assessment Quality by outcome

21 Nuclear Medicine non-uniformity Crosshatched artifact pattern (system has square PMT s)

22 Nuclear Medicine non-uniformity Integral Uniformity = N max N min N max + N min 100% 2.87% 2.81% Action limit = > 5.00%

23 Nelson JS, et al, JNM, 2014 Structured Noise Index (SNI) Eye Filter FT NPS S Subtract Quantum Noise Filtered NPS S Artifact Image Eye Filter Flood Image NPS T Filtered NPS T

24 Flood Image Filtered NPS Artifact Image

25 Is SNI valid? Observer study: 55 images, 5 observers, 2 settings Estimated Integral UFOV Estimated Integral CFOV Sensitivity 62% 54% 100% Specificity 90% 83% 95% PPV 67% 50% 87% NPV 88% 85% 100% Accuracy 84% 76% 96% R Structured Noise Index

26 Integral Uniformity vs SNI CFOV IU = 2.81% Action limit = > 5.00% SNI= 0.78 Action limit = > 0.5

27 SNI Tracking and Analytics Scanners can send QC data to physics server Physics Server

28 SNI Tracking and Analytics Scanners can send QC data to physics server Physics Server

29 Improved consistency across NM fleet 29

30 SNI Tracking and Analytics Improved consistency across systems Decreased the time technologists and physicists spent analyzing QC Successfully alerted staff to issues which would have been overlooked Expedited communication between clinical staff, physicist, and clinical engineers 30

31 What is image quality? ICRU Report 54, 1996 Any general definition of image quality must address the effectiveness with which the image can be used for its intended task.

32 CNR vs observer performance Christianson, et al, Radiology, 2015

33 Parameters that affect taskbased image quality 1. Contrast 2. Lesion size 3. Lesion shape 4. Lesion edge 5. Resolution 6. Viewing distance 7. Display 8. Noise magnitude 9. Noise texture 10. Operator noise Feature of interest Image details Distractors

34 Parameters that affect taskbased image quality 1. Contrast 2. Lesion size 3. Lesion shape 4. Lesion edge 5. Resolution 6. Viewing distance 7. Display 8. Noise magnitude 9. Noise texture 10. Operator noise Feature of interest Image details Distractors CNR 1. Contrast 2. Lesion size 3. Lesion shape 4. Lesion edge 5. Resolution 6. Viewing distance 7. Display 8. Noise magnitude 9. Noise texture 10. Operator noise

35 Iterative Reconstruction FBP Reconstruction Noise texture? IR FBP Images courtesy of Dr de Mey and Dr Nieboer, UZ Brussel, Belgium Low Dose 114 DLP, 1.9 msv

36 Noise Power Spectrum

37 Noise Power Spectrum

38 Resolution and noise, eg 1 Comparable resolution Lower noise but different texture MTF FBP -IRIS -SAFIRE NPS x FBP -IRIS -SAFIRE Spatial frequency Spatial frequency

39 Resolution and noise, eg 2 Higher resolution Lower noise but different texture x MBIR -ASIR -FBP MTF MBIR -ASIR -FBP NPS Spatial frequency Spatial frequency

40 Task-based quality index Resolution and contrast transfer Attributes of image feature of interest Image noise magnitude and texture ( ' d ) 2 NPWE = MTF 2 2 (u,v)w Task [ MTF 2 (u,v)w 2 Task (u,v) E 2 (u,v)dudv] 2 (u,v) NPS(u,v)E 4 (u,v) + MTF 2 2 (u,v)w Task (u,v)n i dudv Richard, and E. Samei, Quantitative breast tomosynthesis: from detectability to estimability. Med Phys, 37(12), (2010). Chen et al., Relevance of MTF and NPS in quantitative CT: towards developing a predictable model of quantitative... SPIE2012

41 CNR vs observer performance Christianson, et al, Radiology, 2015

42 d vs observer performance Christianson, et al, Radiology, 2015

43 Simple Metrics Fourier-Based Image Based Detection Accuracy CNR R 2 = E = 0.3 CI 9 5 % = 5.60e-03 Humans vs. Models R 2 = NPW E = 0.25 CI 9 5 % = 3.32e CHO R 2 = E = 0.4 CI 95 % = 4.25e A CNR A NPW A CHO CNRa NPWE CHOi Detection Accuracy R 2 = E = 0.15 CI 9 5 % = 1.20e R 2 = 0.77 E = 0.3 CI 9 5 % = 5.83e R 2 = 0.72 E = 0.45 CI 95 % = 4.59e A CNRa A NPWE A CHOi

44 Task-based measurements Mercury Phantom 3.0 Diameters matching population cohorts Depths consistent with cone angles Designed for size, AEC, and d evaluations 44

45 Design: Resolution, HU, noise Representation of abnormality-relevant HUs Sizes large enough for resolution sampling Maximum margin for individual assessment Iso-radius resolution properties Matching uniform section for noise assessment 45 Wilson et al, MedPhys 2012

46 imquest (image quality evaluation software) HU, Contrast, Noise, CNR, MTF, NPS, and d per patient size, ma modulation profile Wilson et al, MedPhys 2012

47 Detectability index across systems 12 Detectability 10 8 All Systems System 1 System 2 System 3 System 4 System 5 System 6 System 7 System 8 System 9 System 10 System 11 System 12 System Date Intra system variability: 1-4% Inter system variability: 6% Winslow et al, AAPM, 2014

48 Image quality in the presence of anatomical heterogeneity 48

49 Truth

50 Truth System A System B System C System D System E

51 Ensuring quality, safety, consistency, optimality Prospective performance assessment Quality by inference Protocol optimization Protocol consistency Prospective protocol definition Quality by prescription Retrospective quality assessment Quality by outcome

52 Optimization Quality indices (d, A z, ) framework Benefit regime Different patients and indications Optimization regime Scan factors (mas, kvp, pitch, recon, kernel, ) Risk regime Safety indices (organ dose, eff. dose, )

53 Optimization of kvp-recon 1 Feature with no iodine 1 Feature with iodine A Z A Z Eff dose (msv) Eff dose (msv)

54 Patient size? 54 for-your-child.blogspot.com

55 Optimization of dose per size Iso-gradient optimization points to define ALARA Size

56 Optimization for quantification Quantitative CT volumetry PRC: Relative difference between any two repeated quantifications of a nodule with 95% confidence

57 Texture similarity Sharpness Solomon, Samei, Med Phys, 2012

58 Texture similarity matched recons GE Siemens Minimum RMSD (mm 2 ) Minimum PFD (mm - 1 ) SOFT B35f STANDARD B43f CHEST B41f DETAIL B46f LUNG B80f BONE B75f BONE+ B75f EDGE B75f Solomon, Samei, Med Phys, 2012

59 Texture similarity matched recons GE Siemens Minimum RMSD (mm 2 ) Minimum PFD (mm - 1 ) SOFT B35f Standard B43f ~50 mm STANDARD B43f CHEST B41f DETAIL B46f STANDARD B43f LUNG B80f BONE B75f BONE+ B75f EDGE B75f nnps f (mm 2 ) Spatial Frequency (mm 1 ) Solomon, Samei, Med Phys, 2012

60 Texture similarity matched recons GE Siemens Minimum RMSD (mm 2 ) Minimum PFD (mm - 1 ) SOFT B35f Lung B80f ~50 mm STANDARD B43f CHEST B41f DETAIL B46f LUNG B80f LUNG B80f BONE B75f BONE+ B75f EDGE B75f nnps f (mm 2 ) Spatial Frequency (mm 1 ) Solomon, Samei, Med Phys, 2012

61 Resolution/noise across recons

62 Resolution/noise across recons

63 Ensuring quality, safety, consistency, optimality Prospective performance assessment Quality by inference Dose monitoring - analytics Quality tracking - analytics Prospective protocol definition Quality by prescription Retrospective quality assessment Quality by outcome

64 Dose monitoring components A. Access: Connection and collection of doserelevant data B. Integrity: Data quality and accuracy C. Metrology: Meaningful quantities to monitor D. Analytics: From data to knowledge E. Informatics: Dose monitoring as a secure, integrated solution

65 Dose analytics 1. Benchmarking institution against national DRLs 2. Defining protocol- and size-specific DRLs 3. Identifying outliers 4. Ascertaining trends over time 5. Ascertaining inter-system variability 6. Tracking protocol discrepancy 7. Investigating individual doses 8. Improving operational consistency Frush, Samei, Medscape Radiology, in print 2015

66 Benchmarking institution against national DRLs CTDI Institution DIR

67 Size (cm) Upper Lower

68

69 Dose monitoring in mammo

70 Chest exposures by operator and Clinic CR DR

71 CT Table Height 71

72 PE Chest Protocol CTDIvol (mgy) VCT - Old SOMATOM Definition Effective Diameter (mgy)

73 PE Chest Protocol CTDIvol (mgy) VCT - Old SOMATOM Definition VCT - New Effective Diameter (mgy)

74 Clinical image quality?

75 Clinical image quality monitoring Laplacian Pyramid Decomposition 1. Lung grey level 2. Lung detail 3. Lung noise 4. Rib-lung contrast 5. Rib sharpness 6. Mediastinum detail 7. Mediastinum noise 8. Mediastinum alignment Thorax linear landmark detection 9. Subdiaphragm-lung contrast 10. Subdiaphragm area Lin et al, Med Phys 2012 Samei et al, Med Phys 2014

76 Mediastinum noise

77 Image quality reference levels lung noise Quality Acceptable Poor

78 Lung detail

79 Average Lung Noise Acceptable Quality Average Lung Detail

80 0.45 Average Subdiaphragm Area AJ19 AK151 AM235 BARWI003 BPS12 DAVIS090 DG97 DW77 ED108 HITE0001 JACOB074 KCG24 KMD48 KS324 LAM41 LB111 LH154 MALES003 MCG MH225 MOORE237 MW187 NA RICKM005 SC134 TA34 TE45 TML113 VJ28

81 Overall operational consistency - CT Abdomen Pelvis Exams (n=2358) 50 Scanner 1 50 Scanner 1 45 Scanner 2 45 Scanner Scanner Scanner 3 SSDE (mgy) Noise (HU) Effective Diameter (cm) Effective Diameter (cm) Christianson et al, AAPM, 2014

82 Protocol Report Cards Protocol Definition Image Quality (Prospective) Radiation Dose Protocol name 23_ST ANDA RD_CHEST _WIT HOU T _I V 22_FLASH_PE_CHEST5.1 STANDARD CHEST5.1 STANDARD CHEST5.1 STANDARD CHEST STANDARD CHEST Scan 5 type mm Helical Helical Helical Helical Helical n = 4229 annually Rotation time Scan parameter Definition Flash Pitch 750 HD VCT 0.8 LS Protocol name NDARD_CHE ST _WIT STANDARD 23_STA HOUT CHEST5.1 CHEST _IV22_FLASH_PE_CHEST5.1 STANDARD CHEST5.1 Table speed Helical STANDARD 61.4 Helical Scan type Helical Helical Helical 0.4 Beam width Rotation time Pitch kvp Table speed Beam width Slice thickness kvp Auto ma On On On On On On Slice thickness Auto ma On On On On ma/ni/masref ma/ni/masref SFOV Wedge_3 Wedge_2 Body Body SFOV Standard Body Wedge_3 Standard Wedge_2 Body Body Body Kernel B31f B31f Standard Iterative Level FBP FBP FBP ASiR 30% Kernel FBP B31f B31f Standard Standard Standard Iterative Yes Level FBP FBP ASiR 30% FBP FBP M atches e-protoco l Yes No Yes Yes % of STANDARD CHEST exams 14.3% 27.4% 47.9% 5.3% 5.2% Kernel/IR NPSavg MTF50 d' (5mm) 5% Match M atches e-protocol Yes No Yes Yes Yes Target (CT 750HD) STD/40% N/A % of STANDARD CHEST exams 14.3% 27.4% 47.9% 5.3% 5.2% Current Flash B31/ No Current VCT STD/ No Current LS16 STD/ No Recommended Flash i40/ Yes Recommended VCT STD/ No Recommended LS16 STD/ No To match noise magnitude of 750HD, increase ref mas to 155 If STD behaves on VCT as on 750HD,, decrease VCT NI to 11.8 CTDIvol (mgy) cm (27%) 30-35cm (51%) 35-40cm (21%) Protocol Definition Image Quality (Prospective) 40-45cm (1%) Scan parameter Definition Flash 750 HD VCT LS Recommended Flash 45 i40/ Yes Definit ion 40 Recommended Flash 35 VCT STD/ No Definit ion 750 HD 30 Recommended 25 LS16 VCT STD/0Fla sh No HD LS To match noise magnitude of 750HD, VCT increase ref mas to DIR 75th % 5 LS 16 If STD DIR 50th% behaves on VCT as on 750HD,, decrease VCT NI to DIR 25th% Patient diameter (cm) CTDIvol (mgy) STANDARD CHEST 5 mm n = 4229 annually Kernel/IR NPSavg MTF50 d' (5mm) 5% Match Target (CT 750HD) STD/40% N/A Current Flash B31/ No Current VCT STD/ No Current LS16 STD/ No Noise (Retrospective) Global Noise Level cm (27%) 30-35cm (51%) 35-40cm (21%) 40-45cm (1%) Definit ion Fla sh 750 HD VCT LS Global Noise Level Patient diameter (cm) Definit ion Fla sh 750 HD VCT LS 16 Definit ion

83 Protocol Report Cards Image Quality (Prospective) Recommended LS16 STD/ No To match noise magnitude of 750HD, increase ref mas to 155 If STD behaves on VCT as on 750HD,, decrease VCT NI to 11.8 Protocol Definition Image Quality (Prospective) Radiation Dose STANDARD CHEST 5 mm 30 n = 4229 annually Scan parameter Definition Flash 750 HD VCT LS 16 Protocol name 23_STA NDARD_CHE ST _WIT HOUT _IV22_FLASH_PE_CHEST STANDARD CHEST5.1 STANDARD CHEST5.1 STANDARD CHEST Scan type Helical Helical Helical Helical Helical Rotation time Pitch Table speed Beam width kvp Slice thickness Auto ma On On On On On ma/ni/masref SFOV Wedge_3 Wedge_2 Body Body Body 10 Kernel B31f B31f Standard Standard Standard Iterative Level FBP FBP ASiR 30% FBP FBP M atches e-protoco l Yes No Yes Yes Yes % of STANDARD CHEST exams 14.3% 27.4% % 5.3% 5.2% Kernel/IR NPSavg MTF50 d' (5mm) 5% Match Target (CT 750HD) STD/40% N/A Current Flash B31/ cm cm 35-40cm No 40-45cm Current VCT STD/ (27%) 35.6 (51%) (21%) No (1%) Current LS16 STD/ No Recommended Flash i40/ Yes Recommended VCT STD/ No Recommended LS16 STD/ No To match noise magnitude of 750HD, increase ref mas to 155 If STD behaves on VCT as on 750HD,, decrease VCT NI to CTDIvol (mgy) cm (27%) 30-35cm (51%) 35-40cm (21%) Radiation Dose Noise (Retrospective) 40-45cm (1%) CTDIvol (mgy) Global Noise Level Definit ion Flash 750 HD VCT LS 16 DIR 75th % DIR 50th% DIR 25th% CTDIvol (mgy) cm (27%) Patient diameter (cm) 30-35cm (51%) 35-40cm (21%) Definit ion Fla sh 750 HD VCT LS cm (1%) Definit ion Flash 750 HD VCT LS 16 DIR 75th % DIR 50th% DIR 25th% Definition Flash 750 HD VCT LS 16 Global Noise Level CTDIvol (mgy) Patient diameter (cm) Patient diameter (cm) Definition Flash 750 HD VCT LS 16 Definition Flash 750 HD VCT LS 16 Noise (Retrospective) Global Noise Level cm (27%) 30-35cm (51%) 35-40cm (21%) 40-45cm (1%) Definit ion Fla sh 750 HD VCT LS 16 Global Noise Level Patient diameter (cm) Definit ion Fla sh 750 HD VCT LS 16 83

84 Challenges on the path to MP2.0 Cultural inertia Availability of effective tools Availability of QA informatics Lagging guidelines, accreditations, regulations Lagging certification requirements Lagging education Effective model(s) of practice

85 Conclusions 1 Clinical Imaging physics is a severely untapped resource insufficiently integrated into the patient care process. We need a new paradigm to define and enact how the clinical physicist can engage as an active, effective, and integral member of the clinical team.

86 Medical Physics 2.0 An existential transition of clinical physics in face of the new realities of value-based, evidence-based, personalized medicine. Clinical imaging physics expanding beyond insular models of inspection and acceptance testing, oriented toward compliance, towards Team-based models of operational engagement Prospective assurance of effective use Retrospective evaluation of clinical performance

87 Conclusions 3 Skilled expertise in imaging physics is needed to understand the nuances of modern imaging equipment to Ensure quality (with relevant metrology) Ensure consistency Define and ensure conformance Inform effective use, at outset and continually Optimize quality and safety Monitor and ensure their continued performance Enable quantitative and personlized utilization