11/21/19 Role of image quality in dose management via/through DRL Ehsan Samei, PhD, FAAPM, FSPIE, FAIMBE, FIOMP Department of Radiology Duke University Health System 1 2 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 1
11/21/19 3 The challenge of dose optimization: the monotonic relationship between quality and dose! Image Quality ? Radiation Dose (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 2
11/21/19 The challenge of dose optimization: the monotonic relationship between quality and dose! Image Quality relevance robust ? smart relatable Radiation Dose practical The challenge of dose optimization: the monotonic relationship between quality and dose! Image Quality ? relevant robust ? smart relatable Radiation Dose practical (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 3
11/21/19 Image quality 7 The challenge of dose optimization: the monotonic relationship between quality and dose! Image ? Quality ? relevance robust ? smart relatable Radiation Dose practical (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 4
11/21/19 What dose is optimum? High dose Low dose Low dose High dose Safety (dose) is inherently linked to indication-based image quality. (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 5
11/21/19 Factors that govern quality and safety of medical imaging Dose An Ideal Image (high) Noise (low) Resolution What is image quality? • Aesthetic: Subjective perception of quality • Task-generic: The realism of the image to represent the reality of the object being imaged • Task-specific: The ability of the image to render the information pertinent to the task at hand 12 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 6
11/21/19 What is the right quality metric? Goal Intrinsic Phantom Anthro. Animal Virtual Case Clinical Clinical Specs Measures Models Models Clinical studies Trials Quality Trials What is the right quality metric? Goal Intrinsic Phantom Anthro. Animal Virtual Case Clinical Clinical Specs Measures Models Models Trials studies Trials Quality Complex Simple Relevant Relevance inferred (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 7
11/21/19 What are the right metrics? 1. Relevant: As much as possible, patient-/indication- centric (not modality or machine) 2. Robust: To ensure reliability and applicability (quantitative not subjective) 3. Smart: Maintained balance between robustness and relevance 4. Relatability: Surrogates relatable to clinical task 5. Practical: Economic to measure Balance robustness and relevance ARARA! Simple Complex Relevance inferred Relevant • To extent possible, we need to move toward relevance while keeping robustness in check • As Relevant as Reasonably Achievable (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 8
11/21/19 Image quality metrics Task Generic Task Specific 1. Contrast 1. Threshold Contrast 2. Resolution 2. Detectability, d’ 3. Noise 3. Estimability, e’ 4. SNR, CNR, SdNR 4. d’ in vivo 5. DQE, eDQE, eDE 6. TG IQ in vivo 17 1. Contrast 18 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 9
11/21/19 Contrast • Fractional difference in the signal or brightness between two regions of an image Low Contrast High Contrast 19 Contrast • Best characterized by fractional signal difference (ie, subject contrast) or fractional brightness difference (ie, display contrast) of a target in comparison to background: ! = # $%&'($ − # *%+,'&-./0 # $%&'($ = log # *%+,'&-./0 # *%+,'&-./0 20 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 10
11/21/19 Contrast # $%&'($ − # *%+,'&-./0 # $%&'($ ! = = log # *%+,'&-./0 # *%+,'&-./0 21 2. Resolution 22 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 11
11/21/19 Resolution • Ability to resolve distinct features of an image from each other Low resolution High resolution 23 Resolution • Best characterized by the modulation transfer function (MTF): – The efficiency of an imaging system in reproducing subject contrast at various spatial frequencies MTF(f) = F{LSF(x)} LSF = response of a system to a perfect line 24 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 12
11/21/19 INPUT OUTPUT Imaging System MTF x = f 25 INPUT Imaging System 26 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 13
11/21/19 High MTF 27 Low MTF 28 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 14
11/21/19 MTF and limiting resolution • Limiting resolution ~ Frequency at 10% MTF • Mammography 5-10 lp/mm • Radiography 2-5 lp/mm • Fluoroscopy 1-2 lp/mm • CT 0.5-1 lp/mm 29 Effect of added blur 30 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 15
11/21/19 3. Noise 31 Noise • Unwanted signals that interfere with interpretation Low resolution High resolution High res/ high noise 32 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 16
11/21/19 An underexposed image is “too noisy” Underexposure by 4x Correct exposure 33 120 kVp, photo-timed 120 kVp, 25% less mAs 34 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 17
11/21/19 Noise • Best characterized by the noise power spectrum (NPS): – The variance of noise in an image in terms of the spatial frequencies c h = òò + c + h ACF ( , ) f ( x , y ) f ( x , y ) dxdy NPS(f) = F{ACF(x)} 35 Noise • NPS Image Data ACF NPS Example 1 Uncorrelated Noise x x f Example 2 Correlated Noise x x f 36 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 18
11/21/19 Image without noise 37 Uncorrelated noise 38 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 19
11/21/19 Correlated noise 39 Effect of added noise 40 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 20
11/21/19 4. SNR, CNR, SdNR 41 SNR • Signal to noise ratio % &'()*& !"# = + ,'-.)(/012 42 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 21
11/21/19 CNR and SdNR I target − I background I background = I target − I background CNR = = SdNR σ background σ background I background • In Linear systems: – N in CNR is relative noise – N in SdNR is noise • In log systems, Sd in C and N in relative noise 43 SNR(f) • Noise Equivalent Quanta (NEQ) and frequency- dependent SNR • Affected by the collective effects of resolution and noise and associated contributing factors 2 ( f ) ( f ) = MTF 2 NEQ ( f ) = SNR actual NPS ( f ) 44 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 22
11/21/19 CNR(f) CNR 2 ( f ) = C 2 ( f ) MTF 2 ( f ) = C 2 ( f ) SNR 2 ( f ) NPS ( f ) 45 5. DQE, eDQE, eDE 46 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 23
11/21/19 6. In vivo task-generic metrics in vivo image quality Noise Resolution Christianson et al., AJR, 2014 Sanders et al., Medical Physics, 2016 Perceptual Quality Organ-based HU Abadi et al., Medical Physics, 2017 Samei et al., Medical Physics, 2014 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 24
11/21/19 in vivo noise prediction in vivo NPS Segment the liver (Fu 2018) Organ Segmentation Sub-organ Segmentation Identify liver parenchyma – Avoid vasculature, fatty deposits Sub-segment parenchyma and de-trend image data – Use local polynomial fits for segment Wiener-Khinchin Theorem: Goal NPS( u ) = FT {R N ( x 1 ; x 2 )} = FT {E[N( x 1 ).N( x 2 )]} 1. Anatomical structure Estimate NPS from non-square patches in the liver 2. Large scale trends (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 25
11/21/19 in vivo resolution Sanders et al., Medical Physics, 2016 in vitro (phantom) noise Siemens Flash GE HD750 35 50 45 Phantom 30 40 25 35 noise (HU) 30 20 25 15 20 Phantom 15 10 10 5 5 0 0 20 25 30 35 40 45 20 25 30 35 40 45 patient diameter (cm) patient diameter (cm) Ria et al, AAPM 2018 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 26
11/21/19 in vivo (phantom) noise Siemens Flash GE HD750 Phantom 35 50 Patients 45 30 40 25 35 noise (HU) Patients 30 20 25 15 20 Phantom 15 10 10 5 5 0 0 20 25 30 35 40 45 20 25 30 35 40 45 patient diameter (cm) patient diameter (cm) Task-based metrics 61 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 27
11/21/19 Task-based indicators • Direct measures of task performance • Direct measures of task-like performance • Derivatives of task performance from generic indicators 62 Direct measures of task performance • Most reliable • $ and time-consuming • Not translatable to other tasks or conditions 63 (c) Ehsan Samei, 2019. Use for non-personal purposes by prior permission only. 28
Recommend
More recommend
