Image Quality – Quantifying Quality?

(1)

Stephan Scheidegger, 2015

Image Quality – Quantifying Quality?

Contents

 Motivation

 Mathematical representation of image and imaging process

 From quality to quantity?

 Low contrast detectability:

Contrast, SNR, CNR,

 Noise characteristics: NPS

 High contrast detectability:

MTF

 CDC, DDC

 Model observers – observer models

 Phantoms

ROENTGENTECHNIK

STRAHLENBIOLOGIE

GRUNDLAGENRADIOLOGIE

STRAHLENPHYSIK

(2)

Why are we trying to measure Image Quality?

Different reasons – different tasks:

 Performance: system suitable for clinical rasks?

Clinical relevant quality (Model Observer)

 Optimisation: system working at optimal point (ALARA)?

Comparison of quality with applied dose (e.g. DDC with CTDI)

 Quality control: Change in systems performance?

Comparison of a measure representig systems performance with the base line (e.g. DQE, NPS, MTF etc.)

Mathematical Image Representation

Different concepts (models):

 2Dim.‐Functions (continuous models)

 Matrices (discrete representation – pixel‐model)

 Vector‐representation (Dim.=k x l x 1 resp. kx lx 3 for RGB)

   

( , ) ( , , )

...

( , ) ( , )

( , )

I x y t O x y z t

I x y F P x y S u v FFT F FFT P

 

  

 

 

A  ^{( , , )}

...

( , ) ( , )

H

kl

kl k l

A

I O x y z

I P x y W x x y y dx dy



    

 



I S N

M

(3)

From Quality to Quantity?

technical clinical

Observer impression

numbers

MTF, NNPS DQE, SNR CNR CDC / DDC

Alternative Force Choice AFC

Visual Grading Analysis VGA

Receiver Operating Characteristics ROC

Line pair TO

Image Quality – Image Characteristics

Qualities to quantities

Scharpness (spatial resolution, esp. High‐contrast resolution)

Gray level (distribution), dynamic range

Contrast

Noise

… others like uniformity, lag &

ghosting

(4)

Image Signal – Gray Levels, Blackening & Co.

Image signal representation

Blackening in film‐screen systems: optical density

Gray level displayed on a monitor (8 or 10 bit resolution, 256 – 1024 levels)

Gray level stored in a image file (12 bit – 4096 levels)

CT numbers, HU (12 bit, extended 12 bit)

Image Signal – Gray Levels, Blackening & Co.

Gray level – optical density

By definition adapted to visual process (and range!)



 



 

I D_opt log I^ref 0.5

1.0 2.0 3.0

(5)

Image Signal – Visual Contrast Detection

Visual range & «contrast resolution»

Visual range approx. 10 bit (env. 900 gray scales)

Thresholds!

(1) (2)

max min

(Michelson)

opt opt

nm kl

m

C D D

C I I

I I

C I I

 



Image Signal – Visual Contrast Detection

LC detectability Contrast C

(6)

Contrast & Dose: Gradation Curve

 ^^_^







 

1 1000

Dopt

D S Gy

usable range

(7)

Gradation Curve

☼flat

☼steep

Image range

= usable range?

(8)

Gray Scale to Dose: CR System

0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012 0.0014 0.0016 0.0018 0.002

2000 2500 3000 3500 4000

x (12bit)

c / mGy/x(12bit)

cf (meas) cf (calc)

Good contrast?

(9)

Image Quality and Optimisation

 

1 1

, ,...

( ; ,...)

( ; )

PMMA PVC

HU HU

SNR HU kV CNR HU kV

1 1

( ), ( )

c kV w kV

Parameter Set 1

kV1 Parameter Set 2

kV2

 

2 2

, ,...

( ; ,...)

( ; )

PMMA PVC

HU HU

SNR HU kV CNR HU kV

1 1

( ; ,...)

( ; )

SNR I kV CNR I kV

2 2

( ; ,...)

( ; )

SNR I kV CNR I kV

2 2

( ),

( )

c kV w kV

(10)

Low‐Contrast‐Detectability: Signal – to – Noise Ratio SNR

Sources of noise:

Quantum noise

Detector noise

More clinical: «decision noise!»

Often used: additive noise model: I_N(x,y) = I(x,y) + N(x,y)

1 1

1 ( , )

( ) ( )

N M

n m

nm nm

HU x y N M HU

SNR s HU s HU

 

   

Noise level 20%, r= 2 pixel Noise level 5%, r= 2 pixel

Noise level 10%, r= 2 pixel

(11)

Dose-dependence of noise (CR-system)

1 1

1 ( , )

N M

N M n m

n m

R I x y

s N M



 

 

  

0 20 40 60 80 100 120

0 0.5 1 1.5 2 2.5

Dose / mGy

R(101x101) 73 kV

90 / 125 kV (125 kV)

(12)

CT: Tube current and noise

Low‐Contrast‐Detectability: Contrast – to – Noise Ratio CNR

Simple approach: Difference of SNR in two compared ROI’s

Usefull for relative signal detection with threshold?

2 1

12

, 2( ) , 1( )

Pos Pos

nm Pos nm Pos

HU HU

CNR SNR

s HU s HU

   

(13)

Low‐Contrast‐Detectability: Contrast – to – Noise Ratio CNR

CNR = ‐0.007

CNR = 0.03

Is CNR a usefull quantity?

upper threshold

lower threshold signal

upper threshold

lower threshold signal

Increasing noise

(14)

+ Noise Stochastik Resonance!

Nosie Characteristics?

(15)

Noise Characteristics CT

Miéville et al (2012): Effects of computing parameters and measurement locations…Phys Med

(16)

Marshall N: The diagnostic Xray perspective, UZ Leuven

Noise Characteristics: NPS

High‐Contrast‐Resolution

Observer‐based vs.

calculated:

 Line pairs (lp / mm)

 Modulation transfer function MTF HC resolution

(lp / mm)

a) 73kV 32mAs b) 90kV 8mAs

(17)

FT

(18)

Modulation Transfer Function MTF

(19)

Results without Noise

• ROI with high contrast edge:

no filter (1); r= 0.5 p (2); r= 1 p (3); r= 2 p (4); r= 4 p (6); r= 12 p (8)

MTF high C

-0.2 0 0.2 0.4 0.6 0.8 1 1.2

0 0.2 0.4 0.6 0.8 1

lp / pixel

MTF

Reihe1 Reihe2 Reihe3 Reihe4 Reihe5 Reihe6 Reihe7 Reihe8

initial

Gaussian filter r= 4 pixels

-50 0 50 100 150 200 250 300

-8 2 12 22 32

Position / pixel

Grey Value

MTF high C

-0.2 0 0.2 0.4 0.6 0.8 1 1.2

0 0.2 0.4 0.6 0.8 1

MTF

(20)

Gaussian filter r= 12 pixel initial

Gaussian filter r= 4 pixels

-50 0 50 100 150 200 250 300

-8 2 12 22 32

Position / pixel

Grey Value

MTF high C

-0.2 0 0.2 0.4 0.6 0.8 1 1.2

0 0.2 0.4 0.6 0.8 1

lp / pixel

MTF

MTF high C

-0.2 0 0.2 0.4 0.6 0.8 1 1.2

0 0.2 0.4 0.6 0.8 1

lp / pixel

MTF

MTF(50)

(21)

MTF high C

-0.2 0 0.2 0.4 0.6 0.8 1 1.2

0 0.2 0.4 0.6 0.8 1

lp / pixel

MTF

MTF(50)

Modulation Transfer Function MTF

(22)

Results with Noise

• Gaussian filter with r= 0.5 pixel, no noise (blue), 10% Gaussian noise (yellow), 20% Gaussian noise (pink)

0 0.2 0.4 0.6 0.8 1 1.2

0 0.2 0.4 0.6 0.8 1

Results with Noise

• ImageJ: MTF with no noise vs.

MTF with noise, at high

contrast level: (a) Gaussian filter with r

= 4 pixels, noise level 5%; (b) Gaussian filter with r = 0.5 pixel, noise level 20%; (c) Gaussian filter with r = 0.5 pixel, noise level 10%

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

0 0.2 0.4 0.6 0.8 1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

0 0.5 1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

0 0.5 1

(23)

Optimisation of CR-Systems

b) 73kV 2mAs

a) 73kV 32mAs c) 90kV 8mAs d) 125kV 4mAs

Optimisation of CR Systems

(24)

Modulations-Transfer- Funktion

0 0.5 1 1.5 2 2.5 3 3.5 4

0 0.5 1 1.5 2 2.5

Dose / mGy

MTF / lp/mm

MTF50 MTF80 1.0

0.5

0.3 0.8

5 10 15 20

0 0.0

Bildfrequenz  / (lp/mm)

MTF

( ) ( ) 2 ^ix

MTF  ^ L x e^ ^{ }^ dx



  

Modulation Transfer Function MTF:

Iterative Reconstruction MTF^total  k MTF^k

MTF 50 Abdomen (Siemens Force Abdomen 2.0 Br36 3) 0.227 Abdomen (Siemens Force Abdomen 2.0 Br36 4) 0.239 Abdomen (Siemens Force Abdomen 2.0 Br36 5) 0.2435

Richard et al. (2012): Towards task‐based assessment of CT performance. Med Phys

(25)

Alternative Approach

Images:[1]

Contrast‐detail curve[1, 2, 3]

(26)

Thorax Protocol

@ 120 kV B70s

92.5mA/7.62CTDI

69mA/5.83CTDI

58.75mA/4.93CTDI

No Window Window C40 W300

Window C300 W1500

(27)

Images using different windows

Pelvis Protocol

@

80 kV

100 kV

No Window Window C40 W300 Window C300 W1500

Aproach for obtaining CDC and DDC

Registration

CT Image

Template Image

(28)

Approach for CDC and DDC

Paired point matching using ICP

Affine Transformation T

(29)

Model Observers – Observer Models

Aims

 Standardized observer

 To mimic psychophysiological aspects of recognition

 To cover image quality close to the clinical need

1 0

0 1

2 2

: H

: H( )

;

1 1

2 2

b

b s

H H

t

H H

t t

t SNR

 

 

  

  



I I N

I I I N

T I

Choosing the «Right Phantom» … ?

(30)

Dental Volume Tomography DVT

ZHAW Head Phantom

‐ High contrast resolution:

Bony structures (skull and spine / vertebral body)

‐ 10 cm and 30 cm DLP accessible

‐ SNR and MTF

(31)

EMI‐Scanner 1972

(32)

CT ‐Definitionen

1 ( )

CTDI D z dz h





   

Computed Tomography Dose Index CTDI:

• Gesamte Dosis (inkl.

Streustrahlung) auf Schicht aufgerechnet, entlang einer Linie parallel zur Rotations- achse

• Dosismass pro Schicht

z D(z)

CTDI

CT - Definitionen

100, 100,

1 2

3 3

w c p

CTDI  CTDI  CTDI

Gewichteter CTDI:

• Mittelung zwischen peripheren und zentralen CTDI

• getrennt für Kopf- und Rumpfphantom gebildet (PMMA, Durchmesser 16 cm bzw. 32 cm)

(33)

Standard CTDI Phantom

‐ too short (15 cm)

‐ does not correspond to patients anatomy

‐ does not allow testing AEC

‐ no image quality measured simultaneously

Clinical Protocolls: AEC, Overranging, Dose Profiles

(34)

Optimisation: Impact toDLP, CTDI, MSADandE

z D(z)

CTDI

Risk (E)

( ,...)

Luft Mittel

EDLP  f kV

1 ( , , ,...)

Organ z Luft

z

H

CTDI f Organ z kV p







 

T T T

Ew H

CT - Dosis-Ermitlung

Abschätzung der effektiven Dosis:

• aus Dosis-Längen-Produkt DLP

• und aus Mittelwert der Dosiskonversionsfaktoren

Luft Mittel

E  DLP  f

(35)

Körper- Abschnitt

Frauen (mSv/

(mGy*cm)

Männer (mSv/

(mGy*cm)

Kind (7J) weibl.

(mSv/

(mGy*cm)

Kind (7J) männl.

(mSv/

(mGy*cm)

Säugling weibl.

(mSv/

(mGy*cm)

Säugling männl.

(mSv/

(mGy*cm)

Schädel 0.0022 0.0020 0.0028 0.0028 0.0075 0.0074

Hals 0.0051 0.0047 0.0056 0.0055 0.018 0.017

Thorax 0.0090 0.0068 0.018 0.015 0.032 0.027

Ober-

bauch 0.010 0.0091 0.020 0.016 0.036 0.034

Becken

(Frau) 0.011 0.0062 0.018 0.011 0.045 0.025

Abdomen 0.010 0.0072 0.019 0.014 0.041 0.031

Mittelwerte f_mittel

Frage: Welche effektive Dosis ist zu erwarten?