Image Quality assessment in
digital X-ray detection
systems
AAPM 2004
Summer School
Pittsburgh PA
29 July – 1 August
31-07-2004 Tom Bruijns / Dick Stueve
Philips Medical Systems
Outline
- Introduction
- Technologies in Rad and RF
- Performance Characteristics
- IQ assessment
- IQ design: a system approach
- Summary
- Evening session QC tools
19:00-21:00
TB/DS 24/09/2004
2
Overview Digital Technologies
1981
First CR
1970s
1980s
1977
Digital
subtraction
angiography
1981
Slit-scan
chest unit
1997
Se-based
flat-panel
1992
Se drum
detector
1990s
1997
CCD-based
imaging
2000
1998
CsI-based
flat-panel
Neitzel
TB/DS 24/09/2004
3
Overview Digital Technologies
CR in 1983
~ 10-20x reduction in size and price
CR in 2004
TB/DS 24/09/2004
4
Product range overview
Digital Technologies
Rad systems:
- Thoravision (selenium drum)
- Computed Radiography
- Flat Detector technology
RF systems
- IITV technology (CCD based)
- Flat detector to come
CV systems
- IITV technology (CCD based)
- Flat Detector technology
TB/DS 24/09/2004
5
Product range overview
Digital Technologies
Rad systems:
- Thoravision
- Computed Radiography
- Flat Detector technology
RF systems
- IITV technology (CCD based)
- Flat detector to come
CV systems
- IITV technology (CCD based)
- Flat Detector technology
TB/DS 24/09/2004
6
- Introduction
- Technologies in Rad and RF
- Performance Characteristics
- IQ assessment
- IQ design: a system approach
- Summary
- Evening session QC tools
TB/DS 24/09/2004
7
Flat Detector technology in Digital Radiography
TB/DS 24/09/2004
8
CR and DR
CR
DR
CR :
• DQE will increase
• Line scan
CR will coexist next
to DR for many
years
Frost & Sullivan
TB/DS 24/09/2004
9
Trixell
Moirans France
European Consortium
Thales, Philips, Siemens
Products
Static & Dynamic Flat x-ray
Detectors (FD)
TB/DS 24/09/2004
10
Flat Detector Technology
Scintillator
(CsI)
a-Si Sensor
Matrix
550 µm thickness
Address
Lines
Amplifiers
A/D-Converter
Glass Plate
43 cm x 43 cm static
18 cm x 18 cm dynamic
30 cm x 40 cm dynamic
TB/DS 24/09/2004
11
Large area (43 cm x 43 cm) 9 Mpixel Flat Detector
- For radiographic applications
- Cesium Iodide scintillator (600 µm)
- Amorphous silicon photodiode array
- Array size: 43 cm x 43 cm
cm
- Bit depth:14 bits
m
43 c
43
- Pixel size: 143 µm
- Image matrix: 3k x 3k
- Low noise electronics
- High sensitivity
TB/DS 24/09/2004
12
5 Mpixel Dynamic Flat Detector
- For vascular (and RF) applications
- Cesium Iodide scintillator (550 µm)
- Amorphous silicon photodiode array
- Array size: 30 cm x 40 cm
- Pixel size: 154 µm
- Bit depth:14 bits
- Image matrix: 2.5 k x 2 k
- Low noise electronics
- High sensitivity
TB/DS 24/09/2004
13
1 Mpixel Dynamic Flat Detector
- For cardio and vascular applications
- Cesium Iodide scintillator (550 µm)
- Amorphous silicon photodiode array
- Array size: 18 cm x 18 cm
- Pixel size: 184 µm
- Bit depth:14 bits
- Image matrix: 1 k x 1 k
- Low noise electronics
- High sensitivity
TB/DS 24/09/2004
14
CCD based IITV technology for RF applications
TB/DS 24/09/2004
15
CCD based IITV technology for RF applications
- Used for dynamic applications
- II: Cesium Iodide scintillator
- II size: 38 cm diameter
- Up to 5 zoom fields
- CCD Pixel size: 12,8 µm
- CCD Full well capacity: 170 ke- CCD read out noise 40 e- Bit depth:12 bits
- Image matrix: 10242
TB/DS 24/09/2004
16
FD technology versus CCD based IITV
technology for RF applications
+ No vignetting & no distorsion for FD
+ High resolution + coverage
+ High DQE for FD
+ Flat
- Price level high for FD
TB/DS 24/09/2004
17
- Introduction
- Technologies in Rad and RF
- Performance Characteristics
- IQ assessment
- IQ design: a system approach
- Summary
- Evening session QC tools
TB/DS 24/09/2004
18
Image Quality Triangle
Resolution
NPS(f)
MTF(f)
DQE(f)
Contrast
SNR(f)
Noise,dose,
spectrum
Neitzel, Malmö 2004
TB/DS 24/09/2004
19
Large area (43 cm x 43 cm) 9 Mpixel Flat Detector
Low noise
- For radiographic applications
- Cesium Iodide scintillator (600 µm)
- Amorphous silicon photodiode array
- Array size: 43 cm x 43 cm
cm
- Bit depth:14 bits
m
43 c
43
- Pixel size: 143 µm
- Image matrix: 3k x 3k
- Low noise electronics
- High X-ray sensitivity
TB/DS 24/09/2004
20
Large area (43 cm x 43 cm) 9 Mpixel Flat Detector
- For radiographic applications
- Cesium Iodide scintillator (600 µm)
Resolution
(and Coverage)
- Amorphous silicon photodiode array
- Array size: 43 cm x 43 cm
cm
- Bit depth:14 bits
m
43 c
43
- Pixel size: 143 µm
- Image matrix: 3k x 3k
- Low noise electronics
- High sensitivity
TB/DS 24/09/2004
21
CCD based IITV technology for RF applications
- For dynamic applications
- II: Cesium Iodide scintillator
Low
Noise
- II size: 38 cm diameter
- Up to 5 zoom fields
- CCD Pixel size: 12,8 µm
- CCD Full well capacity: 170 ke- Low dark noise 40 e- Bit depth:12 bits
- Image matrix: 10242
TB/DS 24/09/2004
22
CCD based IITV technology for RF applications
- For dynamic applications
- II: Cesium Iodide scintillator
Resolution
(and coverage)
- II size: 38 cm diameter
- Up to 5 zoom fields
- CCD Pixel size: 12,8 µm
- CCD Full well capacity: 170 ke- Bit depth:12 bits
- Image matrix: 10242
TB/DS 24/09/2004
23
- Introduction
- Technologies in Rad and RF
- Performance Characteristics
- IQ assessment
- IQ design: a system approach
- Summary
- Evening session QC tools
TB/DS 24/09/2004
24
Image Quality Triangle
Resolution
NPS(f)
MTF(f)
DQE(f)
Contrast
SNR(f)
Noise,dose,
spectrum
Neitzel, Malmö 2004
TB/DS 24/09/2004
25
Detective Quantum Efficiency
The detective quantum efficiency (DQE) is considered to
be the fundamental performance parameter of digital Xray detectors.
TB/DS 24/09/2004
26
But
There are many ways to come to many different answers
TB/DS 24/09/2004
27
Detective Quantum Efficiency
Working group FD (DR)
IEC standard 62220-1
TB/DS 24/09/2004
28
Detective Quantum Efficiency
3 methods for
analysing 1 dataset
Differences +/- 15%
After using standard
IEC 62220-1
Differences +/- 5%
Neitzel, Günther-Kohfall, Borasi, Samei
Medical Physics August 2004
TB/DS 24/09/2004
29
Linking DQE and observer tests
TB/DS 24/09/2004
30
DQE versus EAK for dynamic 30x40 FD and IITV
DQE (@ low spatial frequency) versus EAK
100
DQE (lf)
80
60
FD
IITV
40
20
0
1
10
100
1000
10000
EAK [nGy]
TB/DS 24/09/2004
31
Observation tests (using Treshold Contrast Detail Detectability)
H t ( A)
A
1
H t ( A) =
Ct ( A) × A
TB/DS 24/09/2004
32
Observation tests IITV (L) and FD (R)
TB/DS 24/09/2004
33
Observation tests IITV and FD
0.9 uGy
3.5 uGy
100
100
---- FD
Ht(A)
Ht(A)
---- IITV
10
1
1
0.1
1
SQRT (A)
TB/DS 24/09/2004
10
10
0.1
1
SQRT (A)
10
High dose
34
DQE versus EAK
TCDD versus EAK
Ht(max) versus EAK
Ht(max)
100
DQE (@ low spatial frequency) versus EAK
FD
II
100
10
80
0.1
1
10
DQE (lf)
EAK [ uGy ]
60
FD
Ht(max) versus EAK
IITV
40
100
0
1
10
100
EAK [nGy]
1000
10000
Ht(max)
20
FD
10
II
1
1
10
100
EAK [nGy]
TB/DS 24/09/2004
35
- Introduction
- Technologies in Rad and RF
- Performance Characteristics
- IQ assessment
- IQ design: a system approach
- Summary
- Evening session QC tools
TB/DS 24/09/2004
36
Rationale of Image Quality (IQ) Model (Kroon)
- IQ analysis of (non-)existing systems
- system (de)composition for design process
- comparison of present versus future systems
- Fast acquisition of IQ characteristics
- optimization requires extensive data amount
- simulation (seconds) versus experiment (hours-days)
- Various IQ related studies
- design of test objects and methods
TB/DS 24/09/2004
37
Objectives of Image Quality (IQ) Model
• Combines the IQ requirements of components into
system level IQ specification
• All IQ main items are analyzed simultaneously,
leading to a.o. DQE
• Permits tolerance and parameter studies
• Allows optimisation and prevents sub-optimisation
• Design of test objects & methods
TB/DS 24/09/2004
38
Image Quality Model Main Items
TB/DS 24/09/2004
- X-radiation
spectrum, dose, AEC
- Contrast
range and transfer
- Sharpness
MTF of stationary object
- Motion blur
MTF of moving object
- Noise
dynamic & structure WS
- Mixed
geometrics & cosmetics
39
Input → Image Quality Model → Output
X-ray
Energy
Spectrum
TB/DS 24/09/2004
Intensity
Transfer
Function
Modulation
Transfer
Function
Temporal
MTF
- Model input:
- Components
- Configuration
IQ model
Sinar-X
- Tuning
- IQ model:
- Architecture
- IP functions
- Model output
Wiener
Spectrum
- IQ descriptors
40
Image Quality Model Implementation
-
PC with LabVIEW ®
Visual programming
Clear hierarchy
IQ analysis << 1 sec
350 program parts
About 300 variables for
settings, UI and system
definition
TB/DS 24/09/2004
41
- Introduction
- Technologies in Rad and RF
- Performance Characteristics
- IQ assessment
- IQ design: a system approach
- Summary
- Evening session QC tools
TB/DS 24/09/2004
42
Summary
We discussed:
-
Products FD and IITV and their proporties
DQE and the present limitations
DQE versus observation tests
IQ modeling for fully optimized system IQ
TB/DS 24/09/2004
43
TB/DS 24/09/2004
44
Thank you for your attention
See you this evening at our booth
for the session “QC tools”
Tom Bruijns
Dick Stueve