Digital Detector Arrays (Flat Panel Detectors) · 2007-07-31 · Digital Detector Arrays (Flat...

BI

29.0

8.03

©K

B, A

pril

, 200

7

YXLON. The reason why 1

Digital Detector Arrays

Digital Detector Arrays(Flat Panel Detectors)

Klaus Bavendiek,YXLON International, Hamburg, GermanyUwe Ewert, Uwe Zscherpel, BAM, Berlin, Germany

April 2007

BI

29.0

8.03

©K

B, A

pril

, 200

7



>110 years:

X-Ray film was developed formedicial applications

- High Speed Films for minimumdose for patient with flourecentsscreens

~50 years:

X-Ray film was adapted toindustrial applications (Pb screens):

- Higher spatial Resolution

- Higher Energy Range

- Lower Noise (higher dose)

~ 20 years:

DDAs* were developed formedicial applications

- High Speed DDAs for even lessdose for patient

- Highly optimized for a limitedrange of energy and resolution

*DDA = Digital Detector Array or Flat Panel Detector

~10 years:

- DDAs were taken as there are for industrial applications

~ 3 years:

- DDAs were adapted to industrialrequirements

History of X-Ray applications

BI

29.0

8.03

©K

B, A

pril

, 200

7



Radiographic Image Quality is the basis for all comparisons

New Table in ASTM E94

Radiographic Image Quality

Radiographic Contrast

Film Contrast

Subject Contrast

Film System GranularityRadiographic Definition

Inherent Unsharpness

Geometrical Unsharpness

DDA Image Quality

QuantumNoise

DetectorContrast

Subject Contrast

DetectorUnsharpness

Projected Unsharpness

DetectorNoise

BasicNoise

StructualNoise

BI

29.0

8.03

©K

B, A

pril

, 200

7



Image Quality Parameter with DDAs

QuantumNoise

Detector Contrast

Subject Contrast

DetectorUnsharpness

Projected Unsharpness

DetectorNoise

Basic Spatial ResolutionSRB (effective Pixel Size)

CNR

SNR

CNR = Contrast to Noise RatioImportant Image Quality Paramter for DDA:

1. SRB Basic Spatial Resolution

2. SNR Signal to Noise Ratio / Efficiency

3. CS Contrast Sensitivity

4. DR Dynamic Range

5. LAG Behaviour in time

6. IScR Internal Scatter Radiation

7. BP Bad Pixel Distribution

SNR = Signal to Noise Ratio

CS = Contrast Sensitivity= 1 / CNR

BI

29.0

8.03

©K

B, A

pril

, 200

7



Image Quality Parameter: Basic Spatial Resolution

SRB

Spatial detector resolution of the image in x- and y-direction

Smallest detail perpendicular to the X-ray beam resolved in the image

Also known as effective pixel size

Corresponds to ½ of the detector unsharpness

Measured with the Duplex Wire (EN 462-5 or ASTM E2002)

May be calculated out of the visibility of duplex wires

BI

29.0

8.03

©K

B, A

pril

, 200

7



Measurement of the SRB with the Duplex Wire (here: D2 Film)

wire distance wire distanceD4 400µm D9 130µmD5 320µm D10 100µmD6 250µm D11 80µmD7 200µm D12 63µmD8 160µm D13 50µm

BI

29.0

8.03

©K

B, A

pril

, 200

7



SRB of film is much better than with DDAs without magnification

SRB of a 200µm detector. The SRB is 200µm.

With magnification & small focal spot similar SR with a 200µm detector:

BI

29.0

8.03

©K

B, A

pril

, 200

7



Film like SR with DDA using µ-focus system and magnification

Spatial Resolution of 50µm using a of a 200µm SRB detector and µ-Focus tube

But: The very high SRB is not necessary for most applications

BI

29.0

8.03

©K

B, A

pril

, 200

7



σDLoss of information / perceptibility

Example of different SNR at film

Image Quality Parameter: Signal to Noise Ratio (SNR)B

I 29

.08.

03©

KB

, Apr

il, 2

007



Signal to Noise Ratio (SNR) = 2

Examples for different SNR in the Image

BI

29.0

8.03

©K

B, A

pril

, 200

7




Examples for different SNR in the ImageB

I 29

.08.

03©

KB

, Apr

il, 2

007





BI

29.0

8.03

©K

B, A

pril

, 200

7



D-D

0

Signal

2*Noise== 2 SRMS

SNR = S

SRMS

Det

ecto

r U

nits

1024

2048

Image Quality Parameter: Signal to Noise Ratio (SNR)

Noise is in every image. Reasons for noise are:

Quantum Noise (X-Ray) Quantum Noise (X-Ray)Grain Distribution Structure NoiseFog Contribution Electronic Noise

BI

29.0

8.03

©K

B, A

pril

, 200

7



Noise in the Film

Optical Densitiy above fog and base SignalGradient GD @ D=2 and D=4 above fog and base Dose ResponseGranularity σD @ D=2 Noise

Most important parameter for the perception of fine flaws == G2 / σD

Conversion to SNR possible by (linear systems)

SNR = (G2/σD) / ln(10) = (G2/σD) / 2,3026

Diaphragm area (aperture) converted to square shaped area of 88.6 x 88.6µmfor comparison of digitized films with DDAs and different SRB

Measured SNR has to be corrected by

SNRnorm = SNRmeas * 88,6µm / SRB

BI

29.0

8.03

©K

B, A

pril

, 200

7



The limit in SNRnorm for film is some-where in the range of 150 - 200.

With more dose the SNR would

become better but film with densities >4.8 is not readable.

Normalized SNR in the Standards (at Film Density D-D0=2; EN 584-1)

EN 14784-1 Computed Radiography

ASTM E 2446 Computed Radiography

BI

29.0

8.03

©K

B, A

pril

, 200

7



DDA: Electronic Noise Quantum Noise Structure Noise

Reasons for Noise in the film and in digital Systems

Film: Base and fog Quantum Noise too dark for readout

Is this the samewith a DDA?

BI

29.0

8.03

©K

B, A

pril

, 200

7



Two important differences between Film DDAs at Noise:

Quantum Noise (X-Ray) Quantum Noise (X-Ray)Film Granularity Structure NoiseFog Contribution Elektronic Noise

FIXED WITH THE SE-LECTION OF THE FILM

With Image Integration in a Computerand very long exposure times ( )it can be reduced to ~0

8

Can be compensated with a propperdetector calibration

BI

29.0

8.03

©K

B, A

pril

, 200

7



300s integration time


Conditions: 1000mm Focus DDA Distance, SNRnorm calculated by substracted images


SNRnorm= 6600

Film:SNRnorm= 150

SNR at a good DDA is mainly limited by Photon Statistics and the Efficiency differs for different Energies:

BI

29.0

8.03

©K

B, A

pril

, 200

7



The Efficiency can be used to select the best DDAfor a dedicated Energy level:

Normalized SNR forcomparison of different systems

BI

29.0

8.03

©K

B, A

pril

, 200

7



Image Quality Parameter: Contrast Sensitivity

Contrast Sensitivity (CS) indicates the smallest material difference which can be resolved in the image

Contrast= ΔS

CS = = SRMS

ΔS1

CNR

BI

29.0

8.03

©K

B, A

pril

, 200

7



Measurement of Contrast Sensitivity

15mmBase material

1.25mm

2.5mm

5.0mm

7.5mm

10mm

12.5mm

15mm

2% Groove

CS measured with

Duplex Stepwedge

(stepwedge with grooves)

CS is calculated of

difference in signal

devided by

mean noise level

per step.

Example: Stainless Steel

BI

29.0

8.03

©K

B, A

pril

, 200

7



Measurement of Contrast Sensitivity

Three different Materials:Aluminium, Titanium and Stainless Steel

Old version: still 2 grooves ...

BI

29.0

8.03

©K

B, A

pril

, 200

7



Result of Contrast Sensitivity Measurement

In the range of 1.25 to 12.5 mm stainless steel a 0.2% material thicknesscan be resolved.Film level is >1% and only possible within a smaller material thickness range

BI

29.0

8.03

©K

B, A

pril

, 200

7



Image Quality Parameter: Dynamic Range

Three different Materials:Aluminium, Titanium and Stainless Steel

SNR on each step is calculated for different exposure times ...

... also called in ASTM: „Specific Material Thickness Range“

BI

29.0

8.03

©K

B, A

pril

, 200

7



Measurement of the Dynamic Range B

I 29

.08.

03©

KB

, Apr

il, 2

007



Measurement of the Dynamic Range

Dynamic Range from 10 to 72mm for 1% applications with 4s exposure timeDynamic Range from 10 to 90mm for 2% applications with 4s exposure time

BI

29.0

8.03

©K

B, A

pril

, 200

7



Image LAG or Ghosting

Image LAG is caused by the electronic of a DDA

BI

29.0

8.03

©K

B, A

pril

, 200

7



Image LAG or Ghosting

Image LAG is caused by the electronic of a DDA

BI

29.0

8.03

©K

B, A

pril

, 200

7



LAG and Burn In

Negative image of a turbine blade two days ago. Corrected one day before

Lead letter on the step wedge from the day before. No new calibration was done ...

Example for „Burn In“

Burn In is a long term effect in the Scintillator of a DDA

BI

29.0

8.03

©K

B, A

pril

, 200

7



Internal ScatterRadiation

a

b

Internal Scatter Radiation increases with the energy.

Important Parameter for CT machines ...

BI

29.0

8.03

©K

B, A

pril

, 200

7



Bad Pixel

... only a dead pixel is a bad pixel?

Not really ...

BI

29.0

8.03

©K

B, A

pril

, 200

7



Bad Pixel

BI

29.0

8.03

©K

B, A

pril

, 200

7



Groups of Bad Pixel (Cluster)B

I 29

.08.

03©

KB

, Apr

il, 2

007



Main Difference between Film and DDAs

Two unique property of DDAs:

Take multiple images at identical pixel positions and cumulate.

Calibrate on pixel basis with

suitable energy and

material in the beam similar to the test object

The key for best image quality with contrast sensitivity and superior SNR is

the exact calibration of each detector pixel

to obtain the same radiation response as the neighbourhood detector pixel.

BI

29.0

8.03

©K

B, A

pril

, 200

7



Basic Functionality of all Types of Detectors

X-Rays reaching the sensitive area of the detector are converted toelectrical charge.

The amount of charge is measured.

It is proportional to the Dose.

Different Technologies for Flat Panel Detectors are available

intrinsic Method(„direct Method“) - very inefficient, no products

direct Method with Photoconductor (Amourphous Selen or CdTe ) - example: AJAT DICT100 CdTe

indirect Method (Converting X-Ray to Light with Scintillator, „normal“ Photodiodes)

BI

29.0

8.03

©K

B, A

pril

, 200

7



Matrix-Structure of all Flat Panel Detectors

.. with Photo Diode for „Light Capture“ and TFT for selective Read out

BI

29.0

8.03

©K

B, A

pril

, 200

7



Basic Principle: Indirect Method with Scintillator

X-ray quantum

Scintillator (screen) Coating

Photo Diode Layer

Pixel width (indicated geometrical resolution)

A thick scintillator results in many photons, but with a loss of geometric resolution(noise vs. resolution)

The effective geometric resolution depends on various parameters, these must be measured!

Cro

ss S

ectio

nal V

iew

of D

etec

tor light quantum

BI

29.0

8.03

©K

B, A

pril

, 200

7



Basic Principle: Indirect Method with Scintillator

Scintillator (screen) Coating

Photo Diode Layer

Pixel width (indicated geometrical resolution)

Cro

ss S

ectio

nal V

iew

of D

etec

tor

light quantum

X-ray quantum

Cesium iodide can be formed into needles on top of the diode structure to direct the lightto the photodiodes without significant light scatter. The needles ensures a high total internal reflection of the light and direct it to the photodiode.

Disadvantage of CsJ: ImageLag and Burn In

BI

29.0

8.03

©K

B, A

pril

, 200

7



Photodiode

TFT

~8pF

U+

TFTSwitch

LineSelection

Read-OutAmplifier

U

Photo-Diode

2. Incoming Light discharges the capacity;more light means less charge left in theCapacity

3. Now the cell should be read out:The TFT becomes conductive and the capacity is recharged through the Read-Out Amplifier up to U+ ...

TFT

1. Capacity of Photodiode is charged up to U+

3a. The amount of Charge for Rechargingin measured in the Read-Out Amplifier.It is transfered as „U“ to the ADC

What happens:

What happens during operation?B

I 29

.08.

03©

KB

, Apr

il, 2

007



1. Capture the Dark Image

Uncorrected Image with Flaws, Patterns, Bad Pixel, ...without X-Ray

==> Offset Image

How do we do the calibration

BI

29.0

8.03

©K

B, A

pril

, 200

7



Offset corrected Image without X-Ray

1. Substract the Dark Image from any further Image

How do we do the calibrationB

I 29

.08.

03©

KB

, Apr

il, 2

007



2. Capture the Gain Images

Offset corrected Image with X-Ray andwithout Gain Correction

==> Gain Image


BI

29.0

8.03

©K

B, A

pril

, 200

7



2. Multiply each Pixel with Gain Image for further Images

Offset and Gain corrected Image with X-Ray


I 29

.08.

03©

KB

, Apr

il, 2

007



energy

Det

ecto

r O

utpu

t Sig

nal

[AD

U]

full dose

reduced dose during calibration

real dose withmaterial in the beam

2b. Correct Non-Linearities with Energy adapted Gain Images


BI

29.0

8.03

©K

B, A

pril

, 200

7



Signal after Correction with Offset and Multi-Gain

Signal after Correction with „Multi-Gain“

„Gain Line“, calculated of Correction Points 1 to 3

Real (non-linear) Detector Output Signal

Raw Detector Signal

Cor

rect

ed D

etec

tor

Out

put

r4

Offs

etn 1

r1

Correction Point 1

r2

Correction Point 2

r3

Correction Point 3


I 29

.08.

03©

KB

, Apr

il, 2

007



(SNR is limited to ~1000 because of the Structure of the Material)

With 16s same quality like

with 500s andnormal calibration

A High SNR is optained with Optimized Calibration

BI

29.0

8.03

©K

B, A

pril

, 200

7



Aluminum, 10 - 110 mm, PerkinElmer RID 512 / AF1 and Image 3500 DD

in %

Inspection wide Range of Material with optimal CalibrationB

I 29

.08.

03©

KB

, Apr

il, 2

007



With a high SNR details can be seen more easily and even objects smaller than the pixel size can be detected

800µm

200µm

127µm

400µm

1200µm

Aluminum wedge with 5 long holes inside;

Image from 400µm pixel pitch panel PerkinElmer RID512/400 AF1

40mm Al

5mm Al

with Highpass Filter

Can a High SNR Compensate a Lower SRB?

5mm

40mm

BI

29.0

8.03

©K

B, A

pril

, 200

7



800µm

200µm

127µm

400µm

1200µm

This is just a third of the pixelsize.How is this possible?

With a high SNR details can be seen more easily and even objects smaller than the pixel size can be detected

Can a High SNR Compensate a Lower SRB?

with Highpass Filter

Aluminum wedge with 5 long holes inside;

Image from 400µm pixel pitch panel PerkinElmer RID512/400 AF1

BI

29.0

8.03

©K

B, A

pril

, 200

7



With a high SNR it is enough that just a part of a pixel gets the information.

The long hole, just a quarter of a pixel width

Example:

The hole has 2% less density than the rest.

The hole covers ¼ of the pixel.

The visibility would be 2% * ¼ = 0.5%.

With a SNR > 200 the 0.5% is visable.

Pixel Grid of the Detector

Yes, a High SNR can Compensate a Lower SRB

BI

29.0

8.03

©K

B, A

pril

, 200

7



Compensate a lower SRB with a higher SNR

Another example with 8mm steel weld (test weld BAM5)

Film AGFA D2, class C1 EN 584-1

330s exposure time

PE XRD1620 detector (200µm SRB), magnification of 3.0 100s exposure time only

BI

29.0

8.03

©K

B, A

pril

, 200

7



Compensate a lower SRB with a higher SNR

Film AGFA D2, class C1 EN 584-1

330s exposure time

PE XRD1620 detector (200µm SRB), magnification of 3.0 100s exposure time only

Another example with 8mm steel weld (test weld BAM5)

BI

29.0

8.03

©K

B, A

pril

, 200

7



Advantages of Digital X-Ray

Time of Inspection will be reducedExample (from Boeing): 6“ diameter Inconel, 10 welds

Film Process: 2-3 hoursDigital X-Ray: 12 minutes 10 seconds

Avoiding high Recurring Costs (Chemicals and Film)

No Chemical Processing incl. associated facilities + waste water handling

Lower Storage Costs - cheaper Archiving

Faster Access to an Image in the Database - if needed

Image Transfer from one Location to another in few seconds (network)

Very short set-up time with manipulation system (no film transport)

Ability to use Digital Image Processing like Filter, Zoom, Measurement

Smaller Footprint

Inspection of larger Range of Material Thicknesses within one Image

Higher recognition rate of defects in production ...

BI

29.0

8.03

©K

B, A

pril

, 200

7



Comparison of Film (single wall, double wall), RT und DR

done by Bill Meade, Clay Kidwell, Greg Warren (Boeing Commercial Aircraft)

100%

Digital System: Thales FS35 with YXLON Image 3500 DD

BI

29.0

8.03

©K

B, A

pril

, 200

7



Result of the POD Study by Flaw „Volume“ (Depth x Length)B

I 29

.08.

03©

KB

, Apr

il, 2

007



Saving time using DDA systems

Necessary time for inspection of a part:

2 – 3 hours with Radiography using the current film process

35-40 minutes with current radioscopy inspection

12 minutes are sufficient for the same task with the DDA system

Hint:

The POD was not done withthe best available DDA.

New gereration of DDAs increase SNR and CS byfactor of 2.5.

BI

29.0

8.03

©K

B, A

pril

, 200

7



Example: Small Turbine Blade with ASTM Penetramter „5“

Film Agfa D2,

420s exp. timedigitized with50µm and 16 Bit

DDA + µ-Focus

PE XRD1620,

60s exp. time

Imag

es h

ighp

ass

filte

red

BI

29.0

8.03

©K

B, A

pril

, 200

7



Cutout of Penetrameter with the 3 Holes

Film Agfa D2,

420s exp. timedigitized with50µm and 16 Bit

DDA + µ-Focus

PE XRD1620,

60s exp. time

254µm

BI

29.0

8.03

©K

B, A

pril

, 200

7



Example: Small Titanium Duct with Penetramter „5“

Film Agfa D2

100s exposure timeDDA System

16s exposure time

Imag

es h

ighp

ass

filte

red

BI

29.0

8.03

©K

B, A

pril

, 200

7



Cutout of the defects in the weld

Film Agfa D2

100s exposure time

DDA System

16s exposure time

50µm

BI

29.0

8.03

©K

B, A

pril

, 200

7



Inspection of Ducts, potential for ADR ?B

I 29

.08.

03©

KB

, Apr

il, 2

007



Inspection of Ducts, potential for ADR ?

BI

29.0

8.03

©K

B, A

pril

, 200

7



Diameter 0.05mm

Inspection of Ducts, potential for ADR ?B

I 29

.08.

03©

KB

, Apr

il, 2

007



The new DDA systems are suitable for film replacement, ADR and CT

Basic Parameters for Digital Radiology are- Basic Spatial Resolution (SRB)- Contrast Sensitivity (CS)- Dynamic Range Specific Material Thickness Range- Normalized SNR (SNRNorm) Efficiency

SNRNorm of DDA system depends on - integration time, - calibration procedure and - radiation quality

Conclusion:

DDA systems can exceed Film systems by factor >20 in SNRNorm

Very high SNR gives a superior Contrast Sensitivity CS

A high Contrast Sensitivity can compensate a lower SRB

With > CS fine details <1 pixel size give enough signal to become visible

Adapted Filters increase the Visibility of Inhomogenities

Digital Detector Arrays (Flat Panel Detectors) · 2007-07-31 · Digital Detector Arrays (Flat...

Documents

Transcript of Digital Detector Arrays (Flat Panel Detectors) · 2007-07-31 · Digital Detector Arrays (Flat...