Technical development of a high resolution CCD-based scanner for 3-D gel dosimetry:

(II) Problems encountered

S J Doran, K K Koerkamp*,

S Department of Physics,University of Surrey,Guildford, GU2 7XH, UK

Department of PhysicsUniversity of Surrey

Department of Applied PhysicsUniversity of Twente, NL

*

S J Doran, K K Koerkamp Dept of Applied Physics,University of Twente,Enschede, NL

Structure of talk

• Factors determining signal detected

• Detector and projection screen characteristics

• The “ring artifact” and how to remove it

• The “correction scan” procedure

• Sample containers

• The dynamic range problem

• Conclusions

Signal measured in CCD tomography

Hglamp

Cylindrical lens, pinholeand filter pseudo

point-source

Lens parallel beamScanning tank withmatching medium

Exposed gel

Unexposed gel Diffuser screen on which realshadow image forms

CCDdetector

Standard 50mmcamera lens

PC with frame-grabber card

Turntable controlled by acquisitioncomputer via stepper motors • Light field L(x,y)

• Projection screenPS(x,y)

• Detector responseD(x,y,S)

• Gel absorptionG(x,y,)

• Reflection and refraction• None of these quantities is known a priori.

• We can estimate L(x,y) relatively easily.

• PS(x,y) and D(x,y,S) can be a problem.

CCD detector characteristics: (1) Dark response

• We started out using a cheap CCD detector (~£120).

• The “noise” from the detector has a clear structure.

• This has serious consequences for improvement in signal by averaging frames.

(naverages)1/2

0 5 10 15 20

1 /

SNR

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

for pixels in detector column

for pixels in detector row

-5

+5

-5

+5

Image from detector with lens cap on!

CCD detector characteristics: (2) Light response

• Response measured by exposing CCD to different light levels, obtained using two polaroids rotated w.r.t. each other.

• No need for parallel beam here. Collimating optics and projection screen not used.

• Does the response vary with pixel position?

0

200

400

600

800

1000

1200

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Relative intensity

Pix

el v

alu

e

Measured data points6th order polynomial fit

Open light field and projection screen

• Relatively easy to separate slowly varying L(x,y) from PS(x,y) and D(x,y).

• Oscillating the projection screen up and down “smears out” some of the granularity.

• How much of the residual noise comes from PS(x,y) and how much from D(x,y)?

• Replacing the projection screen shows the extent of the problem.

P ix e l n u m b e r in x -d ire c t io n0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 3 0 0 3 5 0 4 0 0 4 5 0 5 0 0

Pix

el v

alu

e

6 0 0

7 0 0

8 0 0

9 0 0

1 0 0 0

1 1 0 0

P ix e l n u m b e r in y -d ire c t io n

0 5 0 1 0 0 1 5 0 2 0 0

Without oscillating projection screen

Oscillating projection screen

Pros and cons of two different projection screens

• Screen 1 (engineering tracing film) is granular, but produces sharp images.

• Screen 2 (opal white perspex) has much less granularity, but the projection images are blurred.

Pixel number in x-direction0 50 100 150 200 250 300 350 400 450 500

Pix

el v

alu

e

600

700

800

900

1000

1100

Consequences of PS(x,y) and D(x,y)

• The “noise” generated is coherent between projections.

• This gives rise to a characteristic ring artifact.

Ideal object Simulated artifact Experimental artifact Artifact removed by “wobble”

“Correction” of the ring artifact via “wobble”

• At each projection step, the detector moves randomly relative to the tank by a few pixels.

• Can be achieved either by moving tank or camera.

• This allows us to sample the response functions of different pixels over the course of the acquisition.

• Coherent noise turns into random noise!

Hglamp

CCDdetector

PC with frame-grabber card

Containers and the correction scan

• Sample container has refractive index different to that of the sample and the matching medium.

• This causes partial reflection and refraction.

• Containers are imperfect, leading to artifacts in the projections.

• Problems can be partly overcome by taking the ratio of images before (“correction scan”) and after irradiation.

Before irradiation After irradiation Processed sinogram

Scratch mark

Artifacts due to imperfect containers

• Minute scratches on the container wall cause spurious reflection and refraction.

• These are easily seen as parallel tracks in the sinograms.

• They lead to characteristic artifacts at the edge of the field-of-view

- 6 Gy

6 Gy

Dynamic range problems

• Video capture card has limited dynamic range (10-bit).

• Light travelling through low-dose region saturates ADC.

• Light travelling through high-dose region registers a very low signal that is strongly affected by noise.

• Extremely important to make absorption of matching medium same as that of unirradiated gel.

Conclusions

• We now understand many of the causes of artifacts in the OCT images.

• Most of the artifacts can be removed by investment in higher quality components (particularly the CCD, projection screen and sample container).