British chapter slides in correct 35 mm format

1
CCD-based optical CT scanning of highly
attenuating phantoms
or All the most interesting stuff happens when
the experiment doesnt work
SIMON J DORAN and Shamsa Al-Nowais
2
Summary of talk

• Background What were we trying to do?
• What system were we using?
• What was the problem?
• Blooming an interesting wild goose chase
• So where did we get to?
• Here are our guesses? Can you help?

3
Background (1) Interesting results

• It all started with a talk at DOSGEL 2006 ...
• Initial experience with a commercial cone beam
  optical CT scannerDe Jean et al., J. Phys.
  Conf. Ser. 56, 179-182
• These are the results for a cone-beam scan of a
  scattering gel.

4
Background (2) Study motivation

• This observation leads to several interesting
  questions
• What is causing the problem?
• Does this just happen with scattering
  gels?Subtext Is PRESAGETM better than BANGTM ?
• Does this just happen with cone-beam
  scanners?Subtext Is my scanner better than
  yours?

• Series of four experiments designed to test this.

5
Background (3) Experimental apparatus

• Parallel-beam optical CT scanner
• Krstajic and Doran, Phys. Med. Biol. 51 2055
  (2006) and 52, 3693 (2007) 

6
Background (3) Experimental apparatus

• Parallel-beam optical CT scanner
• Krstajic and Doran, Phys. Med. Biol. 51 2055
  (2006) and 52, 3693 (2007) 

• Cylindrical sample
• Teflon FEP cylinder (Kevin Jordan)
• Water outside and inside, then gel

Teflon FEP
PET
7
So we started with the easy case ...

• Absorbing blue dye solution no scatter
• Expectation was that we would see no scatter
  effects and no cupping.
• What we saw was a surprise.
• Cupping artefacts lead to incorrect measurements
  of optical density.

8
Why does this happen?

• Incorrect measurements of attenuation in
  projections
• Cupping is a direct consequence of a distortion
  in projection shape.
• We know theoretically what the shape should
  be y(x) 2(r2-x2)1/2 ? S(x)
  exp- 2?(r2-x2)1/2

9
1st possible explanation

• The absorbance at which this happens is high gt
  2.0
• This means that our measured signal is only lt1
  of peak.
• Are we measuring the signal correctly at these
  low levels?
• Dark current signal is ?0.3 peak. Are we falling
  below the floor?

10
Can we confirm this?

• Not a satisfactory solution.
• Shouldnt be necessary.
• There is an adjustable parameter that needs
  tweaking.
• Fortunately, we can test this.
• Method for extending dynamic range (Krstajic and
  Doran, Phys. Med. Biol. 52, 3693 (2007))
• Acquire two projections.
• Measure edge regions (high signal) of phantom as
  normal.
• Measure central region (low signal) with
  increased light level.
• Reconstruct profile by splicing the two
  measurements together.
• This helps, because now all signals contributing
  to profileare significantly above the signal
  floor.

11
Interlude

• Making this measurement turned out to be harder
  than expected because of blooming.
• When using the high light level, we expect
  saturation. Algorithm copes fine.
• Blooming is the deleterious effect on regions
  around the saturated pixels.
• Problem only for CCD scanners, not laser systems.

12
How does blooming affect our data?

• We obtain spurious values for the absorbance.
• In some regions a region that should be saturated
  shows a mid-range signal.Breakdown of
  amplification chain. Known effect.
• Other regions show saturation where there
  shouldnt be.Signal overflow into adjacent
  wells. Expected effect.

13
Results of extended dynamic range

• Conclusion We have measured the correct profile!
• It does not match the theoretical model at for
  high absorbance phantoms.
• It does match for low concentrations of dye, so
  theory is OK, but incomplete.

14
Why does this happen?

• We know we are making the correct CCD
  measurements.
• Hence the signal at with high concentrations of
  dye does not match theory.
• Possible reasons
• Genuine deviation from Beers Law for high
  concentrations of dye.
• High concentrations may cause change in
  refractive index.
• Refractive index mismatch at edge of cylinders
  causing longer path of light through absorber
  than expected.
• John, Kevin and Clive are right all along ...
  Grrr! ... and it is all to do with scattering!
• Obvious test is to try the same experiment with
  the laser scanner.
• Any further suggestions appreciated!

15
Conclusions

• De Jean et al. pointed out an interesting effect
  for cone-beam scanners with polymer gels.
• We found that the same symptoms occur for
  nominally absorbing gels with a parallel-beam CCD
  scanner.
• We have ruled out measurement error as a source
  of the artefact.
• In the process we have demonstrated graphically
  the influence of blooming.
• We have put forward some suggestions for what is
  causing the effect and hope to test these
  hypotheses.