FORMULARIO DE INSPECIÓN (CHECK LIST) DE COMPAÑÍAS

Phantom 3D Simulation

Mesh (3D) Cylindrical 140mm tall, 17157 nodes, 92160_{linear (tetrahedral) elements}

Basis Forward mesh 3x3x3 median filtered

Starting parameters: 0 .0 1 1mm '

Starting parameters: p's 0,8 5 m m '

Iteration 16

Sources 3 2 x 5

Detectors per source 22

Datatypes Mean (simulated)

Calibration Simulated, but inverse crime (difference)

2d 3d correction applied? No

Simultaneous Pa and p's

Acquisition time (per source) -

Wavelength -

E. M. C. H illm an . PhD th esis 2002 C h ap ter 2 . 3 — 1 3 4

2 .3 .2 .4 .2 D isc u ssio n (3D simulated)

T h e im ages show n in F igure 2.3.9 are very sim ilar to the ex p erim en tal d ifferen ce im ages show n in F igure 2.3.8, although the Pa sim u lated im ages are b etter q uality th an the ex p erim en tal ones. C ross-talk betw een Pa and p's is evident, w hich im plies that cro ss talk in the im ages from ex p erim en tal m ean-tim e is p ro b ab ly due to the use o f a single d ataty p e and not a result o f u sin g experim ental data.

N ote that even in these 3D im age re c o n stru ctio n s, we still see o u t-o f-p lan e stru ctu re. T hat is that the inclusions appear to ex ten d fu rth er vertically that they in fact do. A s w ith the 2D m ulti-slice im ages show n in F igure 2 .3.6 the true location o f the inclusion can be determ in ed by fin d in g the plane w here the ‘b lo b ’ is brightest. O f co u rse this req u ires the assum ption that the inclusion is in fact discrete, w hich m ay not be the case in clinical subjects.

T he 3D im aging results p resented in F ig u re 2.3.7, F igure 2.3.8 and F igure 2 .3.9 w ere published in (A rrid g e et al, 2000b) (su b m itted F eb ru ary 2000)

2 .3 .3 Breast phantom imaging

S ection 1.1.3.2 d escribed why the breast w as an ap p ro p riate target for optical im aging. In o rd er to test the practical aspects o f breast d ata acq u isitio n and im age re c o n stru ctio n , a num b er o f conical resin breast phantom s w ere m anu factu red to fit snugly into a specially co n stru cted conical fibre holder (see section 2 .7 .3 .2 ). T he set co m p rised o f tw o co n ical phantom s w ith back g ro u n d optical properties o f P;, = 0.007 m m ' ± 0.001 mm'* and p's = 0.8 mm * ± 0.1 m m * at 800 nm. O ne phan to m c o n tain ed three 10 m m high, 10 m m d ia m e te r cylinders: one w ith tw ice background Pa, on e w ith tw ice b ackground p's and one w ith tw ice both Pa and p's. A ll three objects w ere placed in the sam e plane as show n in F igure 2 .3.10. T he oth er p hantom w as identical but w as co m p letely hom ogenous.

2 x g , Conical optical fibre holder 2xga 2xg\ M easurem ent plane: 82m m d iam eter Background: Pa = 0 .0 0 7 mm ' ± 0 .0 0 0 7 m m p's = 0.8m m ' ± 0.1 m m ' 78 mm 100 mm 135 mm

Figure 2.3.10 The b reast phantom (containing inclusions) de sig n e d to evaluate b re a st imaging capabilities and test the conical fib re holder shown arou nd the resin phantom (right).

T he extension o f the cone shape o f the p h an to m into a cy lin d rical base w as a d elib erate attem pt to co n sid er the effects o f the chest w all. L ight pro p ag atio n w ill not be en tirely lim ited

E. M. C. H illm an . P hD thesis 2002 C h ap ter 2. 3— 1 35

to the b reast tissue alone and if the p hantom w ere m anufactured w ith o u t such as base, d ata w ould be affected by unrealistic in teractio n s w ith this boundary.

C alib ratio n for a conical phantom is co m p licated by the fact that the sources and d e te c to rs are no longer held in a ring that will fit around the ab so lu te calib ratio n tool d escrib ed in section 2.1.2.1. N ote that calib ratio n m easurem ents acq u ired w ith the fibre b u n d les and source fibres in the 70m m d iam eter ring, using the ab so lu te calib ratio n p hantom ,

w o u ld be valid fo r d ata acquired using the conical fib re holder. H o w ev er pro b lem s arise since th e fib re bundles and source fibres are frag ile and d ifficu lt to m ove betw een the tw o g eo m etries. In the tim e taken to adapt the set-up fo r d ata acquisition fo llo w in g calib ratio n , the system IR Fs m ay not rem ain stable. N ote that the sub seq u en t d ev elo p m en t o f the m onstode (see sections 2.1.2.2 and 2.7.2.2.1) allow ed sources and d etecto rs to be m oved m ore freely, and also offered the opportunity to perform som e in-situ calibration m easurem ents.

H o w ev er the problem s w ith absolute calib ratio n and the av ailab ility o f a referen ce ph an to m m eant that data could be m ost readily calib rated u sin g d ifferen ce m easu rem en ts (see 2.1 .4 .3 ). T h is differen ce d ata can be reco n stru cted in 2D or 3D as show n below .

2.3.3.1 2D breast phantom imaging

A s a prelim inary ex perim ent, d ata w ere acq u ired using all sources and d etecto rs in a sin g le plane (using the m iddle ring o f the fib re ho ld er, such that the m easu rem en t plane co in cid ed w ith the inclusions). M O N S T IR ’s 32 source and d etecto r fibres w ere all positio n ed w ithin the single m iddle ring. T he V O A s w ere used to shut o ff 5 d etecto rs each side o f an activ e source. D ata w ere acquired on one p h an to m , fo llo w ed by the other. R aw m ean-tim e d ata w ere ex tracted from both sets o f m easured T P S F s.

Targets:

0:3

Pa = 0.014mm '

®

0 .0 0 6 8 m m ' 0.0073m m ' 0 .9 4 1 m m '' 1.143m m '

|ia image ji's image

Figure 2.3.11 2D im ages o f the breast phantom at the height o f the inclusions, reconstru cted f o r the difference in m ean-tim es between a phantom with structure and an oth erw ise iden tical hom ogenous phantom.

A 2D circu lar FEM m esh w as generated to co rre sp o n d to the correct size o f the sam ple plane (41m m radius), and difference im ages w ere reco n stru cted . T h e 10“^ iteration im ages are

E. M . C . H illm a n . P h D th esis 2 0 0 2 C h a p te r 2 . 3 — 1 3 6

shown in Figure 2.3.11. 10 iterations took 36 minutes using a 400MHz Sun Ultra-10

workstation.

2 .3 .3 .1 .1 Im age sum m ary (2D b reast phantom )

The images in Figure 2.3.11 were published as part of (Hillman et al, 2001c) (submitted

January 2001). Note that the breast phantoms were designed, constructed and measured by

Hylke Veenstra as detailed in (Hebden et al, 2001).

Phantom Breast phantom s

M esh Circular with 3403 nodes, 6615 linear elem ents, 82m m diam eter (2D)

Basis Pixel 24 X 24

Starting parameters: |ia 0 .0 0 7 m m '

Starting param eters: I m m '

Iteration 10

Sources 32

Detectors per source 2 2

D atatypes (raw) Mean

Calibration D ifference betw een B reast - H om og breast

2d 3d correction applied? No

Sim ultaneous Pa and p ' s

A cquisition time (per source) 30 secs

W avelength SOOnm

Table 2 .3 .7 P ro p e rtie s o f im ages sh ow n in F igure 2.3.11

2 .3 .3 .1 .2 D iscu ssion (2D breast phantom )

The images shown in Figure 2.3.11 are fairly free from artefact, suggesting that good

calibration has been achieved through use of the homogenous reference phantom. There is

some distortion in the positions of the inclusions, possibly due to the use of a 2D mesh.

Quantitatively the images are comparable to the 2D multi-slice images shown previously

(Figure 2.3.6Figure 2.3.6). Peak values corresponding to the true inclusion properties are not

retrieved due to blurring and the low iteration (10) chosen. Separation of |ia and |a's is not

complete, probably due to the fact that only mean-time data were used.

Image quality has not deteriorated notably due to the larger size of the object being

imaged (82mm diameter rather than 70mm). Results with this level of parameter separation

and resolution could potentially provide sufficient information for a suspicious lesion in the

breast to be identified (particularly if multi-wavelength acquisitions were performed, see