Preclinical spectral computed tomography of gold nano-particles

Preclinical Spectral Computed Tomography of Gold Nano-Particles Ewald Roessla∗, David Cormodeb†, Bernhard Brendela‡, Klaus J¨ urgen Engelc§, Gerhard Martensa¶, Axel Thranak, Zahi Fayadb∗∗ , Roland Proksaa†† a

Philips Research Europe - Hamburg, Digital Imaging, R¨ ontgenstrasse 24-26, D-22335 Hamburg, Germany b

Translational and Molecular Imaging Institute,

Mount Sinai School of Medicine, One Gustave L. Levy Place, New York, NY 10029, USA. c

Philips Research Europe - Aachen, X-ray imaging systems, Weisshausstr. 2, D-52066 Aachen, Germany

September 27, 2010

1

Abstract

2

Today’s state-of the art clinical computed tomography (CT) scan-

3

ners exclusively use energy-integrating, scintillation detector technol∗ Corresponding author: Tel. +49 40 5078 2965; Fax. +49 40 5078 2510, Email: [email protected] † [email protected] ‡ [email protected] § klaus.j.engel@@philips.com ¶ [email protected] k [email protected] ∗∗ [email protected] †† [email protected]

1

4

ogy, despite the fact that a part of the information carried by the

5

transmitted x-ray photons is lost during the detection process. Room-

6

temperature semiconductors, like CdTe or CZT, operated in energy-

7

sensitive photon-counting mode provide information about the energy

8

of every single x-ray detection event. This capability allows novel,

9

promising approaches to selectively image abnormal tissue types like

10

cancerous tissue or atherosclerotic plaque with the CT modality.

11

In this article we report on recent dual K-edge imaging results ob-

12

tained in the domain of pre-clinical, energy-sensitive photon counting

13

CT. In this approach, the tuning of threshold levels in the detector

14

electronics to the K-edge energy in the attenuation of contrast agents

15

(CA) offers highly specific, quantitative imaging of the distribution of

16

the CA on top of the conventional, morphological image information.

17

The combination of the high specificity of the K-edge imaging tech-

18

nique together with the powerful tool of targeting specific diseases in

19

the human body by dedicated contrast materials might enrich the CT

20

modality with capabilities of functional imaging known from the nu-

21

clear medicine imaging modalities, e.g., positron-emission-tomography

22

but with the additional advantage of high spatial and temporal res-

23

olution. We also discuss briefly the technological difficulties to be

24

overcome when translating the technique to human CT imaging and

25

present the results of simulations indicating the feasibility of the K-

26

edge imaging of vulnerable plaque using targeted gold nano-particles

27

as contrast materials. Our experiments in the pre-clinical domain

2

28

show that dual-K edge imaging of iodine and gold based CAs is fea-

29

sible while our simulations for the imaging of gold CAs in the clinical

30

case support the future possibility of translating the technique to hu-

31

man imaging.

32

Keywords: photon-counting x-ray detectors; K-edge imaging; gold nano-

33

particles; x-ray computed tomography

34

1

35

Recent years have seen increased research efforts in energy-sensitive, photon

36

counting detectors [5,6,8,17] and their application to computed tomography

37

imaging [11, 13–16].

Introduction

38

As today’s commercially available computed tomography scanners exclu-

39

sively employ scintillation detectors operated in current-integrating mode, a

40

part of the information about the attenuation by the object carried by the

41

transmitted flux of x-rays is lost during the detection process. Hence, it is

42

reasonable to ask to what extent the additional information acquired with

43

the help of more advanced detection technology can be used to the benefit

44

of a better clinical diagnosis, the reduction of the dose administered to the

45

patient or to the development of novel CT imaging applications.

46

Intuitively it is clear that the optimal data acquisition in CT would con-

47

sist of a precise measurement of the energies of all individual x-ray photons 3

48

in the transmitted x-ray spectrum, thus operating each detector pixel in

49

spectroscopic mode. In practice, however, it turns out that near-optimal

50

imaging performance can be obtained with much less sophisticated detection

51

techniques [2], because the relative contributions to the total x-ray attenua-

52

tion caused by the photo-electric effect and the Compton effect do not vary

53

strongly for different types of human tissue. For the case of basis material

54

decompositions it has been shown recently that all the information contained

55

in the transmitted spectrum regarding a given N-dimensional basis material

56

decomposition of the object can be retrieved by performing a number of N

57

appropriately weighted spectral measurements. However, the experimental

58

realization of these remains challenging [22]. It has been further shown that

59

the number of energy thresholds required to acquire most of the available,

60

additional information by energy-discriminating photon-counting detectors

61

is rather small [10]: about 4-6 if the energy thresholds are optimized for the

62

imaging task.

63

In this paper we focus on a new application of the energy-selective de-

64

tection of x-rays by photon counting detectors, the so-called spectral K-edge

65

imaging technique. In this technique, a contrast medium is administered

66

having, as is usual for CT, an element with high atomic number as its active

67

component. The energy of the attenuation K-edge is characteristic for the

68

element. By adjusting one of the energy thresholds in the detector to the

69

K-edge of the contrast element, the technique allows to distinguish between

70

the attenuation caused by the contrast medium and the surrounding non4

71

contrasted tissue. Thus, a physical mechanism is provided that allows the

72

selective imaging of the material being administered. The decomposition of

73

the attenuation can be achieved, e.g., with the help of the maximum likeli-

74

hood method, the details of which can be found in [11] and an experimental

75

realization thereof in [13].

76

In this paper we are also considering the situation in which two different

77

contrast media (based on two different elements and, thus, different K-edge

78

energies) are present at the same time. Then, a first of two available energy

79

thresholds is adjusted to the K-edge energy of the first element and a second

80

available energy threshold to the K-edge energy of the second element. Hence,

81

it becomes possible to separate between the attenuation coming from the

82

first contrast agent, the second contrast agent and the surrounding tissue.

83

Note that in the latter case of two contrast agents, at least four energy

84

thresholds are required as the correct modeling of human tissue requires two

85

basis materials [3].

86

We present results obtained in pre-clinical imaging in connection with

87

dedicated, targeted contrast agents based on gold nano-particles. The struc-

88

ture of the paper is the following: in section 2 we briefly describe the scanner

89

hardware on which the reported experiments were performed, as well as the

90

imaging protocol used. In section 3 we summarize our imaging results. In

91

section 4 we present the results of a simulation of a photon-counting hu-

92

man CT system with the detection of vulnerable plaque targeted by gold

93

nano-particles. We will discuss the technological hurdles to be overcome, 5

94

in particular the count-rate problem and possible ideas for mitigations in

95

section 5 and we conclude in section 6.

96

2

97

2.1

98

The experiments we report here were performed with a spectral CT scanner

99

prototype installed at Philips Research in Hamburg. The specification of the

100

scanner have been described elsewhere [13], only the most important infor-

101

mation is summarized here. A KEVEX micro-focus x-ray tube and a planar,

102

single-row, 1024 pixels, 3 mm thick CdTe detector from Gamma Medica -

103

Ideas, Fornebu, Norway were used on a rotating gantry. Spatial resolution is

104

determined by the flexible mounting distances of tube and detector from the

105

iso-center. The effective CdTe pixel height was 1.2 mm and the average pixel

106

pitch was 400 µm, achieving an in-plane spatial resolution of about 100 µm

107

(in the high-resolution arrangement). The electronics performs pulse ampli-

108

fication, shaping and binning of the pulses by the use of six independently

109

adjustable comparators for each of the 1024 pixel.

110

2.2

111

In the pre-clinical dual-K-edge experiment the apo E-KO mouse model of

112

atherosclerosis was used, see [4]. The mice had been on a high-colesterole diet

Materials and Methods Scanner Prototype

Imaging Protocols

6

113

for about 10 months prior to being injected with gold high-density lipoprotein

114

nano-particles (Au-HDL) at a dose of 500 mg Au per kilogram of body weight.

115

For the case shown here, the mice were further injected after 24 hours with

116

150 µl of Fenestra VC (ART Inc., CA), an iodinated bloodpool contrast

117

material for pre-clinical use. The details of the study can be found in [4].

118

The spectral CT imaging parameters were optimized in order to obtain a

119

maximum in specificity with respect to iodine and gold at the same time.

120

This is reflected in the choice of the energy thresholds as follows: 25, 34, 51,

121

80, 91, and 110 keV. Two of the thresholds were set close to the K-edges of

122

iodine at 33.2 keV and gold at 81.0 keV, respectively.

123

The raw data composed of six sinograms, one for each energy bin, were

124

decomposed into a photo-electric effect-, Compton-effect-, iodine- and gold-

125

material basis. The reconstruction of the basis material data sets was per-

126

formed using conventional 2D filtered-back-projection on a reconstruction

127

grid of size 100 × 100 × 200 µm3 . Due to the known noise amplification in

128

the material separation step, the K-edge material images had to be smoothed

129

after reconstruction.

7

130

3

Results

131

3.1

132

In this section we present results of the simultaneous imaging of a non-

133

targeted blood-pool contrast material based on iodine (Fenestra VC) and a

134

macrophage-targeted, high-density lipoprotein nano-particle contrast agent

135

(Au-HDL) [4] as an example of the dual-K-edge imaging technique. Fig. 1

136

shows volume rendered overlays of the photo-effect image showing the bones

137

(grey) the gold material image (yellow) and the iodine material image (red)

138

with an accumulation of gold nano-particles close to the bifurcation of the

139

aorta. In another volume rendered view shown in Fig. 2, we see the uptake

140

of Au-HDL particles in the lymph-nodes of the mouse.

Dual K-edge imaging of Iodine and Gold

141

[Figure 1 about here.]

142

[Figure 2 about here.]

143

In Fig. 3 we show overlays of the gold and iodine data sets above a conven-

144

tional CT image obtained by combining the counts of the six energy windows

145

into one data set and by reconstructing the resulting data.

146

[Figure 3 about here.]

8

147

4

Translation to human imaging

148

In this section we present the results of simulations we performed to inves-

149

tigate the possible translation of the results from the preclinical domain to

150

the domain of human imaging. To be as realistic as possible, we based our

151

simulations on a reconstructed CT volume data-set of a normal-sized human

152

thorax, the central axial slice of which is shown in Fig. 4. [Figure 4 about here.]

153

154

The data set was first low-pass filtered in order to remove the influence

155

of noise in the original reconstructed image and then segmented by a sim-

156

ple thresholding procedure. In this way we obtained a soft-tissue equivalent

157

image and a bone-equivalent image. As the original CT scan was performed

158

using iodinated contrast material, the hyper-dense regions containing iodine

159

were mainly included in this bone-equivalent image. Thus, in these sim-

160

ulations the attenuation coming from the iodine was not modeled as such

161

but was included in the bone-equivalent image. The detection of iodine by

162

its K-edge is not feasible in human CT due to the low energy of the iodine

163

K-edge.

164

In order to simulate a small volume of vulnerable plaque, we added a

165

third basis image set containing only a small cylinder of radius 1 mm and

166

length 10 mm adjacent to one of the blood vessels close to the patients

167

heart. This cylinder contained a 0.1 molar solution of gold on top of a

168

background of water. This concentration was determined by the experience 9

169

gained from the animal experiments an corresponds to about the highest

170

local concentration observed there. At the same time, the image values in

171

the two segmented data sets were put to zero at this very location of the

172

simulated plaque. Then, the three basis data sets were independently forward

173

projected. This was done by determining in a first step the mass density for

174

each voxel in both, the soft-tissue equivalent and the bone-equivalent basis

175

images using the attenuation coefficients denoted Average Soft Tissue male

176

and Skeleton Vertebral Column D6L3 in the ICRU report 46 [1] at an average

177

energy of the simulated spectrum of 71 keV. For this conversion we used:

178

µsof ttissue = 0.0195 mm−1 and µbone = 0.0293 mm−1 . In a second step the

179

basis material densities were then forward-projected to yield complete sets of

180

basis material line-integrals Aj . Finally, the data was spectrally combined to

181

simulate an energy-binned, photon counting detector with 6 energy windows

182

according to the following equation:

λi (Aj ) = Ω IA T

Z∞

−

Φ(E) Si (E) e

3 P

fj (E)Aj

j=1

dE , i = 1, .., 6 .

(1)

0

183

where Ω is the solid angle of the detector pixel, IA the anode tube current

184

and T the time of one integration period. The sensitivity function of the six

185

bins Si (E) were computed using: h

Si (E) =

ZEi

R(E, E ′ )dE ′ .

Eil

10

(2)

186

The detector response function R(E, E ′ ) quantifies the probability of an inci-

187

dent photon of energy E giving rise to an observed energy E ′ . The response

188

functions were obtained on a 1 keV grid by a Monte-Carlo simulation for

189

3 mm thick cadmium-zinc-telluride (CZT) detector material [9]. Fig.

190

shows selected response curves for primary energies of 40, 60, 80, 100 and

191

120 keV. Note that in this simulation we did not normalize the response to

192

unity in order to take into account the case where no interaction takes place.

193

In order to include the effects of electronic noise, the response curves shown

194

in Fig. 5 were smoothed by a Gaussian filter kernel of standard deviation

195

3 keV.

196

5

[Figure 5 about here.]

197

As the source spectrum Φ(E) we followed the model described in [20]

198

with a tube peak voltage of 130 kVp and an additional 2.7 mm of aluminum

199

filtration. The spectrum is depicted in Fig. 6 below.

200

[Figure 6 about here.]

201

Poisson noise was added to the projection data according to an anode tube

202

current of 400 mA. A tube voltage of 130 kVp and a total of 1200 views per

203

turn of the CT gantry and a gantry rotation time of 0.27 s were simulated.

204

The acquisition geometry was assumed to be circular with a detector of

205

64 spectral slices. Physical detector sizes were assumed to be 1.4x1.1mm2 .

206

After a basis-material decomposition using the likelihood method [11], the 11

207

decomposed data sets of the photo-electric effect, the Compton effect and the

208

Au basis-material were reconstructed using a 3D cone-beam reconstruction

209

algorithm described in [7], [21]. The basis material images are shown for one

210

axial slice containing the gold-contrast agent insert in Fig. 7.

211

[Figure 7 about here.]

212

In Fig. 8 the gold basis material image is shown as a colored overlay (yellow)

213

on a quasi-monochromatic image at 60 keV. The latter has been obtained by

214

summing up the contributions of the three material images shown in Fig. 7

215

to the total attenuation of the phantom at the energy of 60 keV.

216

[Figure 8 about here.]

217

A volume rendering of the reconstructed quasi-monoenergetic image at 60 keV

218

and the Au-basis image as a yellow color overlay is shown in Fig. 9. The win-

219

dowing in the rendering display for the 60 keV CT image was chosen such

220

that only the contrasted regions with large attenuation were visible.

221

[Figure 9 about here.]

222

It can be seen from the image that even for the rather moderate tube loading

223

of about 110 mAs, the selective detection of gold can be achieved by means

224

of K-edge imaging even in the case for a human Thorax, assuming that an

225

accumulation of the contrast medium to a local concentrations of 0.1 molar

226

can be reached.

12

227

Despite the fact that we tried to simulate a scenario as realistic as possi-

228

ble, we made certain assumptions that are hardly met in practice. The most

229

important shortcoming of our simulation is that it did not include effects of

230

pulse-pileup at the very high rates employed nowadays in computed tomog-

231

raphy scanners. A comparison of typical x-ray fluxes for three common x-ray

232

imaging modalities is given in Tab. 1 for mammography, radiography and

233

computed tomography [9]. As can be seen, the x-ray flux in computed to-

234

mography is highest. The count-rate performance of state-of-the art photon-

235

counting detectors is summarized in Tab. 2 below [17] together with typical

236

ASIC dimensions and the derived maximally achievable count-rates per de-

237

tector surface area.

238

[Table 1 about here.]

239

[Table 2 about here.]

240

5

Discussion

241

As can be seen from the above tables, there remains a gap between the

242

rates that currently appear handleable and the actual rates in CT scanners

243

of about one order of magnitude. Note, however, that the rates given in

244

Fig. 1, refer to the rates in the unattenuated primary beam. In order to

245

homogenize the x-ray flux onto the detector in the presence of a patient,

246

so-called bow-tie filters are used that attenuate the beam stronger at larger 13

247

fan-angles than at smaller fan-angles (close to the central beam). This has

248

the effect of lowering the maximal rates of photons hitting the detector. In

249

application to photon-counting CT the design of the bow-tie filters will have

250

to be reconsidered under the aspects of the reduced dynamic range of photon

251

counting detectors.

252

Another way to mitigate the count rate problem for human photon-

253

counting CT is the use of detector pixel sizes that are well below the sizes

254

currently used in CT. The aim here is not to gain in spatial resolution but

255

primarily to distribute counts on as many different readout channels as pos-

256

sible [19], [12]. This idea is already reflected in the comparison shown in

257

Tab. 2. Of course this approach is limited by the degrading spectral detec-

258

tor performance due to cross-talk between neighboring pixel, an effect that

259

increases with decreasing pixel size.

260

We should also mention here continuous efforts to reduce the dose ad-

261

ministered to the patient in CT scans for example by means of iterative

262

image reconstruction techniques. This development would benefit the feasi-

263

bility of photon-counting CT as the reduced dose translates to scanning at

264

reduced anode tube current, hence to a reduction of the photon flux hitting

265

the detector. Finally, we note that the effect of pulse pile-up is particularly

266

severe at high-rates. Not only does it complicate the task of registering all

267

of the incoming photons but it also will degrade spectral performance. The

268

superposition of pulses leads to missing events and to systematic errors in

269

the determination of the photon energies. Here, sophisticated pile-up models 14

270

tailored to the particular realization of the detector electronics will help to

271

correctly compute the expected number of counts in a certain energy bin

272

in the presence of first, and higher order pulse pileup. See [18] for recent

273

progress on the subject.

274

6

275

We have demonstrated the feasibility of simultaneously imaging the anatomy,

276

as well as two different contrast materials based on iodine and gold by dual K-

277

edge imaging in the case of pre-clinical imaging. The differentiation between

278

iodine and gold was achieved by tuning two of the available energy thresh-

279

olds to the K-edge energies of iodine at 33.2 keV and of gold at 81.0 keV,

280

respectively. We further investigated the possibility of translating the K-edge

281

technique to human imaging. Our results indicate that at least for the imag-

282

ing of gold as a contrast material, we obtain sufficient sensitivity to separate

283

the Au-targeted plaque structure from the hyper-dense surrounding regions,

284

under the assumption of a gold concentration of 0.1 molar in the vulnera-

285

ble plaque structure. In future work we plan to improve our simulations by

286

taking into account the combined effects of a physical detector response and

287

pulse-pileup.

Conclusions

15

288

Acknowledgment

289

We thank the following people for their support and very helpful discus-

290

sions: Roger Steadman, Christoph Herrmann, Oliver Muehlens (Philips Re-

291

search Europe, Aachen, Germany), Thomas Koehler (Philips Research Eu-

292

rope, Hamburg, Germany) and Naor Wainer and Ami Altman (CT Global

293

Research and Advanced Development, Philips Healthcare, Haifa, Israel).

294

References

295

[1] Photon, electron, proton, and neutron interaction data for body tissues.

296

ICRU report 46. International Commission on Radiation Units and

297

Measurements, 1992. Bethesda, MD.

298

299

[2] R. E. Alvarez. Near optimal energy selective x-ray imaging system performance with simple detectors. Med. Phys., 37(2):822–841, Feb 2010.

300

[3] R. E. Alvarez and A. Macovski. Energy-selective reconstructions in x-ray

301

computerized tomography. Phys. Med. Biol., 21(5):733 – 744, 1976.

302

[4] D. P. Cormode, E. Roessl, A. Thran, T. Skajaa, R. E. Gordon, J.-

303

P. Schlomka, V. Fuster, E. A. Fisher, W. J. M. Mulder, R. Proksa,

304

and Z. A. Fayad. Atherosclerotic plaque composition: analysis with

305

multicolor CT and targeted gold nanoparticles. Radiology, 256(3):774–

306

782, Sep 2010.

16

307

[5] P. Fischer, A. Helmich, M. Lindner, N. Wermes, and L. Blanquart. A

308

Photon Counting Chip with Energy Windowing. IEEE Trans. Nucl.

309

Sci., 47(3):881–884, 2000.

310

[6] J. S. Iwanczyk, E. Nygard, O. Meirav, J. Arenson, W. C. Barber, N. E.

311

Hartsough, N. Malakhov, and J. C. Wessel. Photon counting energy

312

dispersive detector arrays for x-ray imaging. Nuclear Science, IEEE

313

Transactions on, 56(3):535 –542, jun. 2009.

314

[7] M. Kachelrieß, M. Knaup, and W. A. Kalender. Extended parallel back-

315

projection for standard three-dimensional and phase-correlated four-

316

dimensional axial and spiral cone-beam CT with arbitrary pitch, arbi-

317

trary cone-angle, and 100% dose usage. Med. Phys., 31(6):1623 – 1641,

318

2004.

319

[8] E. Kraft, P. Fischer, M. Karagounis, M. Koch, H. Kr¨ uger, I. Peric,

320

N. Wermes, C. Herrmann, A. Nascetti, M. Overdick, and W. R¨ utten.

321

Counting and integrating readout for direct conversion x-ray imaging.

322

Concept, realization and first prototype measurements. IEEE Trans.

323

Nucl. Sci., 54(2):383–390, 2007.

324

[9] M. Overdick, C. Baumer, K. J. Engel, J. Fink, C. Herrmann, H. Kruger,

325

M. Simon, R. Steadman, and G. Zeitler. Status of direct conversion

326

detectors for medical imaging with X-rays. 56(4):1800–1809, 2009.

17

327

[10] E. Roessl and R. Proksa. Optimal energy threshold arrangement in

328

photon-counting spectral x-ray imaging. In Proc. IEEE Nuclear Science

329

Symp. Conf. Record, volume 3, pages 1950–1954, 2006.

330

[11] E. Roessl and R. Proksa. K-edge imaging in x-ray computed tomog-

331

raphy using multi-bin photon counting detectors. Phys. Med. Biol.,

332

52(15):4679–4696, Aug 2007.

333

[12] E. Roessl, J. P. Schlomka, and R. Proksa. Edge-on semiconductor x-ray

334

detectors - towards high-rate counting computed tomography. In Proc.

335

IEEE Nuclear Science Symp. Conf. Record NSS ’08, pages 1748–1751,

336

2008.

337

[13] J. P. Schlomka, E. Roessl, R. Dorscheid, S. Dill, G. Martens, T. Is-

338

tel, C. Bumer, C. Herrmann, R. Steadman, G. Zeitler, A. Livne, and

339

R. Proksa. Experimental feasibility of multi-energy photon-counting K-

340

edge imaging in pre-clinical computed tomography. Phys. Med. Biol.,

341

53(15):4031–4047, Aug 2008.

342

343

[14] P. M. Shikhaliev. Energy-resolved computed tomography: first experimental results. Phys. Med. Biol., 53(20):5595–5613, Oct 2008.

344

[15] P. M. Shikhaliev. The upper limits of the SNR in radiography and

345

CT with polyenergetic x-rays. Phys. Med. Biol., 55(18):5317–5339, Aug

346

2010.

18

347

[16] P. M. Shikhaliev, T. Xu, and S. Molloi. Photon counting computed

348

tomography: concept and initial results. Med. Phys., 32(2):427–436,

349

Feb 2005.

350

[17] R. Steadman, C. Herrmann, O. Mulhens, D. G. Maeding, J. Colley,

351

T. Firlit, R. Luhta, M. Chappo, B. Harwood, and D. Kosty. Chro-

352

mAIX: a high-rate energy-resolving photon-counting ASIC for spectal

353

computed tomography. Proc. of SPIE Medical Imaging Conference,

354

7622(1):762220, 2010.

355

[18] K. Taguchi, E. C. Frey, X. Wang, J. S. Iwanczyk, and W. C. Barber.

356

An analytical model of the effects of pulse pileup on the energy spec-

357

trum recorded by energy resolved photon counting x-ray detectors. Med.

358

Phys., 37(8):3957–3969, 2010.

359

[19] J. E. Tkaczik, J. D. Short, J. W. Rose, X. Wu, and S. K. Basu. Photon

360

counting x-ray detector with overrange logic control, 2007. US Patent

361

No. 2007/0206721 (A1).

362

[20] D. M. Tucker, G. T. Barnes, and D. P. Chakraborty.

Semiempiri-

363

cal model for generating tungsten target x-ray spectra. Med. Phys.,

364

18(2):211–218, 1991.

365

366

[21] H. Turbell. Cone-beam reconstruction using filtered backprojection. Technical Report Thesis No. 672, Link¨oping University, 2001.

19

367

[22] A. S. Wang and N. J. Pelc. Sufficient statistics as a generalization of

368

binning in spectral X-ray imaging. IEEE Trans. Med. Imag., Aug 2010.

20

369

370

List of Figures 1

371 372 373 374

2

375 376 377 378

3

379 380 381 382 383

4

384 385 386

5

387 388 389 390

6

391 392

7

393 394 395 396 397 398 399 400 401 402

8

Volume rendering of photo-effect basis material image (grey), the iodine basis material image (red) and the gold basis material image (yellow) showing the accumulation of Au-HDL particles close to the bifurcation of the aorta. . . . . . . . . Volume rendering of photo-effect basis material image (grey), the iodine basis material image (red) and the gold basis material image (yellow) showing the uptake of Au-HDL in four lymph nodes. . . . . . . . . . . . . . . . . . . . . . . . . . . Color overlay of iodine basis material image (red) and gold basis material image (yellow) on an image obtained by the addition of the six binned photon counting measurements into a single CT image simulating a conventional, non-spectral CT image. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conventional CT image of a human Thorax serving as the starting point for the simulations. Central slice. L/W settings: 100HU/1500HU. . . . . . . . . . . . . . . . . . . . . . . . . Detector response functions obtained by a Monte-Carlo simulation shown for primary energies of 40, 60, 80, 100 and 120 keV. The response was computed for 3 mm of cadmium-zinctelluride (CZT) material. . . . . . . . . . . . . . . . . . . . X-ray tube source spectrum used for the simulation following [20] with an additional filtration of 2.7 mm of aluminum. . Basis material images of the Photo-electric effect (a), the Compton effect (b) and gold (c) obtained by reconstructing the decomposed data sets obtained after a pixel-wise numerical maximization of the likelihood function for the measured binned photon counting data. Only one slice containing the vulnerable plaque insert is shown. The images are displayed on a linear grey scale with black/white corresponding to the minimal/maximal values appearing in the images. . . . . . Colored overlay of the gold basis material image shown in yellow over the mono-energetic image at 60 keV. The plaque structure is indicated by the white arrow. . . . . . . . . . .

21

. 23

. 24

. 25

. 26

. 27 . 28

. 29

. 30

403 404 405

9

Volume rendering of the Au data set together with the rendering of the conventional CT image. The vulnerable plaque is shown in a shade of gold and indicated by the white arrow.

22

31

Figure 1: Volume rendering of photo-effect basis material image (grey), the iodine basis material image (red) and the gold basis material image (yellow) showing the accumulation of Au-HDL particles close to the bifurcation of the aorta.

23

Figure 2: Volume rendering of photo-effect basis material image (grey), the iodine basis material image (red) and the gold basis material image (yellow) showing the uptake of Au-HDL in four lymph nodes.

24

Figure 3: Color overlay of iodine basis material image (red) and gold basis material image (yellow) on an image obtained by the addition of the six binned photon counting measurements into a single CT image simulating a conventional, non-spectral CT image.

25

Figure 4: Conventional CT image of a human Thorax serving as the starting point for the simulations. Central slice. L/W settings: 100HU/1500HU.

26

Response of 3 mm CZT 0.10

40keV

0.08

60keV 80keV

0.06 100keV 0.04

120keV

0.02

20

40

60

80

100

120

Emeasured @keVD

Figure 5: Detector response functions obtained by a Monte-Carlo simulation shown for primary energies of 40, 60, 80, 100 and 120 keV. The response was computed for 3 mm of cadmium-zinc-telluride (CZT) material.

27

photon fluence [photons/(mAs sr keV)]

1.4e+11 1.2e+11 1e+11 8e+10 6e+10 4e+10 2e+10 0 0

20

40

60

80

100

120

140

E[keV]

Figure 6: X-ray tube source spectrum used for the simulation following [20] with an additional filtration of 2.7 mm of aluminum.

28

(a)

(b)

(c) Figure 7: Basis material images of the Photo-electric effect (a), the Compton effect (b) and gold (c) obtained by reconstructing the decomposed data sets obtained after a pixel-wise numerical maximization of the likelihood function for the measured binned photon counting data. Only one slice containing the vulnerable plaque insert is shown. The images are displayed on a linear grey scale with black/white corresponding to the minimal/maximal values appearing in the images. 29

Figure 8: Colored overlay of the gold basis material image shown in yellow over the mono-energetic image at 60 keV. The plaque structure is indicated by the white arrow.

30

Figure 9: Volume rendering of the Au data set together with the rendering of the conventional CT image. The vulnerable plaque is shown in a shade of gold and indicated by the white arrow.

31

406

407

List of Tables 1

408 409 410

2

X-ray flux rates for three different x-ray imaging modalities in comparison taken from [9]. . . . . . . . . . . . . . . . . . . . 33 Count-rate performance of state-of-the-art photon-counting xray detectors in comparison, see [17] for details. . . . . . . . . 34

32

Table 1: X-ray flux rates for three different x-ray imaging modalities in comparison taken from [9].

33

Table 2: Count-rate performance of state-of-the-art photon-counting x-ray detectors in comparison, see [17] for details.

34