Summary

Materials and Methods

Seung Ho Kima, Daecheon Kima, Hanbean Younb, Seungryong Choc, and Ho Kyung Kima,d*

aSchool of Mechanical Engineering, Pusan National University, Busan 46241, South KoreabDepartment of Radiation Oncology, Pusan National University Yangsan Hospital, Kyungsangnam-do 50612, South Korea

cDepartment of Nuclear and Quantum engineering, Korea Advanced institute of Science and Technology, Daejeon 34141, South KoreadCenter for Advanced Medical Engineering Research, Pusan National University, Busan 46241, South Korea

* Corresponding author: hokyung@pusan.ac.kr

Bone-Enhanced Small-Animal MicrotomographyWith Single-Shot Dual-Energy Sandwich Detectors

Poster # 042 2016 CT Meeting, Bamberg, Germany

This work was supported by the National Research Foundation of Korea (NRF) grant funded

by the Korea government (MSIP) (No. 2013M2A2A9046313 and No. 2014R1A2A2A01004416).{http://bml.pusan.ac.kr}Radiation Imaging Laboratory

Research Background

Results

Summary

Research Objective Active multi-layer (“sandwich”) detector can provide “motion-artifact-free” single-shot dual-energy images

DE operation (i.e. subtraction of images) increases image noise

• Measured single-shot DE-CT noise is worse than that estimated double-shot noise model, and which implies

that there exists additional noise due probably to scattered x-ray photons, which are inevitable in the

sandwich structure, and fluorescence x-ray photons from the Cu filter

Single-shot DE-CT imaging can enhance the contrast-transfer ability at non-zero spatial frequencies

• Crude linear approximation shows the enhancement of MTF values at non-zero spatial frequencies

• Demonstration DE bone images of a postmortem mouse well reflect this characteristic

• Adversely, this enhancement causes black-shadow artifacts around edges

This study watches for a potential of microtomography with a sandwich detector for high-resolution

bone-enhanced small-animal imaging without use of elaborate equipment such as micro-focus x-ray

source and high-resolution x-ray imaging detectors

Linear approximation of

single-shot DE-CT imaging

Projection images

Microtomography system

with sandwich detectors

3D Noise power spectrum

To investigate the feasibility of micro-CT with sandwich

detectors for small-animal bone studies

• Apply the sandwich detector to the micro-CT system

• Evaluate the imaging characteristics (MTF, NPS, and NEQ) of

the single-shot DE micro-CT system

• Demonstrate the postmortem mouse images of the single-shot

DE micro-CT system

High-resolution front detector measures low-E

Low-resolution rear detector measures high-E

X-ray photons with different energies

Two detectors constituting the sandwich detector use

different thick phosphors; hence the different spatial-

resolution properties in the corresponding images

Therefore, subtraction of two images obtained from the

sandwich detector can boost up the high spatial-

frequency information in the resultant dual-energy image

X-ray source (50 kVp) Target W

Focal-spot size 0.035 mm

Added filter 1 mm Al

Sandwich detector Front phosphor ~34 mg cm-2 Gd2O2S:Tb

Rear phosphor ~67 mg cm-2 Gd2O2S:Tb

CMOS PD arrays 5121024 pixels with 0.048 mm

Intermediate filter 0 or 0.3 mm Cu

Geometry 𝑑𝑆𝐴 550 mm

𝑑𝑆𝐷 600 mm

Object Postmortem mouse ~40 g

Reconstruction Algorithm FDK with the Hann filter

Performance Evaluation

Projections with the assumption of linear logarithmic

operation for an impulse signal

• 𝑝𝑗 𝐱 = 𝜙𝑗𝜆𝑗𝑔𝑗 𝐱 = 𝜙𝑗𝜆𝑗 𝑔𝑗𝐿𝑗 𝐱

– 𝜙𝑗 incident x-ray photon signal

– 𝜆𝑗(= 𝑑𝑗−1) gain describing the logarithmic operation

(the reciprocal of the pixel value)

– 𝑔𝑗 𝐱 spatial detector response with an LSF, −∞

∞𝐿𝑗(𝐱)d2𝐱 = 1

Filtered projections

• 𝑞𝑗 𝐱 = 𝜙𝑗𝜆𝑗 𝑔𝑗𝐿𝑗(𝐱) ⊗ ℎ(𝐱)

Backprojection of the filtered projections

• 𝑓𝑗 𝐱 = 𝑎 0

2𝜋𝑞𝑗 𝐱 d𝜃 ≈ 𝑏 𝜙𝑗𝜆𝑗 𝑔𝑗𝐿𝑗

′ 𝐱

– 𝑎 & 𝑏 backprojection normalization constants

– 𝐿𝑗′ 𝐱 smeared LSF and −∞

∞𝑏𝐿𝑗

′ 𝐱 d2𝐱 = 1

MTF of the FBP image

• MTF𝑗 𝐮 = FT 𝑓𝑗 𝐱

−∞∞ 𝑓𝑗 𝐱 d2𝐱

≈ 𝑏𝑇𝑗 𝐮

DE projections for an impulse signal

• 𝑝𝐷𝐸 𝐱 = 𝑤𝑝𝐹 𝐱 − 𝑝𝑅 𝐱 = 𝑤 𝜙𝐹𝜆𝐹 𝑔𝐹𝐿𝐹(𝐱) − 𝜙𝑅𝜆𝑅 𝑔𝑅𝐿𝑅(𝐱)

Numerical simulation results with Gaussian 𝐿𝑗(𝑥) DE NPS with the assumption of

the independent process between

front & rear detectors

DE NEQ

MTF of the single-shot FBP DE image

• MTF𝐷𝐸 𝐮 = FT 𝑓𝐷𝐸 𝐱

−∞∞ 𝑓𝐷𝐸 𝐱 d2𝐱

= 𝑤 𝜙𝐹𝜆𝐹 𝑔𝐹MTF𝐹 𝐮 − 𝜙𝑅𝜆𝑅 𝑔𝑅MTF𝑅 𝐮

𝑤 𝜙𝐹𝜆𝐹 𝑔𝐹− 𝜙𝑅𝜆𝑅 𝑔𝑅

• If 𝑔𝑗 =𝜕 𝑑𝑗

𝜕 𝜙𝑗,MTF𝐷𝐸 𝐮 =

𝑤MTF𝐹 𝐮 −MTF𝑅 𝐮

𝑤−1

FYI, the MTF of the double-shot FBP DE image

• If MTF𝑙𝑜𝑤 𝐮 and MTFℎ𝑖𝑔ℎ 𝐮 are similar to each other,

MTF𝐷𝐸 𝐮 ≈ MTF𝑙𝑜𝑤 𝐮 or MTFℎ𝑖𝑔ℎ 𝐮

• NPS𝐷𝐸 𝐮 = 𝑤2NPS𝐷𝐸 𝐮 +NPS𝐷𝐸 𝐮

• NEQ𝐷𝐸 𝐮 = 𝜋𝐮MTF𝐷𝐸

2 (𝐮)

NPS𝐷𝐸 𝐮

Image reconstruction

Image performanceVolume images

DE-CT LSF shows a typical edge-enhancement characteristic

The LSF characteristic is well reflected into the resultant DE-

CT MTF

Linear MTF model agrees well with the numerical simulation

There exists discrepancy with the MTF directly obtained from

the DE-CT LSF images, which are very noisy, and the

discrepancy is probably due to the measurement errors

Linear MTF model describes that the detector design (i.e., Cu-

filter thickness) dose not affect the resultant MTF

characteristics

The measured DE-CT NPS is much worse than that obtained

from the front or rear detector as well as that estimated from

the conventional DE NPS model

There exists an additional noise, except the front and rear

detector noise and noise due to subtraction operation, and it

could be due to scattered and fluorescence x-ray photons

The use of intermediate filter largely increases the rear

detector NPS; hence the DE NPS

The DE-CT MTF shifts the NEQ peak to higher frequencies,

and enhances NEQ performance

Cross-sectional images

u

w

u

v

uNy uNy uNy

2×uNy