Phase-contrast imaging is investigated in order to improve the phase-imaging protocol as well as enhancing contrast in radiographic images. Regarding research purposes, a goal of phase-contrast imaging would be to enable the simple use in laboratory settings with reasonable acquisition time and without complicated setups. On the other hand, one goal of studying phase-contrast imaging is to introduce its use into clinical settings. The research so far has shown that the technique is not suitable or does not compare with absorption imaging for all purposes, but especially in applications where primarily soft tissue is imaged, it could provide images with higher contrast. One such promising application is mammography, where diagnoses suffer from a high rate of false positives and false negatives due to poor absorption contrast .
Perhaps the most critical limitation of phase-contrast imaging is the suitability of X-ray sources for the application. As the propagation-based methods and grating interferome-try require high transverse coherence of the beam and crystal interferomeinterferome-try as well as analyzer-based methods (e.g. diffraction-enhanced imaging) require longitudinal coherence or monochromaticity, clinical X-ray sources by themselves are not suitable for the applica-tion. Sufficient coherence can typically be obtained using conventional sources with the use of, for example, gratings but at the expense of imaging time, which are then not suitable for clinical settings. Coherent radiation can be obtained at large synchrotron facilities with which many of the research regarding phase-imaging has been done, but is not suitable for clinical settings. 
Due to the very different setups of the various phase-imaging techniques, it is difficult to compare them. As the requirements for the setups are fairly high, many facilities can only conduct experiments using one technique and therefore different techniques cannot be compared using for example the same X-ray source. A brief analysis of the methods and their limitations is provided.
centimeters as well as the FOV. Regarding the X-ray beam, it needs to be highly monochro-matic and intense. On the other hand, analyzer-based imaging has been conducted on samples of size 15 cm. One crystal can be used for a wide energy range but the beam requirements are the same as for crystal interferometry, only available at synchrotron radiation sources.
Propagation-based imaging is the easiest to implement as it requires no additional optical devices. On the other hand, propagation-based imaging has source-size limitations in order to achieve good spatial coherence and a sufficient sample-to-detector distance is required for the signal to diffract. The technique is compatible with polychromatic radiation and an image can be reconstructed from the intensity profile generated by the sample or phase retrieval can be implemented to achieve a more accurate map of the electron density in the sample.
Grating interferometry, which has been studied extensively during past years, shows great potential as a phase-contrast technique but has limitation regarding its implementation into clinical use. The fabrication of gratings required for the setup has progressed so that a FOV of 5 cm has been obtained. For clinical applications, this should be even larger. As gratings are manufactured for a narrow energy range, imaging with multiple energies would require a kit of gratings in the clinical settings. For a fast imaging procedure, a sufficiently coherent X-ray beam is required or a source grating needs to be used. Phase-stepping is most commonly used to separate phase and attenuation information but alternative methods are developed as the procedure is time consuming and causes additional radiation to the sample to be imaged.
4 Study Objective
To achieve good contrast with reasonable exposure time, propagation-based phase imaging requires an X-ray source with a small focal spot size and low energy photons. In addition, propagation-based phase imaging requires spatial coherence. This is partially achieved by increasing the source-to-sample distance as well as by decreasing the focal spot size. Given the obvious limitations of laboratoryµCT devices with respect to phase imaging, the objective of this study is to optimize the phase imaging protocol on the Zeiss Xradia MicroXCT-400 device using thin medical grade polylactic acid (PLA) fibers.
The optimization is performed by acquiring projection images of the fibers embedded in ethanol. There are four main variables that are varied in order to optimize the phase imaging protocol. These variables are source and detector distances, the power output, and voltage.
The acquired projection images are then analyzed for optimal contrast and the final results are applied on a concrete application, collagen samples embedded in air, water, and ethanol.
This is of interest as these collagen samples embedded in water are known to absorb X-rays poorly, resulting in poor contrast.
5 Materials and Methods
Phase-contrast imaging on a laboratory device has been previously studied at other facilities  but as it had been utilized only a couple of times at our facility, some initial testing was conducted before the start of the actual optimization measurements. The testing was conducted with medical grade PLA filaments, which are used in 3D printing. As the material has poor absorption properties, due to its density and atomic number, and the visibility can be enhanced with the use of phase imaging, the materials chosen for the optimization process were thin medical grade PLA fibers. The fibers were chosen because of the simple structure that makes the analysis less complicated.
Seven different diameters for the PLA fibers were included in the optimization to visualize the phase effect with decreasing structure size. The chosen fibers had cross-section lengths of 44, 46, 56, 70, 80, 89, and 107µm of which the 46, 56, and 70µm fibers had a star-shaped cross section. The fibers had been manufactured previously by a process called melt spinning . Briefly, polymer granules are placed in a feeder, which feeds the granules to the extruder at a constant rate. The polymer is melted at high temperatures and is then forced through the extruder. After the extrusion, the fiber is cooled down. A more detailed description of the equipment used and the melt spinning process can be found in .
Briefly, the optimization process consisted of four main variables that were optimized.
These were source distance, detector distance, power output, and the voltage settings. From the source distance optimization measurements, three of the potentially interesting and useful distances were chosen for detector distance and power output variation measurements. All of the measurements for one variable were conducted for all fiber sizes. After all measurements were conducted for one variable, the results were analyzed and based on the analysis, the measurements for the next variable were conducted. Finally, CNR’s were calculated from the obtained results. The work flow of the optimization process is presented in Figure 10.
Table 1 summarizes the reasons for the chosen variables to be included in the optimization process.
Figure 10: The optimization work flow depicted as a block diagram. One blue block depicts one optimization variable in the process. After all measurements for one variable were con-ducted, fiber visibility analysis was performed and based on the results, the measurement settings for the next variable were selected. Contrast-to-noise ratios were also calculated from results obtained from all variables.
Table 1: A table summarizing the four main variables included in the optimization process along with the explanation as to why the variable is included in the optimization.
Optimization Variable Reason for Including in Optimization Source Distance
beam results in an enhanced phase effect
source-to-sample distance. Spatial coherence of the X-ray Spatial coherence length increases with increasing
and magnifies the image.
the pixel size decreases with increasing detector distance sample, which results in the edge-enhancing effect. Additionally,
The refracted waves interfere at some distance away from the
This should theoretically enhance the phase effect.
5µ when the power output is decreased to 4W.
focal spot size and the focal spot size decreases to The spatial coherence length increases with decreasing
amount of phase contrast visible.
voltage, an increase in voltage should decrease the wavelength. Since wavelength decreases with increasing
The spatial coherence length increases with increasing
5.1 Imaging Setup and Source Variation Measurements
The imaging setup of one PLA fiber consisted of the PLA fiber inserted into a small 1ml syringe filled with ethanol and sealed with Parafilm. The syringe was then clamped on a metal holder, which can be placed on the sample stage in the µCT. An depiction of the imaging setup and sample is in Figure 11. Due to the biodegradability of the PLA fibers, the fibers were changed daily.
The optimization process began by first imaging all of the PLA fibers of different sizes with varying source distances and constant detector distance on the Zeiss XRadia microXCT-400 device using a magnification of 4x. All projection images and tomographic datasets were acquired with binning 2 in order to decrease the required exposure time. Other imaging parameters were kept constant except for exposure time, which was increased with increasing source distance to achieve required image quality. The exposure time was selected for all images so that the counts in the center of the image were above 5000.
The imaging parameters for the first imaging sequence can be found in Table 2. Reference images were acquired for all images so that the sample was removed from the field of view and 10 images were taken with the same imaging parameters as in the actual imaging of the sample. Those 10 images were averaged by the software and the reference image was then applied to the actual image. When the reference image is applied to the projection image, the intensity values in the projection image is divided by the intensity values of the corre-sponding pixels in the reference image. The resulting value for each pixel is the percentage of transmitted X-rays or the transmission value. The software used for the operation of the µCT device and the imaging as well as the application of reference images was done using XMController.
Figure 11: Imaging setup of the PLA fibers with the fiber immersed in ethanol in the syringe.
The syringe is clamped onto the holder that fits on the sample stage inside theµCT device.
The black object on the left side is the X-ray source and the round object behind the sample is the objective.
Table 2: Imaging parameters for the first imaging sequence with varying source distance, conducted for all fibers.
Image number 1 2 3 4 5 6
Voltage (kV) 40 40 40 40 40 40
Current (µA) 250 250 250 250 250 250
Power (W) 10 10 10 10 10 10
Source Position (mm) 40 90 160 230 300 370 Detector Position (mm) 10 70 70 70 70 70
Exposure Time (s) 1 15 30 50 65 85
Voxel Size (µm) 5.42 3.81 4.71 5.19 5.49 5.69