Determination of Linear X-Ray Attenuation Coeffi cients of Pathological Brain Tissues and Use of Filters in Tissue Contrast Enhancement in Computed Tomography

Objective: X-ray attenuation coeffi cients are used in common radiological, pathological and spectroscopic examinations and in the determination of the radiation dose distribution in biological tissues. In radiology, these coeffi cients enable diagnosis by diff erentiating the abnormal tissues from the normal ones using their morphological structure and contrast diff erences. In this study, our aim is to precisely determine the linear x-ray attenuation coeffi cients of pathological brain tissues and to use x-ray beam fi lters to enhance the tissue contrast in computed tomography. Materials and Methods: To directly measure the relative linear attenuation coeffi cients, an energy dispersive x-ray spectroscopy system (EDXRS-Canberra, Si(Li) with DSA-1000 spectrum analyzer 1998; CT, USA) was used with collimators and a medical-purpose x-ray tube (Siemens, Siremobil, 1985; Erlangen, Germany) in a linear geometry. Results: Using a Mo fi lter with Computed Tomography CT and photon energies from 15 to 25 keV, EDXRS acquisitions were found to signifi cantly distinguish grades of brain tumors (p<0.05). For the data acquired from CT systems with the decreasing fi ltered photon mean energy, the x-ray attenuation coeffi cients (i.e., the Hounsfi eld units) show that the ratio of EDXRS to CT for water’s attenuation coeffi cient are increased. With our suggested x-ray fi lters, the tissue contrast has been found to be increased in ex vivo brain tumor slices compared with slices scanned in conventional CT scanners. Conclusion: X-ray attenuations measured with the EDXRS are found to be statistically more reliable because of the length of acquisition times in this study.


Introduction
Medical imaging, the principal method for noninvasively obtaining anatomical and physiological information about the human body, has experienced considerable advances in technology and clinical applications over the past 25 years [1].X-ray computed tomography (CT) is one modality that has developed in both technique and application.CT imaging of the brain can help in the early diagnosis of malignant tissue abnormalities by distinguishing grades of disease and avoiding invasive interventions into the skull.
The CT x-ray beam spectral shape has been studied by many researchers because the reduction of the radiation dose and an enhancement of the contrast could be achieved in various fields of diagnostic radiology [2].Using different filter combinations, the contrast and dose have been studied in soft tissues using the x-ray tube photon attenuation, both experimental and theoretical [3,4].If the contrast agent enhancement can be increased while using x-ray beam filters, then angiographic examinations may yield more [5].
The spectral shape of the x-ray tube was adjusted by changing the anode and filter materials to obtain maximum image quality and minimum dose delivered to the tissue [6].Image quality is increased and radiation dose is decreased because the photon energy where the detector efficacy is maximized is incident to the detector's surface.A 2-to 3-fold reduction in radiation dose delivered to the tissue has been measured when the x-ray tube spectrum was optimized [7,8].Also, the signal-to-noise ratio (SNR) of acquired images could be optimized using this approach [9].These studies can also be expanded to CT because the x-ray photon attenuation data at different energies also enhances the cross-sectional image quality.Currently, with the help of digital technology, radiodiagnostic image processing and enhancements tend to focus on the spectral analysis and elemental composition of the body region of interest [10].In a separate study, single and dual-energy CT images with monochromatic synchrotron x-rays were compared [11].This technique was used to develop quantitative dual-energy x-ray imaging [12] (e.g., fluorescent x-ray computed tomography (FXCT) with synchrotron radiation).One phantom study reported that FXCT could clearly image the distribution of both iodine and xenon agents, and the contrast ratio was significantly better than that of the transmission CT images [13].To quantitatively select the iodine concentration in the slice, one study's scans used three heavily filtered x-ray beams: two had mean energies that straddle the iodine K-edge (33 keV) and the third energy was slightly higher.The results were independent of tissue and bone attenuation over a broad range of projection path lengths.It was shown for slice diameters up to 30 cm that, to separate iodine from one other material, a two-beam K-edge approach requires less integral dose than a two-beam technique with conventional CT energies.For selective iodine imaging in the presence of more than one other material, the three-spectrum K-edge technique is necessary.The exposure requirements and beam-hardening corrections were discussed in detail, and a computer-simulated CT image, generated by the proposed scheme, was presented in that study [14].Phasecontrast x-ray imaging is a promising technique for observing the structure inside biological soft tissues without the need for staining or serious radiation exposure.Phase-contrast x-ray CT was able to clearly differentiate the cancer lesion from the normal tissue and the fine structures, which correspond to cancerous degeneration and fibrous tissues [15].
Commercial results of these studies are being used to differentiate soft tissues shaded by bone structures.In these systems, dual photon energy techniques produce effective detected energy spectra that result in much lower noise for a given patient radiation dose.[16,17].Some studies have done ex vivo attenuation measurements of x-rays with normal and abnormal breast tissues.It was found that differences between normal and cancerous tissues exist in the linear attenuation coefficients of monochromatic x-rays between 14.15 and 18 keV, but there was some degree of overlap.[18].In a separate study, infiltrating duct carcinomas and fat were well-differentiated by measuring x-ray attenuation.For photon energies used with filmscreen mammography, infiltrating duct carcinomas were found to be more attenuating than fibrous tissue, and above 31 keV, the ranges of attenuation overlapped for the two tissue types [19,20].
As previously reported by various researchers, x-ray spectral analysis in CT is a useful tool for distinguishing pathological soft tissues from normal ones.The materials and methods for exploiting this differentiation are noteworthy.

Materials and Methods
In our study, we used conventional CT scanners to obtain similar results to those from dedicated energy-differentiating scanners.Filters were placed in front of the x-ray tube beam aperture to shape the x-ray beam spectrum.The energy distribution of photons obtained after the filters were applied is shown in Figure 1.
We measured linear attenuation coefficients as Hounsfield Units (HU) by using the CT scanner's software (Toshiba, XVision/GX, 1994; Nasu, Japan; Figure 2), manually evaluating the peak areas of the x-ray tube's filtered continuous x-ray region, and manually evaluating the attenuated peak areas in the exponential attenuation formula.In general, the linear attenuation coefficients were determined by source-sample-detector linear geometry (Figure 3); the x-ray tube source's linear attenuation coefficients of water and paraffin were immersed in tissues (Table 1).Two high purity metal foils and two cellulose matrix pellets were selected as x-ray filters to measure the x-ray tube's continuous energy spectrum.
Experienced pathologists with standard microscopy methods diagnosed and classified pathological brain tissues in three groups: astrocytoma LG (Group 1), astrocytoma G3 (Group 2), and glioblastoma (Group 3).
The linear attenuation coefficients calculations and evaluations were blind to the pathological classification.Statistical tests, including one-way analysis of variance (ANOVA) and Duncan's MRT, were done using SPSS 9.0 software (Chicago, IL, USA).

Results
Although the number of samples was not adequate for an accurate statistical analysis, our results showed similar contrast enhancements of soft tissues as lower x-ray energies.Also, our materials were modified by fixing in paraffin blocks.Water content of these tissues was replaced with paraffin.As seen on Table 1, the linear attenuation coefficient of paraffin was lower than water for all x-ray energies.It was expected that the average linear attenuation coefficient would decrease when a component was replaced by another with a lower density, but the linear attenuation coefficients also increased in contrast.
There was significant (p<0.05)enhancement in brain tumor linear attenuation coefficients (Table 2) of the energy dispersive x-ray spectroscopy system (EDXRS) attenuation measurements for lower energies that were obtained by both physically filtered and computationally filtered continuous x-ray tube spectrums.Compared to tungsten's characteristic x-ray's (W Kα,β) attenuation, the lower energetic x-rays could be used to distinguish two grades of same astrocytoma (Groups 1 and 3).W Kα,β x-rays were less affected by filtering, but these energies were basically used for diagnostic purposes in CT scanners.These energies were able to distinguish Group 1 from Group 3.Although the standard variations of these measurements were small enough, the results may be false because the number of samples was less in Group 3.Although an insignificant finding, Groups 1 and 2 could be distinguished for both energy intervals.

Discussion
Similar results are acquired when the same (Mo, Sn, Ba, Eu) filters are used in conventional CT scanners.The Mo filter's effective photon energy is approximately 18 keV, and at this region of the x-ray tube's spectrum, abnormal soft tissues can be distinguished more significantly (p=0.16)(Table 3).In our Xe detector scanner, the maximum quantum efficiency is near the K-absorption edge of Xe gas, which is about 35 keV.The general background enhancement of slices can be referred to the uncalibrated scanner for specific filters.Our measurements show contrast differences in the same slice, so the differences between groups are significant.
Additionally, it can be inferred from our results that the Ba filter with a cellulose matrix will increase the iodinated agent's contrast enhancement because its effective energy is slightly higher than iodine's absorption edge.However, this should be the subject of another study.
Although there are efforts to add x-ray spectral analysis into conventional imaging units, no commerical prototype currently

Table 1. X-ray beam-shaping filters, the corresponding mean photon energies after filters, and the linear attenuation coefficients of water and paraffin at these energies (WinXCom database)
Filter Mean Energy (MeV) Water μ (cm -1 ) Paraffin μ (cm -

Figure 1 .
Figure 1.Energy distribution of photons obtained with different filters.

Table 3 . Capability of Sn and Mo filters to distinguish study groups with CT scanner data
Duncan's test Type III (error)=2097.669.a-Harmonic mean sample size = 8.800.b-Alpha = 0.05