Abstract
1352
Objectives: Bench-top x-ray fluorescence computed tomography (XFCT) is an emerging imaging modality that could image non-labeled metal elements in 3-D objects, such as lab animals or even patients. However, the sensitivity of bench-top XFCT has been limited primarily by two physical factors — (a) the intrinsically small photoelectric absorption cross-sections of metal elements to X-ray photons, and (b) the down-scattered Compton scattering background superimposed on the fluorescence X-ray signals. These two fundamental factors, along with the imperfections in XFCT detection systems (e.g. finite energy resolution and detection efficiency), limit the sensitivity of bench-top XFCT to the order of 1 mg/mL level. In this study, we explore several techniques that minimize the Compton background in XF signals, which could significantly improve the sensitivity of XFCT. In this presentation, we demonstrate that combining X-ray filters, fine-tuned X-ray energy and backscattering data acquisition geometry, which could allow us to acquire X-ray fluorescence (XF) signal with near zero Compton background. This would allow us to push the sensitivity of XFCT imaging by an order of magnitude and therefore open up the door for potential in vivo study of trace metals in lab animal.
Methods: Noise reduction. According to Eq.1, the increase of detection angle (i.e. detector locates closer to the beam source) results in a shift of the spectrum towards lower energies. As shown in Fig.1, by using 90kVp X-ray source, the detection of Pt-Kα,1 measured at 150° has a large advantage than other detection angle, due to the low noise counts. As demonstrated in Fig.2, the incident spectrum (grey line) is reshaped to a filtered spectrum (blue line) by filter K-edge, remaining a tail part of the spectrum. After backscattering detection, the detected spectrum (orange line) moves to the lower energy side comparing with excitation spectrum. If the excitation spectrum could ideally cut off the tail part by the filter, the noise around the fluorescence signal would be close to zero. Using a simulated 110kVp X-ray source spectrum (Fig.3a), Fig.3b shows the shape of filtered spectrum by a 0.3mm Au filter. Fig.3c shows the corresponding detected spectrum measured at 150°. Fig.4 compares the detected spectrum without and with filter (0.2mmAu), measuring at 150°. The signal (S) and noise (N) were both decrease after using the filter. However, the value of S/N increases. Although it is hard that ultimately getting rid of the tail part of the excitation spectrum, this method provides a very useful indication to minimize noise. Monte Carlo studies. A Geant4-based Monte Carlo (MC) model of XFCT system was developed (Fig. 5).
Results: SNR was presented in Eq. 2. Fig.6 shows the detected spectrum with 0.3 mm Au filter (left) and 0.3 mm Tl filter (right). The Pt fluorescence signals (66.8KeV) showed at spectrum tail where the noise was relatively low. The cutting edge of the detected spectrum was about 63.6 KeV for Au filter and 65.8 KeV Tl filter. Considering the detection efficiency, it is prefer to choose Au filter for this XFCT configuration. Tab.2 lists preliminary results without and with filter (0.3 mm Au and 0.3 mm Tl). Comparing with no filtering, adding a filter increases the S/N by a factor of 3.05 for Au filter, and a factor of 2.14 for Tl filter. In the term of dose-normalized SNR, the configuration with filter (both 0.3 mm Au and 0.3 mm Tl) improved the sensitivity by an order-of-magnitude.
Conclusions: The preliminary MC results showed that SNR has a strong correlation with the filter selection at backward detection angle. In the case of excitation of Pt solution using a 110 kVp X-ray tube, an order-of-magnitude SNR improvement was reached by choosing a 0.3 mm Au filter at a detection angle of 150°. Optimization of filter and detection angle presents a promising effect of noise reduction, which could be taken seriously in the design of future XFCT imaging systems.