Performance of X-Ray Photon-Counting Scintillation Detectors Under Pile-Up Conditions at 60 keV

We investigate silicon photomultiplier (SiPM)-based scintillation detectors for medical X-ray photon-counting applications, where the input count rate (ICR) can reach a few Mcps/mm2 in cone-beam CT for radiotherapy, for example, up to a few hundred Mcps/mm2 in diagnostic CT. Thus, pulse pile-up can severely distort the measurement of counts and energies. Here, we experimentally evaluate the counting and spectral performance of SiPM-based scintillation detectors at 60 keV as a function of ICR/pile-up level. We coupled <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$0.9\times 0.9\times 3.5~{\mathrm { mm}}^{3}$ </tex-math></inline-formula> LYSO:Ce and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$0.9\times 0.9\times 4.5~{\mathrm { mm}}^{3}$ </tex-math></inline-formula> YAP:Ce scintillators to <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$1.0\times 1.0~{\mathrm { mm}}^{2}$ </tex-math></inline-formula> ultrafast SiPMs and exposed these single-pixel detectors to a 10-GBq Am-241 source. We varied ICR from 0 to 5 Mcps/pixel and studied detector performance for paralyzable-like (p-like) and nonparalyzable-like (np-like) counting algorithms, after applying a second-order low-pass filter with cut-off frequencies <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$f_{\mathrm { c}}$ </tex-math></inline-formula> of 5, 10, or 20 MHz to the pulse trains. Counting performance was quantified by the output count rate (OCR) and the count-rate loss factor (CRLF). In addition to the traditional spectral performance measure of the full-width-at-half-maximum (FWHM) energy resolution at low ICR, we propose the spectral degradation factor (SDF) to quantify spectral effects of pile-up at any ICR. Best counting performance is obtained with np-like counting and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$f_{\mathrm { c}}{=}$ </tex-math></inline-formula> 20 MHz, for which the count-rate loss is at most 10% in the investigated range of ICRs, whereas p-like counting yields best spectral performance. Due to less pile-up, the fastest pulses obtained with <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$f_{\mathrm { c}}{=}$ </tex-math></inline-formula> 20 MHz already provide the best SDF values at ICRs of a few Mcps/pixel, despite their worse low-rate energy resolution. Hence, spectral performance under pile-up conditions appears to benefit more from substantially faster pulses than a somewhat better low-rate energy resolution. Moreover, we show that the pulse shape of SiPM-based detectors allows to improve spectral performance under pile-up conditions using dedicated peak detection windows.