Evaluation of a hybrid direct–indirect active matrix flat-panel imager using Monte Carlo simulation

Abstract.

Purpose: Monte Carlo simulations were used to evaluate the imaging properties of a composite direct–indirect active matrix flat-panel imager (AMFPI) with potentially more favorable tradeoffs between x-ray quantum efficiency and spatial resolution than direct or indirect AMFPIs alone. This configuration, referred to as a hybrid AMFPI, comprises a scintillator that is optically coupled to an a-Se direct AMFPI through a transparent electrode and hole blocking layer, such that a-Se acts as both a direct x-ray converter and an optical sensor.

Approach: GEANT4 was used to simulate x-ray energy deposition, optical transport, and charge signal generation processes in various hybrid AMPFI configurations under RQA5 and RQA9 x-ray beam conditions. The Fujita–Lubberts–Swank method was used to quantify the impact of irradiation geometry, x-ray converter thicknesses, conversion gain of each layer, and x-ray cross talk between layers on detective quantum efficiency (DQE).

Results: Each hybrid configuration had a greater DQE than its direct AMFPI layer alone. The DQE improvement was largest at low spatial frequencies in both front- and back-irradiation (BI) geometries due to increased x-ray quantum efficiency provided by the scintillator. DQE improvements persisted at higher frequencies in BI geometry due to preferential x-ray absorption in a-Se. Matching the x-ray-to-charge conversion gains of a hybrid AMFPI’s direct and indirect detection layers affects its Swank factor and, thus, DQE(0). X-ray cross talk has a negligible impact on the $\mathrm{DQE}(f)$ of hybrid AMFPIs with sufficiently high optical quantum efficiency.

Conclusion: An optimized hybrid AMFPI can achieve greater DQE performance than current direct or indirect AMFPIs.

1. Introduction

Active matrix flat-panel imagers (AMFPIs) are currently the dominant technology in digital radiography and fluoroscopy and may be broadly classified as either direct or indirect AMFPIs.¹ Direct AMFPIs [Fig. 1(a)] use a photoconductor to convert absorbed x-ray energy into charge, which may be collected by a 2D array of pixel electrodes and read out using an amorphous silicon (a-Si:H) thin-film transistor (TFT) array. Direct AMFPIs possess a high spatial resolution due to the minimal lateral spread of charge within the photoconductor. However, they suffer from a low x-ray quantum efficiency ($\eta$) due to the low atomic number ($Z=34$) of amorphous selenium (a-Se), which is currently the only photoconductor that is commercially feasible for use in AMFPIs. This fundamentally limits detective quantum efficiency (DQE), a detector performance metric that quantifies the spatial-frequency ($f$) dependent signal-to-noise properties of an x-ray imager. Higher $Z$ alternatives to a-Se, such as ${\mathrm{PbI}}_2$, ${\mathrm{HgI}}_2$, and various perovskites, are currently being researched, but they have not yet been sufficiently developed for practical use.^2–5 Direct AMFPIs using thick a-Se layers (up to $1000\mu \mathrm{m}$)^6–9 have been developed to improve $\eta$; however, further increases in thickness are not possible due to the limited mobility–lifetime product of electrons in a-Se.⁷ As a result, direct AMFPIs are best suited for lower energy (20 to 50 keV) applications, such as mammography and digital breast tomosynthesis.^10,11

Fig. 1.

(a) Schematic of a direct AMFPI. X-rays are converted directly into charge via the photoconductor, which is read out by the TFT array. (b) Schematic of an indirect AMFPI. X-rays are converted into optical photons, which are then converted into charge and read out by a photodiode-TFT array. BI of indirect AMFPIs results in greater light escape efficiency and less blur and image noise than FI.

Indirect AMFPIs [Fig. 1(b)] consist of a scintillator coupled to a pixelated photodiode–TFT array. The scintillator converts the energy of absorbed x-rays into optical photons, which are then converted to charge by the photodiodes. Indirect AMFPIs possess a higher $\eta$ than direct AMFPIs under most imaging conditions due to the high $Z$ of the inorganic scintillators they use, which are typically powder ${\mathrm{Gd}}_2{\mathrm{O}}_2\mathrm{S}:\mathrm{Tb}$ (GOS) or columnar CsI:Tl (herein, referred to as CsI). This advantage is partially offset by the fact that indirect AMFPIs have a lower spatial resolution and more image noise (i.e., Swank noise and the Lubberts effect)^12,13 than direct AMFPIs due to light scattering and absorption within the scintillator. Consequently, indirect AMFPIs are best suited for dose-sensitive applications that require higher x-ray energies (e.g., $>100\mathrm{kVp}$) and heavily filtered beams, such as interventional fluoroscopy and cone-beam CT.

Alternative AMFPI configurations are now being proposed to improve indirect AMFPI performance. For example, Sato et al. proposed applying a back-irradiated (BI) geometry commonly used in screen-film radiography to AMFPIs.¹⁴ In a BI geometry [see Fig. 1(b)], x-rays are first transmitted through the photodiode–TFT array before interacting with the scintillator. This results in preferential x-ray absorption near the optical sensor, which increases light escape efficiency and reduces blur and image noise. Lubinsky et al.¹⁵ and Howansky et al.¹⁶ showed that these advantages increase the DQE of indirect AMFPIs relative to their performance in the usual front-irradiated (FI) geometry, particularly at high $f$. Further improvements in DQE are expected using a properly optimized “sandwich” AMFPI configuration, comprising an FI and BI scintillator coupled to a bidirectional optical sensor.¹⁵ Other investigators have reported the performance of AMFPIs comprising multiple stacks of scintillators and optical sensors. Rottmann et al. investigated an electronic portal imaging device (EPID) comprising four layers, each consisting of a copper build-up layer, a scintillator, and a photodiode–TFT array.¹⁷ A substantially higher DQE was reported compared with an EPID using only one layer. Maurino proposed a similar detector for single-shot dual-energy diagnostic imaging, comprising three layers of stacked indirect AMFPIs using CsI.¹⁸

In this work, we propose a novel indirect–direct detector configuration referred to as a hybrid AMFPI (Fig. 2), which consists of an x-ray scintillator coupled to an a-Se direct detector through thin, optically transparent intervening layers (i.e., bias electrode and blocking layer).¹⁹ In this configuration, both the scintillator and a-Se function as x-ray converters; a-Se generates signal charge via direct interactions, while the scintillator produces optical photons, which are subsequently detected and converted into charge by a-Se. Practical limits on a-Se thickness and, thus, the direct detector’s $\eta$, are overcome using the scintillator to absorb x-rays that would otherwise be undetected to achieve both a high $\eta$ and high spatial resolution. BI geometry is proposed for the hybrid AMFPI both to maximize x-ray attenuation in a-Se, which has more desirable imaging properties than the scintillator, and to maximize the inherent imaging performance of the scintillator (Fig. 2). The purpose of this work is to use Monte Carlo (MC) simulation to evaluate the potential imaging performance of hybrid AMFPIs and the impact of using different design parameters.

Fig. 2.

Schematic of a hybrid AMFPI. Both a-Se and the scintillator function as x-ray converters. X-rays absorbed in a-Se are converted directly into charge. X-rays absorbed in the scintillator are converted into optical photons, which are detected at the scintillator/a-Se interface and converted into charge. Charge is then read out by the TFT array.

2. Methods

MC simulations were used to model the x-ray imaging performance of different hybrid AMFPI configurations (Sec. 2.1) by generating ensembles of the detector’s 2D response to individual normally incident x-rays, i.e., single x-ray responses (Sec. 2.2). Zero-frequency metrics, such as the pulse height spectrum (PHS), Swank factor, and DQE(0), were found by integrating each single x-ray response in 2D to yield the total signal induced by each x-ray (Sec. 2.3). To derive spatial-frequency dependent metrics, such as $\mathrm{MTF}(f)$, $\mathrm{NPS}(f)$, and $\mathrm{DQE}(f)$, the Fujita–Lubberts–Swank (FLS) method was used (Sec. 2.4). The impact of irradiation geometry, x-ray converter thickness, x-ray conversion gain of each layer (Sec. 2.5), and x-ray cross talk between each layer (Sec. 2.6) on these performance metrics was then investigated.

2.1. Overview of the Hybrid AMFPI Simulation

MC simulations were performed using GEANT4, a C++ based MC package designed to simulate particle transport through matter.²⁰ X-ray and electron transport were modeled using the “G4EmPenelopePhysics” GEANT4 physics list. Optical transport was modeled using the “G4OpticalPhysics” list. As shown in Fig. 3(a), the hybrid AMFPI was modeled as a three-layer structure comprising (1) an x-ray scintillator, (2) a photoconductive layer of a-Se, and (3) TFT substrate glass. An overview of simulation parameters is shown in Table 1.

Fig. 3.

(a) Schematic of our hybrid AMFPI model, which consists of GOS, a-Se, and TFT substrate glass. Four types of x-ray interactions are possible. Type 1 interactions only deposit energy in a-Se and only result in a direct signal component. Type 2 interactions only deposit energy in GOS and only result in an indirect signal component. Type 3 and 4 interactions result in x-ray cross talk from a-Se into GOS and GOS into a-Se, respectively, which occur due to either x-ray scatter or fluorescence. (b) Relative intensity plots of RQA5 and RQA9 x-ray spectra, which have average energies of 52 and 76 keV, respectively.

Table 1.

Parameters used to simulate the hybrid AMFPI.

Parameter	Value
GOS thickness ($d_{\mathrm{GOS}}$)	0 to $1000\mu \mathrm{m}$
a-Se thickness ($d_{\mathrm{a}\text{-}\mathrm{Se}}$)	0 to $1000\mu \mathrm{m}$
TFT substrate thickness	$700\mu \mathrm{m}$
GOS density	$4.59\mathrm{g}{\mathrm{cm}}^{-3}$
a-Se density	$4.29\mathrm{g}{\mathrm{cm}}^{-3}$
GOS absorption length	40 mm
GOS scattering length	$3.67\mu \mathrm{m}$
Reflectivity of GOS backing	0.88
Optical quantum efficiency ($\beta$)	0 to 1
GOS scintillation yield ($Y_{\mathrm{GOS}}$)	$60\mathrm{ke}{\mathrm{V}}^{-1}$
X-ray energy per detected ehp in a-Se ($W_{\mathrm{Se}}$)	7 eV to $W_{\mathrm{Se}}\to \infty$
Electric field of a-Se ($E_{\mathrm{Se}}$)	0 to $35\mathrm{V}{\mu \mathrm{m}}^{-1}$
Exponent in $W_{\mathrm{Se}}$ versus $E_{\mathrm{Se}}$ power law ($\gamma$)	0.7
Pixel size	$75\mu \mathrm{m}$, $139\mu \mathrm{m}$

The x-ray scintillator material considered in this work was GOS due to the fact that its x-ray imaging properties have been extensively characterized in the literature both experimentally and theoretically. Optical transport in GOS was modeled using the Mie scattering model developed by Liaparinos et al., with a Mie scattering length of $3.67\mu \mathrm{m}$ and an anisotropy factor of 0.82.^21–23 An absorption length of 40 mm was used.²⁴ The density ($4.59\mathrm{g}{\mathrm{cm}}^{-3}$)²² and backing reflectivity (0.88)²⁵ were taken to match those of the Lanex Fast Back GOS screen (Eastman Kodak Company), which has a thickness of $\sim 290\mu \mathrm{m}$. However, the thickness of GOS may easily be increased up to $1000\mu \mathrm{m}$.²⁶ Accordingly, we simulated a range of GOS thicknesses between 0 and $1000\mu \mathrm{m}$. a-Se was modeled as a uniform slab of selenium with a density of $4.29\mathrm{g}{\mathrm{cm}}^{-3}$.²⁷ The thickness of a-Se used in direct AMFPIs is usually $200\mu \mathrm{m}$²⁸ for mammography and up to $1000\mu \mathrm{m}$ for general radiography and fluoroscopy. Therefore, we also simulated a-Se using a range of thicknesses from 0 to $1000\mu \mathrm{m}$. The TFT substrate was modeled as borosilicate glass²⁹ with a density of $2.23\mathrm{g}{\mathrm{cm}}^{-3}$ and a thickness of $700\mu \mathrm{m}$. Each hybrid AMFPI configuration was simulated using both an FI and a BI geometry.

Each simulation comprised a pencil beam of 20,000 input x-rays using either RQA5 or RQA9 x-ray beam qualities [Fig. 3(b)]. These x-ray energy spectra were computed using the TASMIP spectral model³⁰; the energy of each input x-ray in the simulations was determined by randomly drawing from these distributions. X-rays were sent with normal incidence along the $z$ axis toward the detector at the $x-y$ plane’s origin, and the 2D spatial distribution of signal induced by each interacting x-ray was recorded. The results were then spatially binned in 2D to represent signal integration on a pixel matrix. A default pixel size of $75\mu \mathrm{m}$ (100% fill factor) was considered in this work, although this parameter is readily adjusted during postprocessing.

2.2. Simulation of the Hybrid AMFPI’s Single X-Ray Response

Signal charge is induced directly in a hybrid AMFPI by x-ray energy that is directly converted to charge in the a-Se photoconductor and indirectly by optical photons produced by x-ray interactions in the scintillator that are converted to charge upon reaching the scintillator–selenium interface. The former signal contribution is hereafter referred to as a “direct signal component” and the latter as an “indirect component.” In a hybrid structure, each interacting x-ray may either deposit its entire energy in the a-Se or scintillator layer or deposit a fraction of its energy in both layers through processes such as remote K-fluorescence reabsorption or x-ray scattering [Fig. 3(a)]. Consequently, the 2D spatial distribution of the signal induced by a single interacting x-ray may also be described in terms of its direct and indirect signal components, denoted as $S_{\text{direct}}(\overrightarrow{r})$ and $S_{\text{indirect}}(\overrightarrow{r})$, respectively. The total spatial distribution of signal induced by each detected x-ray is

$$S(\overrightarrow{r})=S_{\text{direct}}(\overrightarrow{r})+S_{\text{indirect}}(\overrightarrow{r}),$$ (1)

where $\overrightarrow{r}=(x,y)$.

For a given x-ray, $S_{\text{direct}}(\overrightarrow{r})$ was derived by collecting raw data sets consisting of the magnitude and location of the energy deposited by primary electrons produced by photoelectric or Compton interactions in a-Se. Charge generation was assumed to be local to each site of energy deposition, and lateral spread of charge during transit to the pixel electrode was assumed to be negligible.³¹ Due to x-ray and electron scattering, each incoming x-ray results in multiple sites of energy deposition. The number of electron–hole pairs (ehps) generated at the $i$’th site of energy deposition in a-Se ($g_{\mathrm{Se},i}$) was modeled as a Poisson-distributed random variable with an average of

$$⟨g_{\mathrm{Se},i}⟩=1000·\frac{E_i}{W_{\mathrm{Se}}},$$ (2)

where $E_i$ denotes the amount of energy in keV deposited at $i$’th deposition site and $W_{\mathrm{Se}}$ denotes the average amount of energy in eV required to generate one detected ehp. A default value of $W_{\mathrm{Se}}=50\mathrm{eV}$ was assumed, corresponding to an electric field of $E_{\mathrm{Se}}=10\mathrm{V}{\mu \mathrm{m}}^{-1}$ within a-Se.³² For an x-ray resulting in $N$ energy depositions in a-Se, $S_{\text{direct}}(\overrightarrow{r})$ is given by

$$S_{\text{direct}}(\overrightarrow{r})=\sum _i^N{g}_{\mathrm{Se},i}\delta (\overrightarrow{r}-{\overrightarrow{r}}_i),$$ (3)

where ${\overrightarrow{r}}_i$ denotes the location of the $i$’th energy deposition. For each energy deposition, $g_{\mathrm{Se},i}$ was computed during postsimulation processing. This approach was used to evaluate the impact of different values of $W_{\mathrm{Se}}$ (e.g., due to differences in applied bias voltage) using otherwise identical data sets.

$S_{\text{indirect}}(\overrightarrow{r})$ was derived by collecting raw data sets consisting of the magnitude and location of energy deposited by primary electrons in the scintillator. The number of optical photons generated at the $i$’th site of energy deposition in GOS was approximated as a Poisson-distributed random variable with an average of

$$⟨g_{\mathrm{GOS},i}⟩=Y_{\mathrm{GOS}}·E_i,$$ (4)

where $Y_{\mathrm{GOS}}$ is the scintillation yield of GOS, which was assumed to be $60{\mathrm{keV}}^{-1}$ based on reports from the literature.^33–36 Optical photons created at the $i$’th energy deposition site in GOS reach the GOS–selenium interface with a depth-dependent probability $\xi (z_i)$, which is determined by the extent of light scattering and absorption within the scintillator. This process may be represented by a Bernoulli distributed random variable ${\xi }_{ij}$, which has a value of 1 if the $j$’th optical photon due to the $i$’th energy deposition reaches a-Se and 0 otherwise. Each optical photon that reaches a-Se will produce an ehp with a probability $\beta$ denoting the optical quantum efficiency of a-Se to GOS luminescence. This process is also represented by a Bernoulli distributed random variable ${\beta }_{ij}$. For a single x-ray interaction resulting in $M$ energy depositions in GOS, $S_{\text{indirect}}(\overrightarrow{r})$ is given by

$$S_{\text{indirect}}(\overrightarrow{r})=\sum _i^M\sum _j^{g_{\mathrm{GOS},i}}{\beta }_{ij}{\xi }_{ij}\delta (\overrightarrow{r}-{\overrightarrow{r}}_{ij}),.$$ (5)

where ${\overrightarrow{r}}_{ij}$ denotes the location of each optical photon when it reaches the GOS–selenium interface. For each optical photon, ${\beta }_{ij}$ was computed during postsimulation processing. This approach was used to evaluate the impact of the optical quantum efficiency between the scintillator and a-Se using otherwise identical data sets.

2.3. Zero-Frequency Properties of the Hybrid AMFPI: PHS, $A_S$, and DQE(0)

Zero-frequency detector performance metrics were determined by first integrating $S(\overrightarrow{r})$ in 2D to find the total signal $S$ induced by each single x-ray. The probability distribution of $S$ is referred to as the PHS and was calculated by creating histograms of all $S$ from a given simulation. The Swank factor, $A_S$, is defined as follows:

$$A_S=\frac{⟨S{⟩}^2}{⟨S^2⟩},$$ (6)

and was calculated directly from this equation.¹² $A_S$ characterizes the relative spread of the PHS; $A_S=1$ corresponds to an infinitely narrow PHS [i.e., $\delta (S)$], while $A_S=0$ corresponds to an infinitely broad PHS. DQE(0) was found according to the following equation:¹²

$$\mathrm{DQE}(0)=\eta A_S.$$ (7)

2.4. Spatial-Frequency Dependent Properties of the Hybrid AMFPI: MTF(f), NPS(f), and DQE(f)

$\mathrm{DQE}(f)$ of each hybrid configuration was calculated using the FLS simulation approach described in detail by Star-Lack et al.²² Briefly, this approach uses the following equation:

$$\mathrm{DQE}(f)=\frac{{\mathrm{MTF}}^2(f)}{q\mathrm{NNPS}(f)},$$ (8)

where $\mathrm{MTF}(f)$ is the detector’s presampling MTF, $q$ is the x-ray input fluence, and $\mathrm{NNPS}(f)$ is the noise power spectrum divided by the mean squared signal. $\mathrm{MTF}(f)$ is determined using the slanted slit method of Fujita et al.³⁷ We used an infinitely thin slit created during postprocessing by uniformly spacing each $S(\overrightarrow{r})$ along a line angled at 3.57 deg with respect to the $x$ axis, corresponding to an oversampling rate of 16.³⁷ Next, the product $q\mathrm{NNPS}(f)$ is determined by first calculating its frequency dependence. This is done by integrating each $S(\overrightarrow{r})$ along the $y$ axis to yield the 1D single x-ray imaging response $S(x)$. The frequency dependence is then given by

$$q\mathrm{NNPS}(f)\propto ⟨{|S(f)|}^2⟩,$$ (9)

where $S(f)$ denotes the discrete Fourier transform of $S(x)$.¹³ This expression is then normalized to the correct zero-frequency value, which is found by combining Eqs. (7) and (8) to yield

$$q\mathrm{NNPS}(0)=\frac{1}{\eta A_S}.$$ (10)

The $q\mathrm{NNPS}(f)$ may then be calculated as follows (Fig. 4):

$$q\mathrm{NNPS}(f)=q\mathrm{NNPS}(0)\frac{⟨{|S(f)|}^2⟩}{⟨{|S(0)|}^2⟩}=\frac{1}{\eta A_S}\frac{⟨{|S(f)|}^2⟩}{⟨{|S(0)|}^2⟩}.$$ (11)

Fig. 4.

A schematic depicting the methods used to calculate various detector performance metrics. Row 1: By integrating each single x-ray response, $S(\overrightarrow{r})$, in 2D, the total signal magnitude due to each x-ray, $S$, is derived. Each measurement of $S$ is then binned in a histogram to determine the PHS, which is used to calculate the Swank factor $A_S$ and DQE(0). Row 2: The $S(\overrightarrow{r})$ ensemble is spaced uniformly along an infinitely thin slanted slit, from which $\mathrm{MTF}(f)$ is computed via Fujita’s method. Row 3: Each $S(\overrightarrow{r})$ is integrated along the $y$-axis to derive $S(x)$. Fourier analysis is performed to compute $q\mathrm{NNPS}(f)$. $\mathrm{DQE}(f)$ is then found from $\mathrm{MTF}(f)$ and $q\mathrm{NNPS}(f)$.

2.5. X-Ray Cross Talk

Due to the multilayer structure of the hybrid AMFPI, processes such as remote K-fluorescence reabsorption and x-ray scatter can result in x-ray “cross talk,” where an x-ray that initially interacts in one x-ray converter also deposits energy in the other converter. As seen in Fig. 3(a), an x-ray’s direct signal component may be generated in two ways through direct interaction with a-Se (types 1 and 3) or through x-ray cross talk from GOS into a-Se (type 4). Similarly, an x-ray’s indirect signal component may be generated in two ways: through x-ray interactions in GOS that result in optical photons that reach the a-Se/GOS interface (types 2 and 4) or through x-ray cross talk from a-Se into GOS (type 3).

The following approach was used to separately evaluate the impact of cross talk (i.e., interactions of type 3 and type 4, respectively) on the performance of a BI hybrid AMFPI comprising $300\mu \mathrm{m}$ of both GOS and a-Se under RQA5 beam conditions. PHS and $\mathrm{DQE}(f)$ were calculated using only the indirect component of each x-ray’s signal and were compared for the cases of either including or excluding all signals due to cross talk. Similarly, PHS and $\mathrm{DQE}(f)$ were calculated using only the direct component of each x-ray’s signal and were compared for the cases of either including or excluding all signals due to cross talk.

2.6. Gain Matching

Each detected x-ray in a hybrid AMFPI will experience a different x-ray-to-charge conversion gain depending on whether it initially interacts with a-Se or GOS. These random gain fluctuations degrade the detector’s $A_S$, thereby reducing its DQE(0) according to Eq. (7). For a given amount of absorbed x-ray energy, the conversion gain of GOS is determined by $Y_{\mathrm{GOS}}$, $\xi (z)$, and $\beta$ [see Eqs. (4) and (5)], while that of a-Se is determined by $W_{\mathrm{Se}}$ [see Eqs. (2) and (3)]. This results in a design consideration referred to as “gain matching” that is unique to multilayer detectors, wherein $A_s$ may be optimized by adjusting $Y_{\mathrm{GOS}}$, $\xi (z)$, $\beta$, and $W_{\mathrm{Se}}$ to minimize variation in conversion gain between layers. We confined our analysis to $\beta$ and $W_{\mathrm{Se}}$ since both can be easily altered by adjusting the internal electric field of a-Se, $E_{\mathrm{Se}}$, while $\beta$ can also be altered by doping a-Se. The following approach was used to evaluate the impact of gain matching on the performance of a BI hybrid AMFPI comprising $300\mu \mathrm{m}$ of both GOS and a-Se under RQA5 beam conditions.

DQE(0) was calculated as a function of both $\beta$ and $W_{\mathrm{Se}}$ using

$$\mathrm{DQE}(0)=\eta \frac{{⟨S(\beta ,W_{\mathrm{Se}})⟩}^2}{⟨S{(\beta ,W_{\mathrm{Se}})}^2⟩},$$ (12)

where $S(\beta ,W_{Se})$ was computed for each x-ray via postprocessing of simulation data. Due to the excessive computation involved in processing the same raw dataset for each pair $(\beta ,W_{Se})$, we first calculated $S_{\text{indirect}}$ and $S_{\text{direct}}$ for $\beta =1$ and $W_{\mathrm{Se}}=50\mathrm{eV}$. We then estimated $S(\beta ,W_{Se})$ as follows:

$$S(\beta ,W_{\mathrm{Se}})=\beta ·S_{\text{indirect}}+(50/W_{\mathrm{Se}})·S_{\text{direct}},$$ (13)

which is valid for large values of $S_{\text{indirect}}$ and $S_{\text{direct}}$.

DQE(0) was also calculated as a function of $E_{\mathrm{Se}}$. Values from $E_{\mathrm{Se}}=0\mathrm{V}{\mu \mathrm{m}}^{-1}$ to $E_{\mathrm{Se}}=35\mathrm{V}{\mu \mathrm{m}}^{-1}$ were considered. $W_{\mathrm{Se}}$ was assumed to follow the empirically derived proportionality:

$$W_{\mathrm{Se}}\propto E_{\mathrm{Se}}^{-\gamma },$$ (14)

where $\gamma$ is a fitting parameter with a value of 0.7.^38,39 $W_{\mathrm{Se}}$ was normalized to a value of 50 eV at $E_{\mathrm{Se}}=10\mathrm{V}{\mu \mathrm{m}}^{-1}$. $\beta$ was assumed to follow the Onsager mechanism, where $\beta (r_0,E_{\mathrm{Se}})$ is a function of both $E_{\mathrm{Se}}$ and an initial separation $r_0$ of ehps, which is both wavelength ($\lambda$) and temperature ($T$) dependent, given by

$$\beta (r_0,E)=\frac{e^{-A}{e}^{-B}}{B}\sum _{m=0}^{\infty }\sum _{n=0}^{\infty }\frac{A^m}{m!}\frac{B^{m+n}}{(m+n)!},$$ (15)

where $A=e^2/4\pi \kappa {\epsilon }_{0}kT{r}_0$ and $B=e{r}_0{E}_{\mathrm{Se}}/kT$.⁴⁰ The values of $\beta$, $W_{\mathrm{Se}}$, and $E_{\mathrm{Se}}$ that resulted in optimal DQE(0) were determined.

Additionally, two PHS were calculated: ${\mathrm{PHS}}_{\mathrm{a}\text{-}\mathrm{Se}}$ which only included x-rays that initially interacted with a-Se (interactions of types 1 and 3), and ${\mathrm{PHS}}_{\mathrm{GOS}}$, which only included those that initially interacted in GOS (interactions of types 2 and 4). The total PHS of the hybrid AMPFI is given by the sum:

$$\mathrm{PHS}={\mathrm{PHS}}_{\mathrm{GOS}}+{\mathrm{PHS}}_{\mathrm{a}\text{-}\mathrm{Se}}.$$ (16)

From Eq. (7), we expected DQE(0) to be highest when values of $\beta$ and $W_{\mathrm{Se}}$ were chosen that resulted in the narrowest combined PHS (i.e., the highest $A_S$).

3. Results

3.1. Effect of Irradiation Geometry on DQE(f)

Figure 5 compares $\mathrm{DQE}(f)$ of a hybrid AMFPI using $300\mu \mathrm{m}$ of both GOS and a-Se in FI versus BI geometry at both (a) RQA5 and (b) RQA9. Optimal gain matching was assumed, using $\beta =0.44$ and $W_{\mathrm{Se}}=50\mathrm{eV}$. For each beam quality and combination of layer thicknesses, BI hybrid AMFPIs had higher $\mathrm{DQE}(f)$ than if they were FI, particularly at higher $f$. We shall, therefore, confine subsequent results to BI hybrid AMFPI configurations.

Fig. 5.

$\mathrm{DQE}(f)$ of a hybrid AMFPI in FI and BI geometries for both (a) RQA5 and (b) RQA9 beam qualities. Both a-Se and GOS thicknesses were $300\mu \mathrm{m}$, and optimal gain matching was assumed using $\beta =0.44$ and $W_{\mathrm{Se}}=50\mathrm{eV}$. BI results in a significant increase in $\mathrm{DQE}(f)$ at higher frequencies compared to FI.

3.2. Effect of Thickness of Both a-Se and the Scintillator on DQE(f)

Figure 6 shows $\mathrm{DQE}(f)$ of hybrid AMFPIs using various a-Se thicknesses and a fixed GOS thickness ($300\mu \mathrm{m}$). Optimal gain matching was assumed, using $\beta =0.44$ and $W_{\mathrm{Se}}=50\mathrm{eV}$. Increasing a-Se thickness increases $\mathrm{DQE}(f)$ at all $f$, with diminishing returns after $\sim 700\mu \mathrm{m}$. Improvement in $\mathrm{DQE}(f)$ is greater when using an RQA5 spectrum compared with RQA9.

Fig. 6.

$\mathrm{DQE}(f)$ of hybrid AMFPIs with a GOS thickness of $300\mu \mathrm{m}$ and various a-Se thicknesses at (a) RQA5 and (b) RQA9. Optimal gain matching was assumed using $\beta =0.44$ and $W_{\mathrm{Se}}=50\mathrm{eV}$. Greater a-Se thickness increases $\mathrm{DQE}(f)$ at all frequencies, particularly when using lower energy x-ray spectra such as RQA5.

Figure 7 shows $\mathrm{DQE}(f)$ of hybrid AMFPIs using various GOS thicknesses and a fixed a-Se thickness ($300\mu \mathrm{m}$). Optimal gain matching was assumed, using $\beta =0.44$ and $W_{\mathrm{Se}}=50\mathrm{eV}$. The addition of GOS to a-Se (solid curves) greatly increases $\mathrm{DQE}(f)$ at lower $f$ with smaller improvements at higher $f$. As $f$ increases, the hybrid AMFPI’s $\mathrm{DQE}(f)$ approaches that of a-Se alone. As GOS thickness increases, $\mathrm{DQE}(f)$ continues to improve, but with diminishing returns; this improvement becomes restricted to increasingly lower $f$. Improvement of $\mathrm{DQE}(f)$ with GOS thickness is more significant when using the RQA9 spectrum as opposed to RQA5. When using an a-Se thickness of $700\mu \mathrm{m}$ and a GOS thickness of $1000\mu \mathrm{m}$ (dotted curves), hybrid AMFPI $\mathrm{DQE}(f)$ outperforms or is comparable to $\mathrm{DQE}(f)$ of a typical CsI indirect AMPFI (dashed curve⁴¹) at all $f$; this is particularly the case for low $f$ when using an RQA9 spectrum, for which an improvement from 47% to 76% in DQE(0) is observed.

Fig. 7.

$\mathrm{DQE}(f)$ of hybrid AMFPIs with an a-Se thickness of $300\mu \mathrm{m}$ and various GOS thicknesses at (a) RQA5 and (b) RQA9. Optimal gain matching was assumed using $\beta =0.44$ and $W_{\mathrm{Se}}=50\mathrm{eV}$. Greater GOS thickness increases $\mathrm{DQE}(f)$ at lower frequencies, particularly when using higher energy x-ray spectra such as RQA9.

3.3. Effect of Gain Matching on DQE(0)

Figure 8 shows ${\mathrm{PHS}}_{\mathrm{GOS}}$ and ${\mathrm{PHS}}_{\mathrm{a}\text{-}\mathrm{Se}}$ of a hybrid AMFPI ($d_{\mathrm{Se}}=300\mu \mathrm{m}$, $d_{\mathrm{GOS}}=300\mu \mathrm{m}$) using different values of $\beta$ and $W_{\mathrm{Se}}$ under RQA5 beam conditions. Average gains are indicated by dashed lines. Reducing $\beta$ from 1 to 0.44 [Figs. 8(a) and 8(b)] brings both averages closer together and improves DQE(0) from 0.589 to 0.678, a $\sim 9\%$ increase on the absolute scale. An identical improvement in DQE(0) is obtained by reducing $W_{\mathrm{Se}}$ by the same factor of 0.44 [Fig. 8(c)].

Fig. 8.

${\mathrm{PHS}}_{\mathrm{GOS}}$ (blue) and ${\mathrm{PHS}}_{\mathrm{a}\text{-}\mathrm{Se}}$ (magenta) using different values of $\beta$ and $W_{\mathrm{Se}}$. Greater overlap increases DQE(0). Dashed lines indicate averages. (a) $\beta =1$, $W_{\mathrm{Se}}=50\mathrm{eV}$; (b) $\beta =0.44$, $W_{\mathrm{Se}}=50\mathrm{eV}$; and (c) $\beta =1$, $W_{\mathrm{Se}}=22\mathrm{eV}$.

Figure 9(a) shows DQE(0) as a function of $\beta$, with $W_{\mathrm{Se}}=50\mathrm{eV}$, compared with DQE(0) of a-Se alone with $W_{\mathrm{Se}}=50\mathrm{eV}$. When $\beta =0$, gain matching is poorest, resulting in a DQE(0) of 0.410. DQE(0) then increases with $\beta$ until it reaches a maximum at $\beta =0.44$, after which DQE(0) decreases. Note that, regardless of the value of $\beta$, DQE(0) of the hybrid AMFPI is still greater than that of a-Se alone.

Fig. 9.

(a) Hybrid AMFPI DQE(0) (dark blue) as a function of $\beta$, with $W_{\mathrm{Se}}=50\mathrm{eV}$. (b) Hybrid AMFPI DQE(0) (magenta) as a function of $W_{\mathrm{Se}}$ with $\beta =1$. DQE(0) of a-Se alone with $W_{\mathrm{Se}}=50\mathrm{eV}$ is shown in teal in both plots.

Figure 9(b) shows DQE(0) as a function of $W_{\mathrm{Se}}$, with $\beta =1$, compared with DQE(0) of a-Se alone with $W_{\mathrm{Se}}=50\mathrm{eV}$. At $1/W_{\mathrm{Se}}=0$ (i.e., $W_{\mathrm{Se}}\to \infty$), DQE(0) of the hybrid AMFPI approaches that of GOS alone (with a reduction due to x-ray attenuation by a-Se), which is slightly less than that of a-Se alone. As $1/W_{\mathrm{Se}}$ increases, DQE(0) increases until a maximum is attained when $W_{\mathrm{Se}}=22\mathrm{eV}$, after which DQE(0) decreases.

Figure 10(a) shows a contour plot of DQE(0) as a function of both $\beta$ and $W_{\mathrm{Se}}$. DQE(0) of a-Se alone (0.410) is attained at $\beta =0$, whereas a DQE(0) of GOS alone (0.301) is attained at $1/W_{\mathrm{Se}}=0$ (with a reduction due to x-ray attenuation by a-Se). A maximum DQE(0) of 0.678 is achieved for all $\beta$ and $W_{\mathrm{Se}}$ combinations lying on the dark red band, corresponding to optimal gain matching. The curves shown in black show $\beta$ and $W_{\mathrm{Se}}$ as functions of $E_{\mathrm{Se}}$ at different wavelengths of optical light, $\lambda$. $E_{\mathrm{Se}}$ intervals of $5\mathrm{V}{\mu \mathrm{m}}^{-1}$ are marked with asterisks on each curve from 0 to $35\mathrm{V}{\mu \mathrm{m}}^{-1}$ (left to right).

Fig. 10.

(a) A contour plot of DQE(0) as a function of $\beta$ and $W_{\mathrm{Se}}$. Warmer colors indicate higher values. DQE(0) reaches a maximum of 0.678 (dark red) whenever optimal gain matching is achieved. Black curves trace paths through $\beta$ and $W_{\mathrm{Se}}$ as a function of $E_{\mathrm{Se}}$ for various values of $\lambda$. $5\mathrm{V}{\mu \mathrm{m}}^{-1}$ intervals are marked by asterisks. (b) DQE(0) versus $E_{\mathrm{Se}}$ for various values of $\lambda$. When $E_{\mathrm{Se}}$ is larger than $\sim 5\mathrm{V}{\mu \mathrm{m}}^{-1}$, DQE(0) is primarily a function of $\lambda$.

Figure 10(b) shows curves of DQE(0) versus $E_{\mathrm{Se}}$, where the optical emissions of GOS have been artificially set to various values of $\lambda$. This is equivalent to taking the value of DQE(0) at each point along the black curve corresponding to $\lambda$ in Fig. 10(a). When $E_{\mathrm{Se}}$ is greater than about $5\mathrm{V}{\mu \mathrm{m}}^{-1}$, DQE(0) primarily depends on $\lambda$ and remains relatively constant with $E_{\mathrm{Se}}$. At a given $E_{\mathrm{Se}}$, DQE(0) is highest between 400 and 450 nm and decreases with $\lambda$.

Figure 11(a) shows the x-ray induced luminescence spectrum of GOS plotted against the spectral sensitivity curves of a-Se both with^42,43 and without⁴⁰ tellurium (Te) doping. While the main peak of GOS is at 550 nm, 38.6% of its emissions are below 500 nm. This results in an effective optical quantum efficiency of $\beta =46.7\%$ and $\beta =13.5\%$ with and without Te doping, respectively. Figure 11(b) shows $\mathrm{DQE}(f)$ of $300\mu \mathrm{m}$ of a-Se alone (dark blue) and a hybrid AMFPI (teal), where $d_{\mathrm{Se}}=300\mu \mathrm{m}$, $d_{\mathrm{GOS}}=300\mu \mathrm{m}$, $W_{\mathrm{Se}}=50\mathrm{eV}$, and $\beta =13.5\%$. $\mathrm{DQE}(f)$ of the same hybrid AMFPI, but with $\beta =46.7\%$, is shown in red. The addition of GOS results in a large increase in $\mathrm{DQE}(f)$ at lower frequencies, particularly for Te-doped a-Se.

Fig. 11.

(a) The x-ray induced luminescence spectrum of GOS and the spectral sensitivity of a-Se both with and without Te doping. (b) $\mathrm{DQE}(f)$ of a-Se alone (blue) and the hybrid AMFPI using GOS both with (red) and without (teal) Te doped a-Se.

3.4. Effect of Cross Talk between a-Se and the Scintillator on DQE(f)

Figure 12(a) shows the PHS of the indirect signal component of a hybrid AMFPI ($d_{\mathrm{Se}}=300\mu \mathrm{m}$, $d_{\mathrm{GOS}}=300\mu \mathrm{m}$) using optimal gain matching ($\beta =0.44$ and $W_{\mathrm{Se}}=50\mathrm{eV}$) under RQA5 beam conditions. Counts due to x-rays initially interacting in GOS are shown in blue, while the additional counts due to cross talk from a-Se are shown in green. The main peak of GOS is centered at $\sim 900\mathrm{ehp}$, while the K-fluorescence escape peak is centered at $\sim 200\mathrm{ehp}$. Cross talk results in an additional peak (shown in green) at $\sim 280\mathrm{ehp}$, which corresponds to the absorption of Se K-fluorescence in GOS. The impact of x-rays scattering from a-Se into GOS is negligible. When $\beta =1$ and $W_{\mathrm{Se}}=50\mathrm{eV}$, cross talk from a-Se into GOS constitutes 0.88% of the total hybrid AMFPI signal (see Table 2). For an optimally gain matched detector, this decreases to 0.59%.

Fig. 12.

(a) PHS consisting of only the indirect component of signal. Counts due to GOS x-ray interactions are shown in blue, while those due to a-Se cross talk are shown in green. Cross talk results in an additional peak at $\sim 280\mathrm{ehp}$ due to a-Se K-fluorescence, which is negligible in magnitude. (b) PHS consisting of only the direct component of signal. Cross talk results in two additional peaks contained within that of a-Se due to gadolinium K-fluorescence. Counts due to a-Se x-ray interactions are shown in blue, while those due to GOS cross talk are shown in green. (c) Hybrid AMFPI $\mathrm{DQE}(f)$ of both direct and indirect signal components is shown with and without cross talk. Only cross talk from GOS into Se was significant, resulting in an increase at low $f$ with a degradation at all other $f$.

Table 2.

A summary of the percent contribution of various types of energy deposition to total signal magnitude using different gain matching parameters. Energy deposited in layer B due to an x-ray initially interacting with layer A is labeled $A\to B$. The contribution of $\mathrm{a}\text{-}\mathrm{Se}\to \mathrm{GOS}$ was negligible. The contribution of $\mathrm{GOS}\to \mathrm{a}\text{-}\mathrm{Se}$ was inversely related to $\beta$ and reaches a maximum of 11.10% at $\beta =0$. The values $\beta =0.44$ and $W_{\mathrm{Se}}=50\mathrm{eV}$ are an example of parameters that lead to optimal gain matching.

	$\beta =1$, $W_{\mathrm{Se}}=50\mathrm{eV}$	$\beta =0.44$, $W_{\mathrm{Se}}=50\mathrm{eV}$	$\beta =0$
% $\mathrm{GOS}\to \mathrm{GOS}$	61.19	41.29	0
% $\mathrm{a}\text{-}\mathrm{Se}\to \mathrm{GOS}$	0.88	0.59	0
% $\mathrm{a}\text{-}\mathrm{Se}\to \mathrm{a}\text{-}\mathrm{Se}$	33.72	51.67	88.90
% $\mathrm{GOS}\to \mathrm{a}\text{-}\mathrm{Se}$	4.22	6.45	11.10

Figure 12(b) shows the PHS of the direct signal component of the same hybrid AMFPI under the same beam conditions. Counts due to x-rays initially interacting in a-Se are shown in blue, while the additional counts due to cross talk from GOS are shown in green. a-Se interactions result in a single peak centered at 925 ehp. K-fluorescence from gadolinium ($K_{\alpha }\approx 43\mathrm{keV}$ and $K_{\beta }\approx 49\mathrm{keV}$) results in two additional peaks, which are contained within the peak of a-Se. Scatter from GOS into a-Se is negligible. When $\beta =1$ and $W_{\mathrm{Se}}=50\mathrm{eV}$, cross talk from GOS into a-Se constitutes 4.22% of the total hybrid AMFPI signal (see Table 2). This grows to 6.45% in an optimally gain matched detector and 11.10% when $\beta =0$.

Figure 12(c) shows the $\mathrm{DQE}(f)$ of direct and indirect signal components both with (solid lines) and without cross talk (dashed lines). The addition of cross talk increases ${\mathrm{DQE}}_{\text{direct}}(f)$ at very low frequencies but slightly degrades ${\mathrm{DQE}}_{\text{direct}}(f)$ for all other frequencies. The impact of cross talk on ${\mathrm{DQE}}_{\text{indirect}}(f)$ is negligible at all frequencies.

4. Discussion

4.1. Effect of Irradiation Geometry on DQE(f)

Superior $\mathrm{DQE}(f)$ was observed in a BI hybrid AMFPI geometry for two reasons: First, BI increases the fraction of x-rays that interact with a-Se, which has a superior spatial resolution compared with GOS. Second, BI increases the imaging performance of GOS by ensuring that more x-rays are absorbed closer to the optical sensor, i.e., a-Se. This results in a greater light-escape efficiency, less optical blur, and a reduction in Lubberts effect.¹⁶

4.2. Effect of Thickness of Both a-Se and Scintillator on DQE(f)

Our results demonstrate that $\mathrm{DQE}(f)$ of a-Se alone provides a lower bound for $\mathrm{DQE}(f)$ of the hybrid AMFPI, while GOS provides additional low frequency $\mathrm{DQE}(f)$. The addition of indirect signal increases the width of a direct AMFPI’s PHS and, thus, decreases its $A_S$ (see Fig. 8). However, the improvement in $\eta$ made through increasing $d_{\mathrm{GOS}}$ apparently outweighs this penalty to improve $\mathrm{DQE}(f)$ overall. Improvement in $\mathrm{DQE}(f)$ with $d_{\mathrm{GOS}}$ is greater for RQA9 due to the fact that GOS is more efficient at absorbing the higher energy x-rays. Conversely, improvement in $\mathrm{DQE}(f)$ with $d_{\mathrm{a}\text{-}\mathrm{Se}}$ is greater for RQA5 due to the fact that x-rays are first incident on a-Se, which is able to absorb the lower energy x-rays in RQA5. Figure 6 suggests that there is little advantage to fabricating hybrid AMFPIs with $d_{\mathrm{a}\text{-}\mathrm{Se}}>700\mu \mathrm{m}$ as $\mathrm{DQE}(f)$ does not significantly increase with $d_{\mathrm{a}\text{-}\mathrm{Se}}$ beyond this point for either RQA5 or RQA9. Figure 7 suggests that a large GOS thickness of $\sim 1000\mu \mathrm{m}$ would be optimal for a hybrid AMFPI utilizing GOS.

4.3. Effect of Gain Matching on DQE(0)

Our results showed that the shapes of ${\mathrm{PHS}}_{\mathrm{GOS}}$ and ${\mathrm{PHS}}_{\mathrm{a}\text{-}\mathrm{Se}}$ may be stretched along the $x$ axis by $\beta$ and $1/W_{\mathrm{Se}}$, respectively. Values of $\beta$ and $W_{\mathrm{Se}}$ that produce the narrowest combined PHS (relative to its average) maximize $A_S$ and, thus, DQE(0). While many pairs of $\beta$ and $W_{\mathrm{Se}}$ produce the same DQE(0), the optimal choice for a practical detector at low dose is that which provides the highest average gain to overcome electronic noise. Though this work used an RQA5 spectrum to demonstrate gain matching, optimal values of $\beta$ and $W_{\mathrm{Se}}$ did not significantly vary between RQA5 and RQA9.

The strong $\lambda$ dependence of $\beta$ [and thus DQE(0)] highlights the importance of scintillator choice in optimizing hybrid AMFPI performance. As seen in Fig. 10(a), a-Se is predominantly sensitive to blue light (below 500 nm) and has limited sensitivity to green light (500 to 565 nm). An ideal scintillator for the hybrid AMFPI would therefore be a blue-emitting scintillator with a high light yield. Because 38.6 % of GOS’s optical emissions lie below 500 nm, $\beta =13.5\%$ and a large increase in $\mathrm{DQE}(f)$ is realized at low $f$. However, if a scintillator’s emissions have a sufficiently large $\lambda$ that $\beta$ approaches 0, merely altering $E_{\mathrm{Se}}$ may not yield optimal gain matching [e.g., $\lambda =550\mathrm{nm}$, Fig. 10(b)]. In such cases, it is possible to augment $\beta$ by doping a-Se. For example, the SATICON camera uses Te doping of a-Se to increase $\beta$ for green and red light, resulting in $\beta =46.7\%$.^42,43 Our model suggests that a hybrid AMFPI using GOS paired with Te doped a-Se achieves near-ideal gain matching.

4.4. Effect of Cross Talk between a-Se and the Scintillator on DQE(f)

Cross talk from a-Se into GOS was not found to be a significant effect for hybrid AMFPI performance. On the other hand, cross talk from GOS into a-Se could potentially impact performance, particularly for low values of $\beta$. This is due to the high energy of GOS K-fluorescence ($K_{\alpha }\approx 43\mathrm{keV}$ and $K_{\beta }\approx 49\mathrm{keV}$), resulting in a higher signal magnitude for cross talk. The slight improvement of ${\mathrm{DQE}}_{\text{direct}}(0)$ with cross talk in Fig. 12(c) is due to the additional $\eta$ provided by the detection of GOS K-fluorescence and scatter. This also explains the fact that DQE(0) improved even when $\beta =0$ in Fig. 9(a). While the cross talk analysis presented in this work considered RQA5 beam conditions, the impact of cross talk was found to be similar under RQA9 conditions as well.

4.5. Limitations and Future Work

Our model used GOS as an example of a scintillator that may be used in the hybrid AMFPI. In the future, we will develop and implement MC simulations of other types of scintillators used in AMFPI, e.g., CsI. Our model did not include electronic noise; thus, our results represent a high dose limit of hybrid AMFPI performance. Since the hybrid AMFPI benefits from signal from both direct and indirect detection layers, we expect the degradation effect of electronic noise to be less than that in either the direct or indirect AMFPI alone. Compared to indirect AMFPI, the effect of electronic noise at high spatial frequencies is mitigated by the direct detection events.⁴⁴ Compared to direct AMFPI, the low dose performance benefits from the boost of x-ray quantum noise at low spatial frequencies, which is afforded by the x-ray absorption in the scintillator.⁴⁴ Effects such as improved low-frequency $\mathrm{DQE}(f)$ via coupling of scintillators to a-Se have been validated in preliminary experimental work using the first ever prototype hybrid AMFPI.⁴⁵ The impact of factors, such as scintillator choice, x-ray converter thickness, gain matching, x-ray cross talk, and electronic noise, on hybrid AMFPI performance will be the subject of future experimental investigations with prototype detectors.

5. Conclusions

In this work, we used MC simulation to model and evaluate the potential imaging performance of the hybrid AMFPI, a novel hybrid direct–indirect AMFPI comprising a scintillator coupled to an a-Se direct detector. In a BI geometry, the addition of the GOS to a-Se substantially improved $\mathrm{DQE}(f)$ at lower $f$ by increasing x-ray quantum efficiency. Larger a-Se thicknesses resulted in better performance, with diminishing returns after $700\mu \mathrm{m}$. Further improvement in $\mathrm{DQE}(f)$ was achieved by increasing GOS from 300 to $1000\mu \mathrm{m}$, which yielded superior or comparable performance to that of a typical CsI indirect AMFPI at all $f$. Gain matching of direct/indirect signal components was shown to play a crucial role in hybrid AMFPI optimization and is affected by choice of scintillator, $E_{\mathrm{Se}}$, and doping of a-Se. When paired with Te doping, GOS achieved near-ideal gain matching with a-Se.

Biographies

Scott Dow has been a graduate student in DRIL Laboratory at Stony Brook University since 2016. His research interests are in medical imaging physics.

Adrian Howansky is a medical physics resident at Stony Brook University Hospital, Department of Radiology. He received his PhD in biomedical engineering from Stony Brook. His research interests are in medical imaging physics.

Anthony R. Lubinsky is a principal investigator at the State University of New York at Stony Brook. He received his PhD in physics from Northwestern University. His research interests are in physics, imaging technology, and materials science for medical imaging systems.

Wei Zhao is a professor of radiology at the State University of New York at Stony Brook. She received her PhD in medical biophysics from the University of Toronto. Her research interests are in medical image sensors and digital breast tomosynthesis.

Disclosures

The authors have no financial interests to disclose.