Two-Step Image Registration for Dual-Layer Flat-Panel Detectors Authors: Dong Sik Kim, Dayeon Lee, Georg Böning, Johannes Kahn Published: 2024 ABSTRACT AbstractBackground: For a single exposure in radiography, a dual-layer flat-panel detector (DFD) can provide spectral images and efficiently utilize the transmitted X-ray photons to improve the detective quantum efficiency (DQE) performance. In this paper, to acquire high DQE performance, we present a registration method for X-ray images acquired from a DFD, considering only spatial translations and scale factors. The conventional registration methods have inconsistent estimate accuracies depending on the captured object scene, even when using entire pixels, and have deteriorated frequency performance because of the interpolation method employed. Methods: The proposed method consists of two steps; the first step is conducting a spatial translation according to the Fourier shift theorem with a subpixel registration, and the second step is conducting a scale transformation using cubic interpolation to process the X-ray projections. To estimate the subpixel spatial translation, a maximum-amplitude method using a small portion of the slant-edge phantom is used. Results: The performance of the proposed two-step method is first theoretically analyzed and then observed by conducting extensive experiments and measuring the noise power spectrum and DQE. An example for registering chest images is also shown. For a DFD, the proposed method shows a better registration result than the conventional one-step registration. The DQE improvement was more than 56% under RQA 9 compared to the single flat-panel detector case. Conclusions: The proposed two-step registration method can efficiently provide aligned image pairs from the DFD to improve the DQE performance at low doses and, thus, increase the accuracy of clinical diagnosis.Keywords: convex combination image, detective quantum efficiency, dual-layer flat-panel detector (DFD), image registration 1. Introduction Dual-layer flat-panel detectors (DFDs) enable single-exposure spectral imaging based on energy-selective imaging from copper filtering or beam hardening [1,2,3]. These detectors have a variety of applications, such as in bone and tissue separations [1], material decomposition [4], and bone mineral density estimation [5]. They can also improve the contrast-to-noise ratio and detective quantum efficiency (DQE) [6,7,8]. Engel et al. [7] showed an increased DQE from DFDs compared to a single-layer detector case having an added thickness of the scintillator layers of DFD. Kim [8] showed that convex combination of upper and lower images acquired from DFD produces higher DQE and reduces noise power spectrum (NPS) [9]. Su et al. [10] obtained super-resolution images based on a DFD. Dual-energy cone-beam computed tomography is a suitable field for the application of DFD [4,5,6]. As shown in Figure 1, the structure of a DFD consists of two layers, where the upper and lower detector layers are attached as close as possible [3,11]. The lower layer usually absorbs relatively high-energy X-ray photons because of the beam hardening compared to the upper layer case. The intermediate layer between the upper and lower layers can prevent the mutual transmission of light photons and can contain a spectral filter for X-rays [1,8,12]. The DFDs introduced in the literature are summarized in Table 1, where indirect conversion flat-panel detectors with CsI(Tl)-scintillator layers and pixel pitches of 0.140 mm and 0.145 mm are used [3,5,6,8,10,11]. Structure of the dual-layer flat-panel detector (DFD) and acquired images (DRTECH Co., Ltd., Seongnam, Republic of Korea, www.drtech.com). (a) Two flat-panel detector layers are attached as close as possible. Conventionally, the upper and lower layers can have the same direction as the incident X-ray. In this example, the lower layer is inverted. The upper layer can also be inverted [8]. dTT is the distance between TFT layers, and dU and dL are the thicknesses of the upper and lower scintillator layers, respectively. (b) An example of chest images acquired from a DFD. The pixel pitch is 0.14 mm, and the image size is 3072×3072 pixels, where dTT=1.1 mm and dU=dL=0.5 mm. Table 1Dual-layer flat-panel detectors (DFDs). DistanceCsI(Tl) ScintillatorIntermediateTFT Detector dTT dU dL FilterPhotodiodeStackingCompanyLu et al. [6], 20192.50.20.551 Cua-SiNormalVarex Imaging, Salt Lake, UT, USAShi et al. [3], 20202.50.20.551 Cua-SiNormalVarex Imaging, Salt Lake, UT, USAKim [8], 20231.3–2.20.50.5No filter, 0.5 Cua-Si/IGZONormal, inverted upper/lowerDRTECH, Seongnam, Republic of KoreaWang et al. [5], 2023-0.2–0.550.55No filter, 1 Cua-SiNormalVarex Imaging, Salt Lake, UT, USASu et al. [10], 20246.60.260.551 Cua-SiNormalCareRay, Suzhou, ChinaLee & Kim [11], 20241.10.35–0.50.5No filtera-IGZOInverted lowerDRTECH, Seongnam, Republic of KoreaThe unit of thickness is millimeter (mm). Dual-layer flat-panel detectors (DFDs). The unit of thickness is millimeter (mm). When stacking the upper and lower detector layers to construct a DFD, the images acquired from the layers should be registered because misalignments occur between the layers. By using a physical positioning device, the rotational deviation between the layers can be controlled to be as small as possible while stacking the layers. However, physically aligning the detector-element positions of the lower thin-film transistor (TFT) layers with respect to those of the upper layer is not easy [8,11]. Hence, a geometric translocation of the lower layer exists in both horizontal and vertical directions with respect to the upper layer. To conduct an image registration, geometric transformations are first estimated between the upper and lower images acquired with an object and then are employed to register the images based on interpolation schemes [13]. A variety of image registration methods have been developed for applications such as medical imaging, general imaging, remote sensing, and measurement [14,15,16,17,18,19]. Recent studies on the registration of detector layers for DFDs are summarized as follows. Shi et al. [3] registered the lower image using an affine transform accounting for translation, rotation, and scale based on interpolation with the IsoCal phantom [20]. Kim [8] and Lee and Kim [11] estimated the translation parameters based on a slant-edge phantom, which was used for measuring the modulation transfer function (MTF), and conducted transforms with those translation parameters. Wang et al. [5] also aligned the lower image based on an interpolation. However, the employed registrations and transforms were usually based on interpolation schemes and, thus, caused registration errors because of the amplitude and phase distortions of interpolation. Note that the traditional intensity-based and feature-based registration approaches [14,21], which use natural scenes, produce large registration errors and, thus, are not appropriate for the registration purpose of DFDs. When registering the upper and lower images acquired from a DFD, the following two items should be considered: Spatial translation while stacking the layers; Spatial translation while stacking the layers; Scale due to the X-ray projection. Scale due to the X-ray projection. The first item is concerned with the registration of the spatial horizontal and vertical translations, as shown in Figure 2a. In a DFD, image registration is a simple process of finding the misaligned spatial translations and using these to transform the lower image. Here, a subpixel registration method [16,22,23,24,25,26,27,28,29,30] is required to accurately find the spatial translation. Subpixel registration is not only important for obtaining aligned image pairs but also for checking the uniform pixel alignment of the stacked detector layers [10]. If the translation parameter of a DFD is accurately estimated, then, using the parameter, the translation of the images acquired from the same DFD can be performed every time to acquire registered image pairs. Hence, there is no need to perform separate image registration in the application of an image processing method. The second item is concerned with a transform of the image scale change due to the X-ray projection, as shown in Figure 2b. Because X-rays generated in the form of a point source in an X-ray tube are projected onto the DFD, the X-ray image obtained from the lower layer is more magnified than the image obtained from the upper layer. Misalignment example for translation and scale parameters in DFD. (a) Translation misalignment while stacking. The lower layer is translated with respect to the upper layer by (2.6, −1.8) pixels to be registered. We assume that the stacked angle between the layers is approximately zero. (b) Scale misalignment due to X-ray projection. For dSID=1600 mm and dTT=1.1 mm, the scale factor is 1.0006875, and the outermost pixel, which has its start location at u=1535 pixels from the center, consequently has an image deviation of 1.055 pixel. In this paper, a two-step image registration method for a DFD is introduced. Conventional one-step registration methods have errors depending on objects in the X-ray images acquired from the upper and lower detector layers [14]. For spectral purposes, these errors can be ignored. However, the purpose of the considered DFD is to improve DQE and obtain high-quality images even at low doses. Therefore, for this purpose, very precise alignment is required to obtain a performance that is close to a theoretically achievable performance. The proposed method for high-precision image registration for DFDs consists of two steps; the first step is conducting a spatial translation according to the Fourier shift theorem [31] with a subpixel registration, and the second step is conducting a scale transformation using cubic interpolation to process the X-ray projections. To conduct an accurate subpixel registration, we employed a method based on the notion of maximum amplitude, where a conventional slant-edge phantom is used as a fiducial mark [11]. This maximum-amplitude method can provide high-precision spatial translation compared to the methods that use the entire pixels of natural scenes [16,22,23,25,26,27,28,30]. The proposed two-step method can achieve more accurate image registration than the conventional one-step case, especially for DFD applications. This paper is organized in the following way. In Section 2, we first describe the registration of images acquired from a DFD. We then propose a two-step registration method for a DFD. Theoretical analysis of the proposed method is also conducted to observe the registration performance. To evaluate the registration accuracies experimentally, extensive experiments using X-ray images acquired from the DFD are presented in Section 3 with discussions. The conclusion is then stated in the last section. 2. Methods In this section, we first comparatively discuss subpixel registration methods for the images acquired from a DFD. We then introduce a two-step registration method that takes into account the magnified image due to the X-ray projection. 2.1. Two-Step Registration for the Dual-Layer Flat-Panel Detector In order to obtain an aligned image pair from the DFD, the lower image acquired from the lower detector layer should be accurately aligned with respect to the upper layer. We first consider the translation of horizontal and vertical directions, as shown in Figure 2a, where the lower detector is translated by (−2.6, 1.8) pixels as an example. We can transform the lower image by using the translation parameter based on the Fourier shift theorem or interpolation schemes, such as linear or cubic interpolations. Here, the translation parameter can be estimated from a subpixel registration method [22]. Conventional intensity-based subpixel registration algorithms generally use the entirety of pixels of natural scenes and, thus, are not suitable for finding such a fine translation of subpixel resolutions for a DFD. Note that if the entire pixels of a typical scene are used for image registration, pixels that are not suitable for alignment are also emphasized and may interfere with precise alignment. To increase the accuracy of subpixel registration for a given small area, we should design a complicated phantom with special patterns as a fiducial mark. However, designing a phantom with such a special pattern is difficult for radiography detector applications. Instead of using such a special phantom, Kim [8] used the conventional slant-edge phantom and maximized the DQE value to obtain a registered convex combination image. Lee and Kim [11] recently conducted a subpixel registration based on a necessary condition on maximizing an amplitude response and extended this notion to developing a high-precision measurement algorithm, where a cyclic-coordinate optimization based on the maximum amplitude is conducted. Note that this method can find local subpixel translations using a small portion of the slant edge as Positions 1, 2, and 3 in Figure 3. Example images of three slant-edge phantoms for measuring horizontal MTF curves at three different locations. Because the X-ray generated in the form of a point source from the X-ray tube is projected onto the DFD, the X-ray image acquired from the lower layer is more enlarged than the image obtained from the upper layer. As shown in Figure 2b, letting dSID denote the source-to-image distance (SID), a horizontal pixel location of u from the aligned center pixel produces the deviation Δu:=udTT/dSID pixels in the lower layer with respect to the upper layer. In other words, compared to the upper image, the lower image is enlarged by a scale factor of 1+dTT/dSID, which is derived from the distance dTT, as shown in Figure 2b. To minimize the deviation Δu due to the projection, we need to minimize the distance dTT between the TFT layers of the upper and lower detector layers. Several examples of the distance are shown in Table 1. We can compensate for the enlarged lower image using the scale factor and an interpolation scheme to conduct a high-precision image registration considering the projection. We can also obtain the scale factor experimentally by measuring pixel deviations from the three slant-edge phantoms of Figure 3. By using the estimated translation and scale parameters, we can conduct an image transformation based on an interpolation [13]. However, the employed interpolation usually deteriorates the frequency responses of the transformed image. To alleviate the deterioration, we propose a two-step method based on the Fourier shift and scaling with an interpolation. The proposed method for a DFD is summarized as follows. Two-Step Registration for the Dual-Layer Flat-Panel Detector: (0)Find the translation of the lower image based on a subpixel registration; calculate the scale factor for a given SID. Find the translation of the lower image based on a subpixel registration; calculate the scale factor for a given SID. (1)Translate the lower image using the translation estimate based on the Fourier shift theorem. Translate the lower image using the translation estimate based on the Fourier shift theorem. (2)Transform the lower image using the scale factor based on a cubic interpolation. Transform the lower image using the scale factor based on a cubic interpolation. Note that the proposed method is composed of two steps: translation based on the Fourier shift theorem and scaling based on a cubic interpolation. Instead of these two steps, we can consider a one-step translation with the translation and scale parameters based on an appropriate interpolation, such as linear or cubic schemes. However, this single translation can reduce the MTF response and distort the phase response because of the employed interpolation scheme [13]. On the other hand, the proposed two-step method can alleviate the degradation problem from the interpolation-based transformation approach. 2.2. Modulation Transfer Function of the Convex Combination Image We analyze the registration performance of the proposed registration method by observing the MTF of a convex combination image. For a pixel position u, let p[u] denote a convex combination [5,8] of the acquired upper and lower images pU and pL, respectively, as with a combination coefficient of λ such that 0≤λ≤1. Here, optimal coefficients of λ can be found in terms of maximizing the NPS or DQE performance [8]. The mean of Equation (1) is given as μ:=λμU+(1−λ)μL, where μU and μL are the means of the upper and lower images, respectively. The frequency performance of imaging systems can be evaluated by measuring the detector MTF, which is the amplitude response acquired from the Fourier transform of the impulse response or the point spread function. We first measure directional MTF curves from the convex combination image p based on a method described in the IEC standard [32]. Here, to avoid overlaps of aliases and increase measurement accuracies [33], the oversampled impulse response is calculated from the derivative of the oversampled step response, which is acquired from upper and lower X-ray images using a tungsten slant-edge phantom. At Position 1 of Figure 3, a directional translation of s0 is estimated based on a subpixel registration method. This step corresponds to Step (0) in preparation for the proposed registration. By multiplying e−jωs in the discrete-time Fourier transform (DTFT) domain, the lower image is aligned based on the Fourier shift theorem [11,31]. This step corresponds to Step (1) of the proposed registration. The directional MTF of the convex combination image, which is denoted as T, then satisfies the following relationship: where ω is the normalized radian frequency given as ω:=2πf/fs and |ω|≤π in the DTFT domain. In (2), a1 and a2 are defined as a1(ω):=2λ(1−λ)μUμLTU(ω)TL(ω) and a2(ω):=λμUTU(ω)2+(1−λ)μLTL(ω)2, respectively, where TU and TL denote the upper and lower MTF curves, respectively [8]. If the estimate s from a subpixel registration method is equal to s=s0, then the maximum MTF of Equation (2) can be rewritten as which is a convex combination and, thus, is between those of the upper and lower layers. Note that a relationship of T(ω,s)≤T(ω,s0) holds. Hence, the lower image should be aligned with the true s0 to maximize the MTF performance. If we conduct the translation based on an interpolation scheme instead of the Fourier shift theorem, then the MTF of Equation (2) is reduced because of the interpolation. Let Ψ denote such a reduced MTF. The MTF from the interpolation is then given as where the constants b1 and b2 are defined as and respectively. In Equations (4)–(6), α and β are the amplitude and phase responses, respectively, from the translation with interpolation. We now observe the amplitude and phase responses of the interpolation-based translation to observe the reduced MTF of Equation (4). Let α1 denote the amplitude response of a translation of s to the right direction based on linear interpolation (B-spline of degree 1 [34]). α1 is then given as for |s|≤1, and is periodic with a fundamental period of 1 [11]. Here, the relationship 0≤α1≤1 is satisfied. In a similar manner, the amplitude response of the translation α3, which is based on a cubic interpolation (Catmull–Rom spline [35]), is given as Because a relationship of α1≤α3≤1 holds, we can alleviate the amplitude reduction problem by using cubic interpolation instead of linear interpolation. In (5) and (6), α can be α1 or α3 for the linear or cubic interpolation, respectively. Furthermore, the phase responses of both linear and cubic interpolations are given as for a translation of s [11]. Examples of the amplitude responses on α1 and α3 and phase responses on β are illustrated in Figure 4. When the translation s is an integer, the amplitude response becomes the largest value, 1, and when the remainder divided by an integer is 0.5, it becomes the smallest value. We can also observe in Figure 4 that the curve near the integer translations in the amplitude response of cubic interpolation is gentler and smoother than the linear interpolation case. Note that Ψ(ω,s)≤T(ω,s) is satisfied, where the equality holds when s is an integer. Hence, if the true translation s0 is not an integer, then Ψ(ω,s0)