**1. Introduction**

Ultra-high-resolution (UHR) computed tomography (CT) has been used in commercial applications since 2017 due to it features of higher image spatial resolution [1]. Kakinuma et al. used a prototype UHR-CT that was operated with 0.25 mm detector pixel size and 0.1 mm reconstructed image pixel interval at 0.25 mm slice thickness [2]. Another experiment with a clinical UHR-CT scanner (Aquilion Precision, Canon Medical Systems) reported that the system was operated in the same condition of that in the previous paper [3]. The clinical UHR-CT has three scan modes: normal, high-resolution (HR), and super-high-resolution (SHR) modes, which support 512 × 512, 1024 × 1024, and 2048 × 2048 image matrixes, respectively, in a given reconstruction field of view (FOV) [4]. Judging from the previously published papers, UHR-CT generally can be distinguished from conventional high-resolution CT (CHR-CT) because it makes use of pixel sizes below 0.25 mm at image matrixes above 512. It has been reported that CHR-CT uses pixel sizes ranging from 0.23 mm [5] to 0.35 mm [6]. The clinical aspects of UHR-CT include reduction of vascular continuity of the coronary arteries, visualization of fine structures of lungs, such as peripheral pulmonary vessels less than 1 mm in size, and artifact reduction such as blooming [7,8].

Recent X-ray detector technology in both multi-row and the flat-panel detectors (FPDs) enables high-resolution acquisition at a small pixel size of less than 0.25 mm [9]. Following these efforts, dedicated UHR cone-beam CT (CBCT), e.g., the OnSight 3D system (Carestream Healthcare, Rochester, NY, USA), has been introduced for extremity scans at lower cost and radiation doses compared to multi-detector CT (MDCT) systems [10]. The OnSight 3D system are mounted with a CsI:Tl scintillator-based complementary metal-oxide semiconductor (CMOS) FPD with a pixel size of 139 µm. UHR-CBCT can ultimately improve the visualization of bone morphometry and contribute to the diagnosis of osteoporosis and osteoarthritis, and detection of fine fractures, which typically require measurements in the range of 0.05–0.2 mm [11]. In general, the FPD could be operated in detector pixel binning mode, which is the process of combining the adjacent electric charges into one pixel [12]. This can reduce both the electronic and quantum noise, and decrease the image readout time at a higher frame rate. The user selects the FPD operation in either full or binning mode, which can optimally satisfy the need of correlation between the resolution and frame rate.

However, the image at higher resolution is not always good, especially for low-contrast detection tasks due to the enhanced noise level during the process of filtered back-projection (FBP) image reconstruction [13]. The "low contrast" of the image can be described as low discrimination between the target and background. The spatial resolution measurement in high-density materials, such as bar pattern and tungsten wire, is an easy task for both standard and UHR CT imaging. However, medical image quality of low-contrast objects is defined in terms of how well the tradeoff relationship between the resolution and noise is obtained from the image [14]. The amount of noise suppression at high frequencies is adjustable by setting either different cutoff frequency levels or different smoothing functions implemented on the CT sinogram. The higher the cutoff frequency level, the sharper but noisier the reconstructed image [15]. This, in turn, results in reconstructed image quality, thereby greatly influencing the detectability of objects by human observers [16]. Unfortunately, choosing an optimal filter scheme relies on experience, because there is no global function that can accept all principal signals underlying the entire frequency range. Therefore, the effect of the reconstruction filters on different materials in UHR-CBCT should be studied to provide useful information when observing a tiny amount of information during UHR acquisition.

In this study, we measured the spatial resolution of five different cylindrical objects according to four different UHR acquisition modes using six different filter schemes. The self-developed UHR-CBCT system, which is installed at the authors' institution (Korea Electrotechnology Research Institute, Ansan, Korea) was used for acquiring the CBCT images in both standard and high-resolution modes. This study aimed to evaluate the effect of filter schemes on the spatial resolution that underlies each imaging object and to suggest the optimal filter scheme in UHR-CBCT depending on the different object materials.
