Internal imaging method based on translation of a compound scan of a source of rays

Abstract

This invention relates to an internal imaging method based on X-ray source translational composite scanning, belonging to the field of internal tomographic imaging technology. The method includes: S1: Constructing a composite sampling mode hySTCT based on X-ray source translation, specifically, using dense sampling of the region of interest (ROI) to acquire high-resolution projection data, and using sparse sampling of the region outside the ROI to acquire a small amount of low-resolution projection data; S2: Based on the linear characteristics of the inverse Radon transform, constructing a V-FBP reconstruction method iV-FBP based on data interpolation and a two-step V-FBP reconstruction method tV-FBP, and applying them to hySTCT scan reconstruction. This invention can effectively suppress truncation artifacts and improve the reconstruction accuracy within the ROI range.

Description

Technical Field

[0001] This invention belongs to the field of internal tomographic imaging technology and relates to an internal imaging method based on X-ray source translational composite scanning. Background Technology

[0002] Micro-CT (micro-nano CT) is a technique for non-destructively imaging the three-dimensional internal microstructure of objects at high spatial resolution, and it is widely used in materials science, microelectronics, geology, non-destructive testing, and earth sciences. However, imaging objects with diameters larger than the imaging field of view (FOV) at high resolution is a challenge. Internal tomography is an effective alternative method that scans only the region of interest (ROI) within the object at high resolution. Furthermore, internal tomography can reduce interference from object structures outside the ROI. For example, in the 3D reconstruction of lithium-ion battery electrodes, internal tomography is a promising technique that can reduce the impact of high-attenuation elements on the image quality of low-contrast carbon binder regions.

[0003] However, internal tomography faces several challenging problems, such as non-unique and unstable solutions, as well as truncation artifacts, which severely impact its application value. Over the past few decades, techniques used to address these problems can be categorized into five types: projection extrapolation, analytical algorithms, iterative algorithms, deep learning, and hardware-adjusted scanning techniques. Projection extrapolation methods mitigate truncation artifacts by extrapolating truncated data using smoothing functions, such as symmetric mirror extrapolation, cosine function fitting extrapolation, cylindrical extrapolation, or deep learning extrapolation. Due to the lack of consistency in projection data extrapolation, these methods are useful for qualitative analysis but inaccurate for quantitative analysis. Analytical algorithms such as differential back projection (DBP) can achieve accurate and stable reconstructions given a known subregion within the ROI; however, these assumptions are not always met in actual imaging. Iterative algorithms such as total variation minimization can obtain unique and stable results when the reconstructed region satisfies the condition of a piecewise constant or polynomial; however, in practical applications, the reconstructed region does not always meet such conditions. Recently, deep learning techniques have offered a novel approach to solving the truncation problem, but their practical deployment can be challenging due to a lack of generalization, interpretability, and sufficient training data. Adjusted hardware scanning is a method for artifact-free imaging of regions of interest (ROIs) with acceptable accuracy. Its mathematical principle is that projected data outside the ROI still produces a non-zero two-dimensional image within the ROI.

[0004] To subtract the ray attenuation value of the outer region from the truncated projection data, a widely used technique is to perform a low-resolution scan of the entire object by changing the geometric magnification ratio or using a large detector. Changing the magnification ratio requires solving the projection data registration problem caused by inconsistent magnification ratios, which increases computational costs and destroys high-frequency information. Using a large detector can alleviate this problem, but it increases the cost and complexity of the system. Dual-field-of-view optical coupling detection systems can simultaneously acquire low-resolution projection data of the entire object and high-resolution projection data of the truncated ROI by dividing the light into two secondary optical paths; however, it suffers from low detection efficiency and high noise. In our previous work, to avoid changing the magnification ratio and increasing system complexity, we used a multi-segment linear translational CT (mSTCT) scanning method to acquire low-resolution projection data of the entire object and a circular scanning method (fixed ray source and detector, rotating object) to acquire high-resolution projection data of the ROI. However, fusing the projection data acquired in the two different scanning modes introduces errors.

[0005] Therefore, a new internal tomographic imaging method is urgently needed to solve the problem of low sampling efficiency in mSTCT. Summary of the Invention

[0006] In view of this, the purpose of this invention is to provide an intra-articular imaging method based on X-ray source translation composite scanning, which uses a V-FBP reconstruction method based on data interpolation (iV-FBP) and a two-step V-FBP reconstruction method (tV-FBP) to realize hySTCT scan reconstruction, effectively suppressing truncation artifacts and improving reconstruction accuracy within the ROI range.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] An internal imaging method based on X-ray source translational composite scanning specifically includes the following steps:

[0009] S1: Construct a composite sampling mode hySTCT based on X-ray source translation, specifically: use dense sampling of X-ray sources to obtain high-resolution projection data for the ROI (Region of Interest), and use sparse sampling of X-ray sources to obtain a small amount of low-resolution projection data for the region outside the ROI.

[0010] S2: Construct a V-FBP reconstruction method based on data interpolation (iV-FBP) and a two-step V-FBP reconstruction method (tV-FBP), and apply them to hySTCT scan reconstruction; where V-FBP represents a filtered back projection algorithm based on virtual projection.

[0011] Furthermore, step S1 specifically includes: in the hySTCT sampling mode, dense sampling is performed on the ROI region, while sparse sampling is performed on the region outside the ROI; the non-equidistant sampling method of the X-ray source can be expressed as:

[0012]

[0013] in, This refers to the position coordinates of the X-ray source during sampling in hySTCT scanning mode, where l is the distance from the X-ray source to the object; It is the downsampling coefficient, used to control the size of the sampling interval. Represents non-zero natural numbers; N1 = 2(s-s0) / κΔλ, and N0 = 2s0 / Δλ, where θ is the trajectory of the X-ray source and the rotation angle of the detector, Δλ is the sampling interval of the X-ray source, s represents half the distance of the X-ray source's linear motion trajectory, and s0 is half the motion distance when the X-ray source scans an ROI with a radius of R0. The geometric relationship is expressed as:

[0014]

[0015] Where h is the distance from the detector to the object, and d is half the width of the detector;

[0016] This sampling mode generates two types of projection data: one is ROI projection data obtained by dense sampling with a sampling interval of Δλ. |λ|≤s0; and another type of sparsely sampled projection data of regions outside the ROI with a sampling interval of κΔλ: s0<|λ|≤s, where λ represents the sampling point of the X-ray source and u represents a single detector unit.

[0017] Furthermore, step S2 specifically includes: the reconstruction of data obtained using the hySTC scanning model can be represented as:

[0018]

[0019] Formula (1) is simply called tV-FBP, which is a two-step V-FBP reconstruction method. According to the linear transformation characteristics of V-FBP(·), the two-step V-FBP reconstruction can be converted into a one-step V-FBP reconstruction, as follows:

[0020]

[0021] However, since hySTCT uses non-equidistant sampling to obtain projection data, formula (2) cannot be directly implemented. This invention uses an interpolation operator I(·) to interpolate the sparsely sampled data, expressed as:

[0022]

[0023] Then, formula (2) becomes

[0024]

[0025] Formula (4) is called iV-FBP, which is the V-FBP reconstruction method based on data interpolation.

[0026] The beneficial effects of this invention are as follows:

[0027] (1) This invention employs a composite sampling mode based on X-ray source translation (i.e., the hySTCT scanning model) for internal tomographic imaging. hySTCT performs dense sampling on the ROI and sparse sampling on the regions outside the ROI, which can reduce truncation artifacts and grayscale deviations in the ROI reconstructed image.

[0028] (2) Based on the linearity of the Radon inverse transform, and inspired by the previous filtered back projection (V-FBP) algorithm based on virtual projection, this invention proposes a V-FBP reconstruction method based on data interpolation (iV-FBP) and a two-step V-FBP reconstruction method (tV-FBP) for hySTCT scan reconstruction. This reconstruction method can effectively suppress truncation artifacts and improve the reconstruction accuracy within the ROI range.

[0029] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0030] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein:

[0031] Figure 1 This is a schematic diagram of internal tomographic imaging.

[0032] Figure 2 The diagrams are schematics of mSTCT and hySTCT, where (a) is the three-dimensional geometry of mSTCT, (b) is the two-dimensional geometry of mSTCT, and (c) is the two-dimensional geometry of hySTCT.

[0033] Figure 3 For a real projection with data truncation (a) and a virtual projection without truncation (b);

[0034] Figure 4 Schematic diagram of V-FBP, tV-FBP and iV-FBP reconstruction process;

[0035] Figure 5 The global reconstruction results are shown for different numbers of external projections using tV-FBP and iV-FBP methods respectively; where (a)-(f) are the reconstruction results of tV-FBP with 1000, 400, 200, 100, 50 and 0 external projections respectively; (g)-(l) are the reconstruction results of iV-FBP in the corresponding cases respectively.

[0036] Figure 6 The images are ROI images reconstructed using tV-FBP and iV-FBP methods for different numbers of external projections; where (a)-(f) and (m)-(r) are ROI images with 1000, 400, 200, 100, 50 and 0 external projections, respectively; (g)-(l) and (s)-(x) are the corresponding difference images;

[0037] Figure 7 The images show the reconstruction results using different methods; the first row is the ROI image, the second row is the corresponding difference image, the first four columns show the ROI image with a field of view radius of 4.5mm, and the last two columns show the ROI image with a field of view radius of 2.1mm.

[0038] Figure 8 The results are global reconstructions using tV-FBP and iV-FBP methods under different amounts of external projection data; where (a)-(f) are the tV-FBP reconstruction results with 800, 400, 200, 100, 50 and 0 external projections, respectively; (g)-(l) are the reconstruction results under the corresponding iV-FBP conditions.

[0039] Figure 9 The images are ROI images reconstructed using tV-FBP and iV-FBP methods under different amounts of external projection data; where (a)-(f) and (m)-(r) are ROI reconstructed images with external projection numbers of 800, 400, 200, 100, 50 and 0, respectively, and (g)-(l) and (s)-(x) are the corresponding difference images;

[0040] Figure 10 To reconstruct ROI images using different scanning methods; the first row displays the ROI image, the second row displays the corresponding difference image, the first four columns display the ROI image with a field of view radius of 3.0 mm, and the last two columns display the ROI image with a field of view radius of 1.2 mm. Detailed Implementation

[0041] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0042] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0043] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0044] Please see Figures 1-10 This invention proposes a composite sampling scanning model based on X-ray source translation for internal tomographic imaging. High-resolution projection data is acquired through dense sampling of the region of interest (ROI), while a small amount of low-resolution projection data is acquired through sparse sampling of the region outside the ROI. This composite sampling mode is called hySTCT. hySTCT can directly acquire data of different resolutions, avoiding the data fusion process. This work is inspired by our previous work on mSTCT scanning mode, where each X-ray source sampling position only illuminates a portion of the object, resulting in truncated projection. Therefore, this invention regroups the X-rays emitted from each X-ray source position into virtual X-rays emitted from each detector unit, forming a non-truncated virtual projection, and further derives a virtual projection-based FBP algorithm (V-FBP). However, mSTCT suffers from low sampling efficiency. hySTCT, which focuses only on imaging within the ROI, can effectively improve efficiency.

[0045] 1. Basic Theory of Internal Tomography

[0046] Suppose a function Its Radon transform can be expressed as

[0047]

[0048] The corresponding Radon transform representation of the internal region of interest is as follows: Where R0 (R0 < R) is the radius of the interior region of interest. Natterer proved that for any result reconstructed from the projection data of the interior region of interest, it can be regarded as an accurate reconstruction result from the complete projection data. Reconstruction result after subtracting one external region projection data Represented as

[0049]

[0050] in, Inverse Lardon transform. Despite However, when it performs the inverse Radon transform, it generates a non-zero two-dimensional local image in the ROI region. This explains why internal tomographic imaging produces truncation artifacts and non-unique solutions, such as... Figure 1 As shown.

[0051] If projection data outside the region of interest is obtained This will give you a rough global image, represented as

[0052]

[0053] Although projection data outside the region of interest (ROI) cannot reconstruct a high-quality external region image, it can improve the quality of the reconstructed ROI image, including reducing truncation artifacts and improving reconstruction accuracy. According to formula (3), the reconstructed ROI image can be expressed as...

[0054]

[0055] like Figure 1 As shown, Figure 1 The first row shows the dense and sparse sampling projections outside the ROI; Figure 1 The second, third, and fourth lines respectively show the reconstruction process represented by equations (2), (3), and (4) in the basic theory of internal tomography. Figure 1 The study demonstrates a significant improvement in ROI reconstruction quality after incorporating a partial external dataset. Therefore, several different methods for collecting partial external data are proposed, including low-resolution scanning of the entire sample and small-scale complete and low-dose scanning of the entire sample. These strategies produce a multi-resolution global image, such as... Figure 1As shown in the third row, the image has high resolution within the ROI and low resolution outside the ROI. This invention proposes using mSTCT scanning to obtain low-resolution external projections to improve the image quality of the ROI. ROI reconstruction methods can be broadly classified into five categories: projection extrapolation, analytical reconstruction algorithms, iterative reconstruction algorithms, deep learning reconstruction, and hardware-adjusted reconstruction techniques.

[0056] 2. hySTCT scanning method and its reconstruction algorithm

[0057] 1) mSTCT and hySTCT imaging

[0058] Figure 2 The image shown is an mSTCT imaging model, which consists of multiple STCTs with different rotation angles. Figure 2 (a)). In each STCT scan, the object is brought close to the X-ray source, while the X-ray source is translated equidistantly along a straight line parallel to the detector plane, thereby acquiring projection data of the entire object. Figure 2 (b)). This invention can represent the position of the discrete source on the trajectory in each STCT as follows:

[0059]

[0060] Where θ represents the trajectory of the X-ray source and the rotation angle of the detector, and Δλ represents the sampling interval of the X-ray source. And N = 2s / Δλ. Here, s represents half the distance of the linear trajectory of the ray source. The ray source can scan and reconstruct an object with radius R at this distance, and the geometric relationship is expressed as:

[0061]

[0062] Where l is the distance from the X-ray source to the object; h is the distance from the detector to the object; and d is half the width of the detector. When Δλ < Δu / 2k (Δu is the detector element size), mSTCT can avoid resolution loss, but high-resolution imaging of large objects will require a large number of sampling points per STCT scan. Therefore, in order to reduce the number of sampling points, this invention proposes a hySTCT scanning model that images only the region of interest (ROI).

[0063] During hySTCT scanning, dense sampling is performed within the Region of Interest (ROI), while sparse sampling is performed outside the ROI. Therefore, the non-equidistant sampling method of the X-ray source can be expressed as:

[0064]

[0065] in, This is the downsampling coefficient, used to control the size of the sampling interval; N1 = 2(s-s0) / κΔλ, and N0 = 2s0 / Δλ. Here, s0 is half the distance traveled when the ROI radius of the X-ray source is R0, and the geometric relationship is expressed as follows:

[0066]

[0067] This sampling mode generates two types of projection data: one is ROI projection data obtained by dense sampling with a sampling interval of Δλ.

[0068]

[0069] And another type of sparsely sampled projection data of regions outside the ROI with a sampling interval of κΔλ.

[0070]

[0071] Where λ represents the X-ray source sampling point, and u represents a single detector unit.

[0072] The differences between mSTCT and hySTCT are as follows:

[0073] (1) When κ=1, in[-s,-s0)∪(s0,s], hySTCT becomes mSTCT;

[0074] (2) When κ=∞, in[-s,-s0)∪(s0,s], hySTCT only obtains ROI projection data.

[0075] 2) Analytical Reconstruction Algorithm

[0076] During mSTCT scanning, the X-ray source emits rays that only cover a portion of the object at each sampling location, resulting in data truncation. Figure 3 As shown in (a), this invention proposes a virtual projection-based filtered back-projection algorithm, V-FBP, for image reconstruction. Virtual projection is as follows... Figure 3 As shown in (b), the relationship between virtual projection and real projection can be expressed as follows:

[0077]

[0078] Then, the V-FBP algorithm can be expressed as:

[0079]

[0080] in, Here, w θ The weighting function used to process redundant data is the q-slope filter. The optimal mSTCT scan reconstruction image is obtained by summing the reconstructed images from each STCT scan segment. The V-FBP reconstruction algorithm is implemented as follows, with the projection data p discretized. θ(λ,u) is:

[0081]

[0082] The projected data, after geometric weighting and redundancy weighting, is represented as follows:

[0083]

[0084] The one-dimensional Fourier transform of the weighted data with respect to the variable λ is:

[0085]

[0086] in The projection data is represented after high-pass filtering as follows

[0087]

[0088] in The reconstruction result of each STCT scan segment can be represented as:

[0089]

[0090] The final mSTCT scan reconstruction result can be expressed as:

[0091]

[0092] Where T represents the number of STCT scans.

[0093] This invention treats the V-FBP reconstruction algorithm as a reconstruction operator V-FBP(·). Clearly, V-FBP(·) is an inverse Radon transform process. Based on formula (3), the data reconstruction obtained using the hySTC scanning model can be expressed as...

[0094]

[0095] Formula (19) is simply referred to as the two-step V-FBP reconstruction method—tV-FBP. Based on the linear transformation characteristics of V-FBP(·), the two-step V-FBP reconstruction can be converted into a one-step V-FBP reconstruction, expressed as:

[0096]

[0097] However, since hySTCT uses non-equidistant sampling to obtain projection data, formula (20) cannot be directly implemented. This invention uses an interpolation operator I(·) to interpolate the sparsely sampled data, denoted as:

[0098]

[0099] Then, formula (20) becomes

[0100]

[0101] Formula (22) is known as the V-FBP reconstruction method based on data interpolation—iV-FBP.

[0102] 3. Experiment

[0103] 1) Simulation Experiment

[0104] To evaluate the effectiveness of the proposed hySTCT and its reconstruction algorithms tV-FBP and iV-FBP in internal computed tomography (ICT) scans, experiments were conducted using a Shepp-Logan simulation phantom with dimensions of 8.57 mm × 8.57 mm. The simulated geometric parameters were: distance from the X-ray source to the object 15 mm, distance from the object to the detector 190 mm, detector size 1024 × 1024 pixels, individual detector size 0.127 mm × 0.127 mm pixels, and an interval angle of 36.5 degrees between adjacent STCT scans. A total of 5 STCT scans were performed. Here, the X-ray source translation distance 2s was set to 30 mm to cover the entire phantom, and 2s' was set to 15 mm to achieve scanning of a ROI with a radius of 2.1 mm. Within the ROI, the X-ray source sampling interval was set to 15 μm, corresponding to 1000 ROI sampling points per STCT scan segment.

[0105] First, to investigate the impact of external projection data on the accuracy of ROI reconstruction, we set downsampling factors to ∞, 20, 10, 5, 2.5, and 1, resulting in external projection counts of 0, 50, 100, 200, 400, and 1000. When the downsampling factor was ∞, hySTCT collected only the projection data of the ROI region; when the downsampling factor was 1, hySTCT became mSTCT, and the entire object was scanned at equal intervals to acquire projection data. All projection data were reconstructed using tV-FBP and iV-FBP, respectively.

[0106] Secondly, to highlight the effectiveness of hySTCT, this experiment compared it with previous work—the mSTCT-PC method—under the same geometric parameters. In mSTCT-PC, an RCT scanning method (object rotation to obtain projection) was used to uniformly acquire 1500 projections over 360 degrees, obtaining ROI projection data with a fixed field of view radius of 4.5mm; using the mSTCT scanning mode, 21 STCT sampling points were used per segment, for a total of 5 STCT scans, to obtain the projection data of the entire object.

[0107] We use three evaluation metrics—root mean square error (RMSE), peak signal-to-noise ratio (PSNR), and structural similarity index (SSIM)—to quantitatively evaluate the image quality of the region of interest. Generally, a lower RMSE and higher SSIM and PSNR indicate higher image quality.

[0108] 2) Actual Experiment

[0109] To demonstrate the feasibility of hySTCT, we conducted a practical experiment on a micro-nano CT system, which consists of a microfocus X-ray source (L10321, Hamamatsu, Japan), a flat panel detector (PaxScan1313DX, Varian, USA), a rotating platform (RGV100BL, Newport, USA), and a translation platform (M-ILS250PP, Newport, USA). In the experiment, the X-ray source operating current and voltage were 60 kV and 70 μA, respectively. The geometric parameters used were consistent with those in the simulation experiment. An experimental sample with a scanning radius of 5 mm was scanned. We set the X-ray source translation distance 2s to 20 mm to cover the entire experimental sample, and set 2s' to 12 mm to scan a ROI with a radius of 1.2 mm. During the ROI scanning, the X-ray source sampling interval was 10 μm, requiring 1200 sampling points for the ROI scan. For scanning the outer region of the ROI, the X-ray source sampling intervals are ∞, 160, 80, 40, 20 and 10 μm, respectively, so that the number of external sampling points is 0, 50, 100, 200, 400 and 800.

[0110] To verify the performance of hySTCT, we conducted simulations and real-world experiments. The reconstruction program was written in MATLAB and tested on a computer configured with an Intel® Core™ i7-4790 CPU @ 3.60GHz. The experiments mainly included evaluating the ROI image reconstruction quality under different sparse sampling point conditions and comparing the reconstruction results between different methods.

[0111] 4. Experimental Results:

[0112] 1) Simulation Experiment Results

[0113] 1.1) Reconstruction under different ROI external projection data volumes

[0114] Figure 5 Global reconstructed images of tV-FBP and iV-FBP under different numbers of external projections are shown in hySTCT scanning mode. For both reconstruction methods, severe artifacts are present in the reconstructed images without external projection data. Figure 5 (f) and Figure 5(l)). When external projection data is available, artifacts in the reconstructed image are significantly reduced. As the number of external projections increases, artifacts gradually decrease. Figure 5 It also shows that when the number of external projections is small, tV-FBP has better artifact suppression performance than iV-FBP. Figure 5 (e) and Figure 5 (k) This is due to inaccurate interpolation of the projection data outside the ROI by iV-FBP. As the number of external projections increases, the difference between tV-FBP and iV-FBP becomes smaller and smaller.

[0115] Figure 6 The RMSE values ​​of the reconstructed ROI images and the difference images with the standard phantom were compared. The results show that without ROI external projection data, the reconstructed ROI images exhibit significant grayscale deviations due to data truncation. With ROI external projection data, the numerical deviations are significantly reduced, which can be verified by the RMSE values ​​and the difference images. When reconstructing ROIs using the tV-FBP method, the RMSE value decreases significantly with the increase of the number of external projections. However, in ROI reconstruction using the iV-FBP method, the RMSE value of the reconstructed image with N1=50 external projections is comparable to that with N1=1000; the RMSE value of the ROI reconstructed image with N1=1000 is equal to that with N1=200. This indicates that, compared to mSTCT, the hySTCT scanning model combined with the iV-FBP reconstruction method can reduce sampling points by hundreds per STCT scan segment while reconstructing the same ROI image quality, thereby improving sampling efficiency.

[0116] Table 1 lists the quantitative evaluation metrics calculated from the reconstructed images based on the ROI. For both tV-FBP and iV-FBP reconstruction methods, as the number of ROI external projections increases, the RMSE value decreases with increasing PSNR and SSIM values, indicating improved image quality. Furthermore, iV-FBP performs better than tV-FBP, especially with a smaller number of external projections. For iV-FBP, the image quality with downsampling factors of 5 or 10 is comparable to that of the reconstructed images from mSTCT.

[0117] Table 1 Quantitative Evaluation Indicators for ROI Reconstructed Images

[0118]

[0119] 1.2) Method Comparison

[0120] Figure 7The ROI reconstruction results of hySTCT were compared with those of RCT and mSTCT-PC. For RCT, two differences were observed: First, at a fixed geometric magnification, the ROI size reconstructed by RCT is fixed, while the ROI size reconstructed by hySTCT can be adjusted by changing the translation distance of the X-ray source. Second, the effects of data truncation differ between the two scanning modes. RCT reconstruction results show cupping artifacts and grayscale deviations at the ROI edges, while hySTCT reconstruction results show grayscale deviations throughout the ROI and a small number of artifacts at the edges. For mSTCT-PC, the RMSE values ​​and difference images show that hySTCT has better accuracy and smaller grayscale deviations than mSTCT-PC. This is because mSTCT-PC inevitably introduces some errors when fusing projection data from two different scanning methods (RCT and mSTCT). Furthermore, these errors are larger than those caused by iV-FBP interpolation. Figure 7 It can also be seen that in hySTCT, the reconstruction accuracy decreases as the degree of data truncation increases.

[0121] 2) Actual experimental results

[0122] 2.1) Reconstruction under different ROI external projection data volumes

[0123] Figure 8 The globally reconstructed image based on actual scan data is displayed. Experimental results show that adding projection data outside the ROI can improve the quality of the globally reconstructed image. When the amount of projection data outside the ROI is small, tV-FBP performs well in the global reconstructed image restoration.

[0124] In actual experiments, mSTCT scan reconstructed images are used as standard reference images to calculate quantitative evaluation indicators and difference images. Figure 9 The ROI reconstruction results are shown. For both tV-FBP and iV-FBP reconstruction methods, when external projection data of the ROI is available, both methods significantly reduce artifacts and grayscale deviations at the edges of the ROI image. Compared to tV-FBP, iV-FBP can better suppress stripe artifacts caused by data truncation at the edges of the region of interest, but slight artifacts may occur within the region of interest due to interpolation.

[0125] Table 2 summarizes the quantitative comparison results of ROI images with different numbers of external projections. Compared with the reconstruction results without external projections, both methods significantly improve the quality of the reconstructed ROI images with external projection data, resulting in lower RMSE and higher PSNR and SSIM. When the number of external projections increases from 50 to 400, the iV-FBP method is more stable than the tV-FBP method, with smaller fluctuations in quantitative evaluation indicators. However, as the amount of external projections gradually increases, the tV-FBP method can further improve image accuracy.

[0126] Table 2 Quantitative Evaluation Indicators of ROI Reconstructed Images under Different Numbers of External Projections

[0127]

[0128]

[0129] 2.2) Method Comparison

[0130] Figure 10 The results of ROI reconstruction using different methods are shown. The proposed method, when external ROI data is available, demonstrates superior ROI image accuracy compared to other methods, with no significant bias in the difference image. ROI images reconstructed by RCT scan exhibit severe cupping artifacts due to truncation of the projection data. The mSTCT-PC method improves the quality of the reconstructed ROI image by completing the truncated ROI data. Even without external projection data, the reconstructed image by the proposed method is still affected by some artifacts at the edges, but these are much less than those of the other two methods. Since the proposed method can flexibly adjust the imaging field of view simply by changing the source sampling distance, we also demonstrate its ROI reconstruction image with a smaller imaging field of view (radius of 1.2 mm). This also proves that data obtained by sparsely sampling the ROI exterior using hySTCT can significantly reduce truncation artifacts and grayscale deviations in the ROI image.

[0131] 5. Comparative Analysis

[0132] In existing technologies, the mSTCT scanning model increases the imaging field of view (FOV) by controlling the translation of the X-ray source, and a corresponding V-FBP reconstruction algorithm is provided to address the problem of projection data truncation. However, as the FOV increases, the number of sampling points also increases, leading to a decrease in sampling efficiency. In some applications, we are only interested in a local area of ​​the scanned object. Therefore, this invention proposes a hySTCT scanning method for internal ROI imaging by shortening the X-ray source translation distance, ultimately reducing the number of projections. However, data truncation is unavoidable in ROI imaging. Due to missing data outside the ROI causing data truncation, we propose using hySTCT to sparsely acquire projection data outside the ROI, reducing the impact of external data on ROI image quality without sacrificing sampling efficiency. Since the sparsely sampled data outside the ROI and the densely sampled data inside are obtained under the same hySTCT scanning mode with the same geometric parameters, complex data fusion is not required. This invention can directly utilize the linear transformation characteristics of the inverse Radon transform to obtain the reconstructed image. Because the sampling rates of the data outside and inside the ROI are different, the reconstructed global image exhibits different resolutions: high resolution inside the ROI and low resolution outside the ROI. Although the reconstructed image outside the ROI has lower quality, a small amount of external data can significantly improve the quality inside the ROI.

[0133] Compared to RCT for acquiring ROI projection data, hySTCT, which acquires ROI projection data via mSTCT scanning, has the following advantages: First, hySTCT allows for changing the size of the imaged ROI by controlling the translation of the X-ray source; second, data truncation has a smaller impact on the reconstructed image, mainly manifesting as stripe artifacts and grayscale deviations at the edges of the ROI image. Compared to mSTCT-PC acquiring ROI data via RCT and acquiring external datasets via mSTCT, hySTCT acquires both ROI data and external data under the same scanning model and parameters, avoiding the loss of accuracy and resolution caused by data fusion.

[0134] Based on the linear transformation characteristics of the V-FBP operator, this invention proposes two algorithms, tV-FBP and iV-FBP, for hySTCT scan reconstruction. Experimental results show that, for both tV-FBP and iV-FBP methods, external projection data helps suppress artifacts at the edges of the ROI image and internal grayscale deviations. Compared to tV-FBP, iV-FBP can better suppress artifacts caused by data truncation because it uses interpolation to estimate missing external projection data. Therefore, iV-FBP exhibits stronger robustness in reconstruction under different numbers of external projections. However, with the help of deep learning techniques, more accurate estimates can be obtained in the future, which may further reduce the number of external projections required for high-precision ROI reconstruction.

[0135] 6. Summary

[0136] This invention proposes a hySTCT scanning mode and presents two reconstruction algorithms: tV-FBP and iV-FBP. Experimental results show that this scanning mode can achieve ROI reconstruction with variable size, few truncation artifacts, and acceptable accuracy. Furthermore, the results indicate that iV-FBP reconstruction is more stable than tV-FBP reconstruction under conditions of minimal external projection supplementation. This method may be used for microscale internal tomographic imaging, such as battery imaging, where internal active particles, binders, and porous phases directly affect battery performance. This method can be used to image the interior of batteries, thereby aiding in battery structural analysis and performance improvement.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An internal imaging method based on X-ray source translational composite scanning, characterized in that, The method specifically includes the following steps: S1: Construct a composite sampling mode hySTCT based on ray source translation, specifically: use dense ray source sampling to obtain high-resolution projection data for the ROI, and use sparse ray source sampling to obtain a small amount of low-resolution projection data for the region outside the ROI; where ROI represents the region of interest. Step S1 specifically includes: in the hySTCT sampling mode, dense sampling is performed on the ROI region, while sparse sampling is performed on the region outside the ROI; the non-equidistant sampling method of the X-ray source is represented as follows: in, This refers to the position coordinates of the X-ray source during sampling in hySTCT scanning mode; l It is the distance from the radiation source to the object; It is the downsampling coefficient. Represents non-zero natural numbers; , ,as well as , It is the trajectory of the X-ray source and the rotation angle of the detector. It is the sampling interval of the X-ray source. s This represents half the distance of the linear trajectory of the ray source. The radius of the ROI scanned by the X-ray source is R Half the distance traveled at 0; This sampling mode produces two types of projection data: one is... The ROI projection data obtained by dense sampling at sampling intervals: ; and another type of Projected data for regions outside the ROI with sparse sampling intervals: ,in, Represents the sampling point of the radiation source. u Represents a single detector unit; The radius of the ROI scanned by the X-ray source is R Half the distance traveled at time 0 can be geometrically represented as: ,in, h It is the distance from the detector to the object. d It is half the width of the detector; S2: Based on the linear properties of Radon inverse transform, construct a V-FBP reconstruction method iV-FBP based on data interpolation and a two-step V-FBP reconstruction method tV-FBP, and apply them to hySTCT scan reconstruction; where V-FBP represents a filtered back projection algorithm based on virtual projection. The data reconstructed using the hySTC scanning model is represented as follows: （1） Formula (1) is simply referred to as tV-FBP, which is a two-step V-FBP reconstruction method; according to Linear transformation features, transforming two-step V-FBP reconstruction into one-step V-FBP reconstruction, are expressed as: （2）。 2. The internal imaging method based on X-ray source translational composite scanning according to claim 1, characterized in that, In step S2, an interpolation operator is used. To interpolate sparsely sampled data, it is represented as: （3） Then, formula (2) becomes （4） Formula (4) is called iV-FBP, which is the V-FBP reconstruction method based on data interpolation.

Images