Offset geometry for extended field-of-view in multi-contrast and multi-scale X-ray microtomography of lung cancer lobectomy specimens
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.
Kai: We're looking at the paper: "Offset geometry for extended field-of-view in multi-contrast and multi-scale X-ray microtomography of lung cancer lobectomy specimens".
Mira: Here's the gist of the paper. X-ray microtomography is a powerful non-destructive technique allowing 3D virtual histology of resected human tissue The authors describe an efficient method to double the maximum field-of-view of a cone-beam X-ray microtomography system, without sacrificing on spatial resolution, and including multi-contrast capabilities They demonstrate an experimental realisation of the method, achieving exemplary reconstructions of a resected human lung sample, with a cubic voxel of 10.5 µm linear dimensions, across a horizontal field-of-view of 4.3 cm The methodology can be applied at a range of different length-scales and geometries, and that it is directly compatible with complementary implementations of X-ray phase-contrast imaging System Limitations and Contrast Enhancement Conventional XCT suffers from inherently low X-ray attenuation cross section when applied to low-density materials such as lung tissue X-ray phase-contrast imaging (XPCI) is proposed as a solution to this problem because it can achieve improvements in contrast-to-noise ratio (CNR) compared to attenuation-based imaging Laboratory sources often impose limitations on the field-of-view due to optical elements, which can impose a further limitation on the field-of-view Techniques employing optical elements rely on either phase or attenuation effects to structure the beam in a way that creates phase sensitivity Proposed Geometry and Reconstruction The proposed method utilizes an offset COR geometry with a full 360° scan to increase the achievable FOV in imaging of lung lobectomy specimens This technique is demonstrated at multiple length scales, showing compatibility with multiple contrast-mechanisms The reconstruction is carried out using offset-COR redundancy weighting, followed by reconstruction without rebinning using a vector backprojection algorithm The offset geometry allows up to 2x the FOV to be imaged without sacrificing pixel size, spatial resolution, or X-ray flux density per detector element Experimental Validation and Results The system was utilized to image a section of resected human lung tissue of approximately 4 x 4 x 2 cm3 For the beam-tracking experiment, the sample was placed inside of a sealed polypropylene container The resulting horizontal FOV increased by a factor of 1.7x up to 4.3 cm, sufficient to cover the entire sample The reconstructed volumes allow for tracing pulmonary vasculature, with both venous (red arrows) and arterial (blue arrows) systems labelled in figure 4b The redundancy weighting procedure allows the reconstruction of accurate volumes without the necessity to use iterative methods This approach is straightforward to adapt to other lab-based X-ray phase-contrast imaging systems, such as those based on free-space propagation The ability to tailor FOV to the particular sample, without sacrificing spatial resolution, would be highly beneficial in such scenarios where rapid reconstructions are required This method is inherently compatible and stackable with FOV-extension methods such as tiled-gratings The technique can be easily adapted to a wide range of existing XPCI and conventional systems Page 1 --- Offset geometry for extended field-of-view in multi-contrast and multi-scale X-ray microtomography of lung cancer lobectomy specimens Harry Allan1,2, Adam Doherty1,2, Carlos Navarrete-Leon1,2, Oriol Roche i Morgo1,27 Yunpeng Jia1,28 Charlotte Percival3 Zoe Hagel3 Kate E J Otter3 Chuen Ryan Khaw3 Kate Gowers3 Helen Hall39 Sam M Janes3 Fleur Monk4 David Moore45 Joseph Jacob6 Marco Endrizzi1,2 1Department of Medical Physics and Biomedical Engineering, University College London, London, UK 2X-ray Microscopy and Tomography Laboratory, The Francis Crick Institute, London, UK 3UCL Respiratory, University College London, London, UK 4Department of Cellular Pathology, University College London Hospitals NHS Foundation Trust, London,UK 5CRUK Lung Cancer Centre of Excellence UCL Cancer Institute University College London,London,UK 6Centre for Medical Image Computing University College London,London,UK X-ray microtomography is a powerful non-destructive technique allowing 3D virtual histology of resected human tissue The achievable imaging field-of-view, is however limited by the fixed number of detector elements, enforcing the requirement to sacrifice spatial resolution in order to image larger samples In applications such as soft-tissue imaging, phase-contrast methods are often employed to enhance image contrast Some of these methods, especially those suited to laboratory sources, rely on optical elements, the dimensions of which can impose a further limitation on the field-of-view We describe an efficient method to double the maximum fieldof-view of a cone-beam X-ray microtomography system, without sacrificing on spatial resolution, and including multi-contrast capabilities We demonstrate an experimental realisation of the method, achieving exemplary reconstructions of a resected human lung sample, with a cubic voxel of 10.5 µm linear dimensions, across a horizontal field-of-view of 4.3 cm The same concepts are applied to free-space propagation imaging of a 2.7 mm segment of the same sample, achieving a cubic voxel of 450 nm linear dimensions We show that the methodology can be applied at a range of different length-scales and geometries, and that it is directly compatible with complementary implementations of X-ray phase-contrast imaging 1 Introduction X-ray computed tomography (XCT), has proven to be an indispensable tool for the non-destructive 3D inspection of samples in a range of fields including medical sciences, materials engineering, and metrology
1: X-ray microtomography (micro-CT)
2: allows the probing of structures on the micro-scale Access to 3D morphology has allowed detailed virtual histology of diseased human lung structure in cases such as fibrosis
3: , Covid-19
4: , and adenocarcinomas
5: Volumetric information from micro-CT enables 3D quantification of disease biomarkers
6: , which can be demonstrably more sensitive than their 2D counterparts
7: enhancing the understanding of disease and the effects of treatment Conventional XCT reconstructs 3D distributions from many 2D projections formed predominantly by the attenuation of X-rays by the object in the beam Yet despite its widespread adoption, the application of XCT to low-density materials, such as lung-tissue, still suffers from their inherently low X-ray attenuation cross section X-ray phase-contrast imaging (XPCI)
8: is capable of achieving improvements in contrast-to-noise ratio (CNR) compared to attenuation-based imaging, and has thus been proposed as a solution to this problem XPCI forms images based on phase-shifts imparted on the X-ray beam by the object, typically requiring beams with high levels of spatial and temporal coherence, such as those found at synchrotron radiation facilities
9, 10: Several techniques have been developed to enable the use of XPCI techniques with conventional sources, increasing their availability in laboratory and clinical settings XPCI based only on free-space propagation (FSP) of the perturbed wavefront is achievable using laboratory sources
11: however this comes at the cost of stringent requirements on source and detector characteristics A number of other methods employ optical elements to alleviate some of these difficulties, including grating interferometry
12: , edge-illumination
13: , beam-tracking
14: , and speckle-tracking
15, 16: Each of these methods is able to recover independent attenuation and differential-phase signals, caused respectively by the imaginary (β) and real (δ) components of the sample’s complex refractive index Laboratory-based XPCI micro-CT has been utilised in a range of applications via a number of methods Page 2 --- (FSP) of the perturbed wavefront is achievable using laboratory sources
11: however this comes at the cost of stringent requirements on source and detector characteristics A number of other methods employ optical elements to alleviate some of these difficulties, including grating interferometry
12: , edge-illumination
13: , beam-tracking
14: , and speckle-tracking
15, 16: Each of these methods is able to recover independent attenuation and differential-phase signals, caused respectively by the imaginary (β) and real (δ) components of the sample’s complex refractive index Alongside these quantities, they are also sensitive to the unresolved refraction and ultra-small angle X-ray scattering (USAXS) that leads to the dark-field contrast channel Laboratory-based XPCI micro-CT has been utilised in a range of applications via a number of methods Page 3 --- Techniques employing optical elements rely on either phase or attenuation effects to structure the beam in a way that creates phase sensitivity The size of this optical element, alongside the size of the X-ray imaging detector, is typically fixed for any given system, therefore defining the usable imaging field-of-view (FOV) To overcome these limitations, it has become relatively routine at synchrotron sources to carry out offset scans In the case of a parallel synchrotron beam, opposing projections can be stitched together to create full-field equivalents, a method not consistent with the divergent geometry observed with most conventional sources Alternatively, a simple linear weighting can be applied to overlap areas and reconstruction carried out across 360° Offset scans have been demonstrated using grating interferometry with a synchrotron source
31: where extended volumes were successfully reconstructed from truncated differential phase-contrast projections The same principles have also been demonstrated using cone-beam X-ray sources with an offset detector
32, 33, 34: By maintaining alignment between the X-ray source, centre-of-rotation (COR), and the perpendicular yet offset detector, a simple cone-beam geometry is maintained This allows a straightforward processing by redundancy weighting of projections
32: , followed by reconstruction using the Feldkamp Davis-Kress (FDK) algorithm
35: or iterative methods Despite its wide adoption, such a geometry suffers a decrease in flux efficiency Up to 50% of the beam is lost as a direct result of offsetting the detector Furthermore, for sources with fixed divergence, the minimum system length must be extended, to ensure that X-rays are incident upon the full extent of the offset detector Alternatively, it is possible to laterally displace the COR with respect to the stationary source and detector This way, there is no requirement to extend the system length and the full X-ray cone can be still captured by the detector To overcome the loss of the conventional cone-beam geometry, acquired data may be rebinned onto a virtual detector, to approximate the offset-detector case
36: The rebinning process however increases computational complexity and may be a source of artefacts and inaccuracies Instead, specific redundancy weights for the offset-COR case have also been derived and demonstrated using a flexible clinical gantry system
37, 38: Page 4 --- Consider a cone-beam beam geometry with its origin at the centre-of-rotation COR, as illustrated in figure 2 In the sample reference frame, the circular tomographic trajectory with projection angles α is described by the relative positions of the source and (centre of) detector, indicated by S(x, y, z) and Dcen(x, y, z), respectively For brevity, the coordinate dependence (x, y, z) is henceforth omitted.
Kai: To define these positions we use the opposing unit vectors Vs =
sin(α), 0, −cos(α): , and Vd =
−sin(α), 0, cos(α): where α is the angle of a given projection We thus can write the source and detector positions as S = (zsm + zmo)Vs + (∆COR)U, and Dcen = (zod)Vd + (∆COR)U where ∆COR is the offset of the COR along x, zsm is the source-to-mask distance, zmo is the mask-to-object distance, and U =
cos(α), 0, sin(α): is the unit vector describing the detector pixel orientation Only the region enclosed by the circle C1 remains within the FOV for the duration of a 360° rotation of the sample about COR Meanwhile, regions enclosed by the circle C2 but not by C1 remain within frame for some subset of the <360° rotation and thus must be weighted differently during reconstruction A redundancy weighting must be applied along x to the acquired projections, to correct for the increased density of rays counted within C1 compared to C2 For offset detector geometries, with COR offset angle τ = tan−1(∆COR/(zsm + zmo)), this must be adjusted to a symmetry in angle β about DCOR, W(DCOR + n) + W(DCOR − n) = 2 The function should furthermore be absent of discontinuities and offer a smooth transition between the maximum and minimum values We fulfill these criteria by adopting a similar smooth weighting function as previously used for complementary offsetCOR short scans
37: but instead we apply this principle to a full 360° rotation We describe this as a function of the detector position x as W(x) = 0, if x Dend, (11) where Ncols is the number of detector columns, and R = zsm + zmo + zod.
Mira: The size of the redundant region is given by D0Dend = Ncols/2 + R tan(tan−1(Ncols/(2R)) − 2τ) Page 5 --- The accuracy of the weighting and reconstruction method was first investigated via a numerical simulation study The central detector row corresponding to an offset geometry such as that shown in figure 2 was defined using a series of vectors, and a truncated sinogram simulated by forward projection of the SheppLogan phantom using the projectors of the Astra Toolbox
55: The geometry parameters were chosen to model the demonstrated experimental system A total of 2701 projections across 360° were simulated, upon a detector of 2424 pixel width Simple X-ray line integrals were simulated without considering phase effects Simulated sinograms were then reconstructed using the described redundancy weighting and the vector cone-beam geometry in the Astra toolbox Page 6 --- The results of the simulation study are illustrated in figure 3 Reconstructions of the large detector conventional scan (3a), offset COR with no weighting (3b), and offset COR with redundancy weighting (3c) are displayed on the top row The corresponding difference images between the reconstructions and ground-truth are displayed on the bottom row While the offset COR with no weighting does reconstruct the approximate morphology of the sample, the unweighted redundant rays produce significant artefacts at the centre of the phantom Furthermore, there is a large loss in quantitativeness, particularly visible in the profiles of figure 3g The offset COR scan with redundancy weighting accurately reconstructs the phantom, with the line profile matching the ground truth, and with a much reduced MSE compared to the no weighting case Page 7 --- Figure 4 illustrates the reconstruction of the beam-tracking scan using the proposed method Axial slices of the attenuation (4a), phase (4b), and dark-field (4c) contrast channels are shown, alongside corresponding zooms (4e, 4f, 6g) Volumetric information allows the tracing of pulmonary vasculature, with both venous (red arrows) and arterial (blue arrows) systems labelled in figure 4b Fine, low-contrast details are made more easily visible in the phase channel In particular, a network of fine vasculature can be seen, which can be followed back through the volume to the vessels labelled in figure 4b A further zoom and line profile (4h) demonstrates an arterial lumen with a full-width-at-half-maximum (consisting of 45 µm) of both the lumen width and the system spatial resolution) of 45 µm A number of enlarged air spaces indicative of centrilobular (Ce) and paraseptal (Pa) emphysema are visible in the slices, and particularly also in the volume rendering (4d) Page 8 --- Figure 5 illustrates the reconstruction of the high-resolution scan using the proposed method The axial slice in figure 5a shows how at higher resolution, the collapsed and closely spaced alveolar septa visible in figure 4 are now better defined The maximum intensity projection in figure 5b highlights brighter structures including calcifications (green arrows) and a number of small vessels (red arrows) with a diameter of ∼ 8 µm, which could represent capillaries, and which are smaller than the beam-tracking voxel size of 10.5 µm Page 9 --- Figure 6: A diagram demonstrating the definitions of the offset tomography geometry.
Title and authors: Kai: They are essentially describing this efficient method to double the maximum field of view of a cone-beam X-ray microtomography system while keeping the spatial resolution exactly as it was before. They also include multi-contrast capabilities in this setup.
Mira: The core idea is that they use an offset COR geometry with a full three hundred sixty degree scan, and then they use offset-COR redundancy weighting for reconstruction, which allows them to achieve up to two times the field of view without sacrificing pixel size or spatial resolution, according to the paper.
Lev: The methodology seems flexible because it can be applied across a range of different length scales and geometries, which is important because we need techniques that work in various lab settings. They also mention direct compatibility with complementary implementations of X-ray phase-contrast imaging.
Kai: They show they tested this on a section of resected human lung tissue, specifically about four by four by two cubic centimeters. The resulting horizontal field of view increased by a factor of one point seven times, reaching four point three centimeters, which was just enough to cover the entire sample.
Mira: And what’s really interesting is that they also applied these same concepts to free-space propagation imaging for a segment of two point seven millimeters from the same sample, achieving cubic voxels of four hundred fifty nanometers linear dimensions.
Lev: That scaling down to nanoscale voxel sizes in the free-space case shows the versatility of their geometric approach, moving between different scales in one framework. It’s a lot to show you can do with this geometry.
The paper's summary: Kai: One major improvement they highlight is that the redundancy weighting procedure allows for accurate volume reconstruction without needing to use iterative methods, which is a big plus for speed and reliability. They found it’s straightforward to adapt this approach to other lab-based X-ray phase-contrast imaging systems, such as those based on free-space propagation.
Mira: Another improvement they focus on is the ability to tailor the field of view specifically to the particular sample you are imaging without sacrificing spatial resolution, which is something that would be really beneficial in scenarios where rapid reconstructions are required. It’s about getting a good overview scan and then using a high-resolution scan to zoom in on different length scales.
Lev: They also mention that this technique is inherently compatible and stackable with other field of view extension methods, like tiled gratings, which means it can be integrated into existing systems without needing a complete overhaul.
Kai: The paper shows they can achieve quantitative separation of the attenuation, refraction, and dark-field contrast channels in a single shot using beam-tracking. This is pretty neat because you get three different information channels from one image acquisition.
Mira: It also shows how they can use beam-tracking to trace pulmonary vasculature by labeling both venous and arterial systems in their reconstructions, which helps visualize fine, low-contrast details more easily in the phase channel.
The paper's improvements: Kai: So to wrap up this paper on "Offset geometry for extended field-of-view in multi-contrast and multi-scale X-ray microtomography of lung cancer lobectomy specimens", they’ve demonstrated a method that doubles the maximum field of view without losing spatial resolution, and they used redundancy weighting to get accurate volumes without iterative methods.
Mira: The implication is that this geometry can be adapted to various existing XPCI systems, and the ability to tailor the field of view for different samples offers a new way to get detailed three dee information efficiently <ref:2502.10322#pg1>. They showed that it works across different length scales and is compatible with other contrast mechanisms.
Lev: For real hardware, running this without iterative methods is a huge advantage for speed, but I think we need to be careful about the accuracy of that weighting function when you move from simulation to actual measurements.
Kai: Right, Lev brings up the practical side—the fidelity of that weighting function under real experimental conditions. We've seen how they achieved those ten point five micrometer cubic voxels across four point three centimeters <ref:2502.10322#pg1>.
Mira: It’s a lot of data to manage, but the ability to get these accurate reconstructions suggests this approach is a solid tool for moving beyond the limitations of fixed detector elements in microtomography.
Lev: Yeah, so the main thing we see here is that it provides a way to get more information from our existing setups without needing massive hardware upgrades just to get that wider view.
Kai: And it opens up possibilities for applying these principles to other soft-tissue imaging problems where contrast enhancement is key. That’s what this work points toward for the next steps in microtomography research.
Conclusion: Kai: So we just covered how they managed to double the field of view on cone beam X-ray microtomography for lung cancer specimens without losing any spatial resolution, using an offset COR geometry and a redundancy weighting procedure.
Mira: Exactly. They showed that this technique works across different length scales, from large samples down to nanoscale segments in free space imaging, and it even works with different contrast mechanisms.
Lev: From my side, the real challenge is making sure that redundancy weighting translates well when you move from simulation to actual hardware runs on a real system.
Kai: Right, Lev brings up the practical side—the fidelity of that weighting function under real experimental conditions.
Mira: But they did manage to reconstruct accurate volumes without needing those computationally heavy iterative methods, which is a big win for speed.
Lev: Yeah, so the main thing we see here is that it provides a way to get more information from our existing setups without needing massive hardware upgrades just to get that wider view.
Kai: It’s interesting how they tied beam-tracking into this, allowing them to quantitatively separate the attenuation and phase contrast channels in one shot.
Mira: That quantitative separation is powerful because it lets you see the pulmonary vasculature, separating the veins from the arteries clearly in the volume rendering.
Lev: I'm impressed that they achieved those fine details, like those forty-five micrometer arterial lumens, using this method instead of something more complex.
Kai: So to wrap up, this paper "Offset geometry for extended field-of-view in multi-contrast and multi-scale X-ray microtomography of lung cancer lobectomy specimens" shows a solid way to push the limits of what we can see in these samples.
Mira: It’s a very flexible approach, and the compatibility with free space propagation imaging is something that opens up a lot of new avenues for lab-based XPCI systems.
Lev: It's definitely relevant because it shows how geometric solutions can complement contrast enhancement methods without needing completely different setups.
Kai: We'll be looking at how this kind of geometry could apply to other soft tissue imaging problems next.
Harry Allan, Adam Doherty, Carlos Navarrete Leon, Oriol Roche i Morgo, Yunpeng Jia, Charlotte Percival, Zoe Hagel, Kate E J Otter, Chuen Ryan Khaw, Êdeteght Gowers, Älen Hall Sam M Janes Fleur Monk David Moore Joseph Jacob Marco Endrizzi
Department of Medical Physics and Biomedical Engineering, University College London · X-ray Microscopy and Tomography Laboratory, The Francis Crick Institute · UCL Respiratory, University College London · Department of Cellular Pathology, University College London Hospitals NHS Foundation Trust · CRUK Lung Cancer Centre of Excellence, UCL Cancer Institute · Centre for Medical Image Computing, University College London
physics.med-ph, physics.ins-det
Submitted: 2025-02-14
Updated: 2025-02-14
Comments: 14 pages (inc. references), 5 figures
Journal ref: Med. Phys. 53, e70335 (2026)
DOI: 10.1002/mp.70335
Code: https://github.com/Hallan99/offset-cone-beam-CT-reconstruction
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 82/100
The gist: X-ray microtomography is a powerful non-destructive technique allowing 3D virtual histology of resected human tissue The authors describe an efficient method to double the maximum field-of-view of a…
Key concepts
- X-ray Phase-Contrast Imaging (XPCI)
- This technique improves contrast in low-density materials like lung tissue by measuring phase shifts imparted on the X-ray beam. This is useful because conventional XCT struggles with soft tissues due to low X-ray attenuation, making phase information a key tool for better visualization.
- Offset COR Geometry
- This geometry involves offsetting the center of rotation (COR) relative to the source and detector. This technique allows for a wider field-of-view—up to 2x the conventional size—while maintaining the same pixel size and spatial resolution, which is crucial for imaging larger samples.
- Redundancy Weighting
- This is a mathematical procedure applied during reconstruction that corrects inaccuracies introduced by using an offset detector. It ensures that accurate volumes are reconstructed from projections without needing complex, computationally expensive iterative algorithms or rebinning.
Terminology
Summary
X-ray microtomography is a powerful non-destructive technique allowing 3D virtual histology of resected human tissue The authors describe an efficient method to double the maximum field-of-view of a cone-beam X-ray microtomography system, without sacrificing on spatial resolution, and including multi-contrast capabilities They demonstrate an experimental realisation of the method, achieving exemplary reconstructions of a resected human lung sample, with a cubic voxel of 10. p. 0 We describe an efficient method to double the maximum fieldof-view of a cone-beam X-ray microtomography system, without sacrificing on spatial resolution, and including
multi-contrast capabilities. We demonstrate an experimental realisation of the method, achieving exemplary
reconstructions of a resected human lung sample, with a cubic voxel of 10. We demonstrate an experimental realisation of the method, achieving exemplary
reconstructions of a resected human lung sample, with a cubic voxel of 10. 5 µm linear dimensions, across
a horizontal field-of-view of 4. We demonstrate an experimental realisation of the method, achieving exemplary
reconstructions of a resected human lung sample, with a cubic voxel of 10. 5 µm linear dimensions, across
a horizontal field-of-view of 4.Where the paper says it
Where the paper says it
Where the paper says it
System Limitations and Contrast Enhancement
Conventional XCT suffers from inherently low X-ray attenuation cross section when applied to low-density materials such as lung tissue X-ray phase-contrast imaging (XPCI) is proposed as a solution to this problem because it can achieve improvements in contrast-to-noise ratio (CNR) compared to attenuation-based imaging Laboratory sources often impose limitations on the field-of-view due to optical elements, which can impose a further limitation on the field-of-view Techniques employing optical elements rely on either phase or attenuation effects to structure the beam in a way that creates phase sensitivity
Proposed Geometry and Reconstruction
The proposed method utilizes an offset COR geometry with a full 360° scan to increase the achievable FOV in imaging of lung lobectomy specimens This technique is demonstrated at multiple length scales, showing compatibility with multiple contrast-mechanisms The reconstruction is carried out using offset-COR redundancy weighting, followed by reconstruction without rebinning using a vector backprojection algorithm The offset geometry allows up to 2x the FOV to be imaged without sacrificing pixel size, spatial resolution, or X-ray flux density per detector element
Experimental Validation and Results
The system was utilized to image a section of resected human lung tissue of approximately 4 x 4 x 2 cm3 For the beam-tracking experiment, the sample was placed inside of a sealed polypropylene container The resulting horizontal FOV increased by a factor of 1.7x up to 4.3 cm, sufficient to cover the entire sample The reconstructed volumes allow for tracing pulmonary vasculature, with both venous (red arrows) and arterial (blue arrows) systems labelled in figure 4b
Conclusion and Future Applications
The redundancy weighting procedure allows the reconstruction of accurate volumes without the necessity to use iterative methods This approach is straightforward to adapt to other lab-based X-ray phase-contrast imaging systems, such as those based on free-space propagation The ability to tailor FOV to the particular sample, without sacrificing spatial resolution, would be highly beneficial in such scenarios where rapid reconstructions are required This method is inherently compatible and stackable with FOV-extension methods such as tiled-gratings The technique can be easily adapted to a wide range of existing XPCI and conventional systems
--- Page 1 ---
Offset geometry for extended field-of-view in multi-contrast and multi-scale X-ray microtomography of lung cancer lobectomy specimens
Harry Allan1,2, Adam Doherty1,2, Carlos Navarrete-Leon1,2, Oriol Roche i Morgo1,27 Yunpeng Jia1,28 Charlotte Percival3 Zoe Hagel3 Kate E J Otter3 Chuen Ryan Khaw3 Kate Gowers3 Helen Hall39 Sam M Janes3 Fleur Monk4 David Moore45 Joseph Jacob6 Marco Endrizzi1,2 1Department of Medical Physics and Biomedical Engineering, University College London, London, UK 2X-ray Microscopy and Tomography Laboratory, The Francis Crick Institute, London, UK 3UCL Respiratory, University College London, London, UK 4Department of Cellular Pathology, University College London Hospitals NHS Foundation Trust, London,UK 5CRUK Lung Cancer Centre of Excellence UCL Cancer Institute University College London,London,UK 6Centre for Medical Image Computing University College London,London,UK
Abstract
X-ray microtomography is a powerful non-destructive technique allowing 3D virtual histology of resected human tissue The achievable imaging field-of-view, is however limited by the fixed number of detector elements, enforcing the requirement to sacrifice spatial resolution in order to image larger samples In applications such as soft-tissue imaging, phase-contrast methods are often employed to enhance image contrast Some of these methods, especially those suited to laboratory sources, rely on optical elements, the dimensions of which can impose a further limitation on the field-of-view We describe an efficient method to double the maximum fieldof-view of a cone-beam X-ray microtomography system, without sacrificing on spatial resolution, and including multi-contrast capabilities We demonstrate an experimental realisation of the method, achieving exemplary reconstructions of a resected human lung sample, with a cubic voxel of 10.5 µm linear dimensions, across a horizontal field-of-view of 4.3 cm The same concepts are applied to free-space propagation imaging of a 2. p. 0 5 µm linear dimensions, across
a horizontal field-of-view of 4. The same concepts are applied to free-space propagation imaging of a
2.Where the paper says it
1 Introduction
X-ray computed tomography (XCT), has proven to be an indispensable tool for the non-destructive 3D inspection of samples in a range of fields including medical sciences, materials engineering, and metrology [1] X-ray microtomography (micro-CT) [2] allows the probing of structures on the micro-scale Access to 3D morphology has allowed detailed virtual histology of diseased human lung structure in cases such as fibrosis [3], Covid-19 [4], and adenocarcinomas [5] Volumetric information from micro-CT enables 3D quantification of disease biomarkers [6], which can be demonstrably more sensitive than their 2D counterparts [7] enhancing the understanding of disease and the effects of treatment Conventional XCT reconstructs 3D distributions from many 2D projections formed predominantly by the attenuation of X-rays by the object in the beam Yet despite its widespread adoption, the application of XCT to low-density materials, such as lung-tissue, still suffers from their inherently low X-ray attenuation cross section X-ray phase-contrast imaging (XPCI) [8] is capable of achieving improvements in contrast-to-noise ratio (CNR) compared to attenuation-based imaging, and has thus been proposed as a solution to this problem XPCI forms images based on phase-shifts imparted on the X-ray beam by the object, typically requiring beams with high levels of spatial and temporal coherence, such as those found at synchrotron radiation facilities [9, 10] Several techniques have been developed to enable the use of XPCI techniques with conventional sources, increasing their availability in laboratory and clinical settings XPCI based only on free-space propagation (FSP) of the perturbed wavefront is achievable using laboratory sources [11] however this comes at the cost of stringent requirements on source and detector characteristics A number of other methods employ optical elements to alleviate some of these difficulties, including grating interferometry [12], edge-illumination [13], beam-tracking [14], and speckle-tracking [15, 16] Each of these methods is able to recover independent attenuation and differential-phase signals, caused respectively by the imaginary (β) and real (δ) components of the sample’s complex refractive index Laboratory-based XPCI micro-CT has been utilised in a range of applications via a number of methods
--- Page 2 ---
(FSP) of the perturbed wavefront is achievable using laboratory sources [11] however this comes at the cost of stringent requirements on source and detector characteristics A number of other methods employ optical elements to alleviate some of these difficulties, including grating interferometry [12], edge-illumination [13], beam-tracking [14], and speckle-tracking [15, 16] Each of these methods is able to recover independent attenuation and differential-phase signals, caused respectively by the imaginary (β) and real (δ) components of the sample’s complex refractive index Alongside these quantities, they are also sensitive to the unresolved refraction and ultra-small angle X-ray scattering (USAXS) that leads to the dark-field contrast channel Laboratory-based XPCI micro-CT has been utilised in a range of applications via a number of methods
--- Page 3 ---
Techniques employing optical elements rely on either phase or attenuation effects to structure the beam in a way that creates phase sensitivity The size of this optical element, alongside the size of the X-ray imaging detector, is typically fixed for any given system, therefore defining the usable imaging field-of-view (FOV) To overcome these limitations, it has become relatively routine at synchrotron sources to carry out offset scans In the case of a parallel synchrotron beam, opposing projections can be stitched together to create full-field equivalents, a method not consistent with the divergent geometry observed with most conventional sources Alternatively, a simple linear weighting can be applied to overlap areas and reconstruction carried out across 360° Offset scans have been demonstrated using grating interferometry with a synchrotron source [31] where extended volumes were successfully reconstructed from truncated differential phase-contrast projections The same principles have also been demonstrated using cone-beam X-ray sources with an offset detector [32, 33, 34] By maintaining alignment between the X-ray source, centre-of-rotation (COR), and the perpendicular yet offset detector, a simple cone-beam geometry is maintained This allows a straightforward processing by redundancy weighting of projections [32], followed by reconstruction using the Feldkamp Davis-Kress (FDK) algorithm [35] or iterative methods Despite its wide adoption, such a geometry suffers a decrease in flux efficiency Up to 50% of the beam is lost as a direct result of offsetting the detector Furthermore, for sources with fixed divergence, the minimum system length must be extended, to ensure that X-rays are incident upon the full extent of the offset detector Alternatively, it is possible to laterally displace the COR with respect to the stationary source and detector This way, there is no requirement to extend the system length and the full X-ray cone can be still captured by the detector To overcome the loss of the conventional cone-beam geometry, acquired data may be rebinned onto a virtual detector, to approximate the offset-detector case [36] The rebinning process however increases computational complexity and may be a source of artefacts and inaccuracies Instead, specific redundancy weights for the offset-COR case have also been derived and demonstrated using a flexible clinical gantry system [37, 38]
--- Page 4 ---
Consider a cone-beam beam geometry with its origin at the centre-of-rotation COR, as illustrated in figure 2 In the sample reference frame, the circular tomographic trajectory with projection angles α is described by the relative positions of the source and (centre of) detector, indicated by S(x, y, z) and Dcen(x, y, z), respectively For brevity, the coordinate dependence (x, y, z) is henceforth omitted. To define these positions we use the opposing unit vectors Vs = [sin(α), 0, −cos(α)], and Vd = [−sin(α), 0, cos(α)] where α is the angle of a given projection We thus can write the source and detector positions as S = (zsm + zmo)Vs + (∆COR)U, and Dcen = (zod)Vd + (∆COR)U where ∆COR is the offset of the COR along x, zsm is the source-to-mask distance, zmo is the mask-to-object distance, and U = [cos(α), 0, sin(α)] is the unit vector describing the detector pixel orientation Only the region enclosed by the circle C1 remains within the FOV for the duration of a 360° rotation of the sample about COR Meanwhile, regions enclosed by the circle C2 but not by C1 remain within frame for some subset of the <360° rotation and thus must be weighted differently during reconstruction A redundancy weighting must be applied along x to the acquired projections, to correct for the increased density of rays counted within C1 compared to C2 For offset detector geometries, with COR offset angle τ = tan−1(∆COR/(zsm + zmo)), this must be adjusted to a symmetry in angle β about DCOR, W(DCOR + n) + W(DCOR − n) = 2 The function should furthermore be absent of discontinuities and offer a smooth transition between the maximum and minimum values We fulfill these criteria by adopting a similar smooth weighting function as previously used for complementary offsetCOR short scans [37] but instead we apply this principle to a full 360° rotation We describe this as a function of the detector position x as W(x) = 0, if x Dend, (11) where Ncols is the number of detector columns, and R = zsm + zmo + zod. p. 3 We thus can write the source and
detector positions as
S = (zsm + zmo)Vs + (∆COR)U, (9)
Dcen = (zod)Vd + (∆COR)U, (10)
where ∆COR is the offset of the COR along x, zsm is the source-to-mask distance, zmo is the mask-to-object
distance, and U = [cos(α), 0, sin(α)] is the unit vector describing the detector pixel orientation, perpendicular
to the line formed by SDcen. From figure 2 it can be seen that only the region enclosed by the circle C1
remains within the FOV for the duration of a 360° rotation of the sample about COR.Where the paper says it
--- Page 5 ---
The accuracy of the weighting and reconstruction method was first investigated via a numerical simulation study The central detector row corresponding to an offset geometry such as that shown in figure 2 was defined using a series of vectors, and a truncated sinogram simulated by forward projection of the SheppLogan phantom using the projectors of the Astra Toolbox [55] The geometry parameters were chosen to model the demonstrated experimental system A total of 2701 projections across 360° were simulated, upon a detector of 2424 pixel width Simple X-ray line integrals were simulated without considering phase effects Simulated sinograms were then reconstructed using the described redundancy weighting and the vector cone-beam geometry in the Astra toolbox
--- Page 6 ---
The results of the simulation study are illustrated in figure 3 Reconstructions of the large detector conventional scan (3a), offset COR with no weighting (3b), and offset COR with redundancy weighting (3c) are displayed on the top row The corresponding difference images between the reconstructions and ground-truth are displayed on the bottom row While the offset COR with no weighting does reconstruct the approximate morphology of the sample, the unweighted redundant rays produce significant artefacts at the centre of the phantom Furthermore, there is a large loss in quantitativeness, particularly visible in the profiles of figure 3g The offset COR scan with redundancy weighting accurately reconstructs the phantom, with the line profile matching the ground truth, and with a much reduced MSE compared to the no weighting case
--- Page 7 ---
Figure 4 illustrates the reconstruction of the beam-tracking scan using the proposed method Axial slices of the attenuation (4a), phase (4b), and dark-field (4c) contrast channels are shown, alongside corresponding zooms (4e, 4f, 6g) Volumetric information allows the tracing of pulmonary vasculature, with both venous (red arrows) and arterial (blue arrows) systems labelled in figure 4b Fine, low-contrast details are made more easily visible in the phase channel In particular, a network of fine vasculature can be seen, which can be followed back through the volume to the vessels labelled in figure 4b A further zoom and line profile (4h) demonstrates an arterial lumen with a full-width-at-half-maximum (consisting of 45 µm) of both the lumen width and the system spatial resolution) of 45 µm A number of enlarged air spaces indicative of centrilobular (Ce) and paraseptal (Pa) emphysema are visible in the slices, and particularly also in the volume rendering (4d)
--- Page 8 ---
Figure 5 illustrates the reconstruction of the high-resolution scan using the proposed method The axial slice in figure 5a shows how at higher resolution, the collapsed and closely spaced alveolar septa visible in figure 4 are now better defined The maximum intensity projection in figure 5b highlights brighter structures including calcifications (green arrows) and a number of small vessels (r…
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Improved AI system can perform 3D virtual histology of resected human tissue with an achievable horizontal field-of-view up to 4.3 cm without sacrificing spatial resolution, as demonstrated by achieving
exemplary reconstructions of a resected human lung sample, with a cubic voxel of 10.5 µm linear dimensions
across this field-of-view. -
Improved AI system can reconstruct accurate volumes using an offset COR geometry and redundancy weighting, which the paper shows
can produce comparable reconstructions to a scan acquired with a detector twice as large,
eliminating the need for computationally expensive iterative methods. -
Improved AI system can quantitatively separate attenuation, refraction, and dark-field contrast channels in a single shot using beam-tracking, allowing for the
quantitative separation of each contrast channel
from a single image. -
Improved AI system can enable selective zooming of areas chosen based on a large field-of-view overview scan and high-resolution scan, allowing investigation of different length scales using the same instrument.
-
Improved AI system can be adapted to a range of existing XPCI systems, including those based on free-space propagation, by applying the offset geometry and redundancy weighting principles without requiring specialized adaptations.
Abstract
X-ray microtomography is a powerful non-destructive technique allowing 3D virtual histology of resected human tissue. The achievable imaging field-of-view, is however limited by the fixed number of detector elements, enforcing the requirement to sacrifice spatial resolution in order to image larger samples. In applications such as soft-tissue imaging, phase-contrast methods are often employed to enhance image contrast. Some of these methods, especially those suited to laboratory sources, rely on optical elements, the dimensions of which can impose a further limitation on the field-of-view. We describe an efficient method to double the maximum field-of-view of a cone-beam X-ray microtomography system, without sacrificing on spatial resolution, and including multi-contrast capabilities. We demonstrate an experimental realisation of the method, achieving exemplary reconstructions of a resected human lung sample, with a cubic voxel of 10.5 μ m linear dimensions, across a horizontal field-of-view of 4.3 cm. The same concepts are applied to free-space propagation imaging of a 2.7 mm segment of the same sample, achieving a cubic voxel of 450 nm linear dimensions. We show that the methodology can be applied at a range of different length-scales and geometries, and that it is directly compatible with complementary implementations of X-ray phase-contrast imaging.
Sources
Related papers
- BART Online Open-Source Sequence Toolbox for Computational MRI
- Constrained estimation of rotational invariants of the cumulant expansion (RICE) for rapid tensor-valued diffusion MRI
- Improved Myelin Water Fraction Estimation and Uncertainty Mapping in Multicomponent T2 Relaxometry via NNLS Bootstrap Aggregation
- Radiation damage to normal mammalian tissue in vivo with laser-driven protons at ultra-high instantaneous dose rate
- BEAM3R: Beam's-eye-view architecture with Mamba-3 for implicit dose reconstruction
- MORSE-PI -- Flexible and artefact-free image reconstruction for structural and functional QSM and other phase-critical imaging applications