Precision Matrix on SOMATOM Force - Ease your workflow in high-resolution CT imaging

White Paper Precision Matrix on SOMATOM Force Ease your workflow in high-resolution CT imaging Marcus Brehm, PhD siemens-healthineers.com/somatom-force International version. Not for distribution or use in the U.S. This product is pending 510(k) clearance, and is not yet commercially available in the United States. SIEMENS Healthineers Precision Matrix White Paper Contents Introduction to high-resolution imaging 3 Role of matrix size in high-resolution imaging 4 Why a larger matrix size is not always the best choice 6 Which matrix sizes are available with Precision Matrix and how to select them? 7 And what if I need support to select the right matrix size? 8 Additional factors to be considered in high-resolution imaging 11 Clinical cases 12 Combine forces 14 2 Precision Matrix White Paper Introduction to high-resolution imaging In 50 years of technical development, computed sampling rate in addition to the quarter detector offset. tomography (CT) set new standards in diagnostic On another front, 3D anti-scatter grids replaced 2D versions. imaging. Fast visualization of the finest anatomical Detector pixel size decreased and channel density structures is, for example, one of the strengths of CT. increased [12]. An attenuating comb filter can be optionally And there are many clinical applications in diagnostic used in front of the detector to reduce the detector aperture, imaging that demand a strong level of high-contrast i.e., the effective detector pixel size [13]. spatial resolution in the sub-mm range. Typical examples are images of the lung, temporal bone, sinus, wrist, Apart from the excellent high-contrast spatial resolution and ankle [1]‒[6]. This need is covered by high- capabilities achieved through technical progress, today’s resolution CT (HRCT) [7] or nowadays even by ultra- scanners also have to meet other demanding challenges high-resolution CT (UHRCT) [4]. in high-resolution imaging – challenges like realizing the lowest possible radiation dose level, providing high Technological advances in the X-ray tube and X-ray flexibility, offering high versatility, and ensuring ease of detector, the main pillars of a CT system, increased the use. In particular, the clinical workflow is increasingly level of high-contrast spatial resolution step by step. important in the daily routine. This is where Precision Focal spot sizes were reduced down to very small sizes Matrix comes into play. But what is Precision Matrix and of 0.6 x 0.7 (IEC) for HRCT and 0.4 x 0.5 (IEC) for UHRCT how does it support clinicians and radiographers in [8]‒[9]. Various vendors introduced X-ray tubes with high-resolution imaging with cutting-edge technology? focal spot deflection [10]‒[11], which can improve the 3 Precision Matrix White Paper Role of matrix size in high-resolution imaging High-contrast spatial resolution depends on various What happens when the three identified parameters – settings in CT imaging. Whereas built-in hardware reconstruction kernel, FoV, and reconstruction matrix – components predefine the limits, the actual resolution do not match? Three experiments based on a line pair within CT images is controlled via parameters freely high-resolution phantom (test module CTP528 of selectable by the user. Use-related trade-off between Catphan 600, The Phantom Laboratory, Salem, NY, USA) spatial resolution and image noise level is an essential demonstrate the interaction between parameters and feature because requirements highly depend on the the perceived spatial resolution. clinical case. Two CT images are compared to each other in Experiment 1 Control is mainly driven by selection of the reconstruction (see Fig. 1); they differ in matrix size (512 vs. 1024) but kernel. A huge variety of options to choose from is share the same choice of reconstruction kernel (Br64) available: examples include bone kernels for fine and FoV (300 mm). Why does one image show greater high-contrast details, smooth kernels for low-contrast detail than the other? The answer is that the maximum objects, as well as dedicated kernels for certain clinical resolution a 512 matrix can depict here is only 8.5 lp/cm. applications (e.g., quantitative imaging). In addition to This is not sufficient compared to the 10.1 lp/cm at kernel selection, the size of the volumetric image pixel 50% resolving power (MTF) of the applied kernel. (voxel) plays an important role for image resolution. The Structural details get lost as a result of the unmatched in-plane size is determined by the displayed field of view parameters, details that are supported by the imaging (FoV) and the size of the reconstruction matrix. The hardware and selected reconstruction kernel. reconstruction matrix is a two-dimensional matrix with a certain number of rows and columns. Typically, it is a square matrix with an equal number of rows and columns. In most CT scanners and clinical applications, the reconstruction matrix is fixed to 512 x 512 voxels (“512”). Sometimes smaller or larger matrices are available, like 1024 x 1024 voxels (“1024”) or 768 x 768 voxels (“768”). 512 matrix 1024 matrix ..... Fig. 1: Section from line pair high-contrast resolution phantom for kernel Br64 (50% MTF: 10.1 lp/cm), FoV of 300 mm, and different matrix sizes. Voxel size for 512 matrix is d = 300 mm / 512 = 0.59 mm and maximum spatial resolution that can be displayed is fmax = 10 mm/cm / (2/lp * d) = 8.5 lp/cm. For the 1024 matrix, fmax is 17 lp/cm. 4 Precision Matrix White Paper 512 matrix 1024 matrix Fig. 2: Section from line pair high-contrast resolution phantom for kernel Br40 (50% MTF: 4.0 lp/cm), FoV of 300 mm, and different matrix sizes. A similar setup is used in Experiment 2 (see Fig. 2), where What are the lessons learned from these three experiments? both CT images are again based on joint kernel selection, An unmatched parameter selection can result in a loss however a smoother version (Br40) is chosen. Here, the of spatial resolution. In return, a larger matrix size may two reconstructions do not show any noticeable difference. improve the resolution of CT images, but not in every The reason is the limited resolution of the kernel with case. One can alternatively adjust the FoV in order only 4.0 lp/cm at 50% MTF. This can already be properly to achieve the same result, but larger matrix sizes enable represented by a 512 matrix. coverage of larger FoVs at the same image quality. The respective workaround in high-resolution imaging In Experiment 3, two image reconstructions are compared is thus no longer needed, i.e., adding further image to each other that again share the same sharp reconstruction reconstructions with dedicated smaller FoVs. But at the kernel (Br64) but now also have the same voxel size. same time one has to understand that system hardware Therefore individual FoVs are used: 300 mm @ 1024 and remains the limiting factor of maximum resolution in 150 mm @ 512. No differences are evident between CT imaging and this cannot be improved by introducing both images within the overlapping region (see Fig. 3). larger matrix sizes. But here the 1024 matrix covers an area four times larger than the 512 matrix. 512 matrix 1024 matrix Fig. 3: Section from line pair high-contrast resolution phantom for kernel Br64 (50% MTF: 10.1 lp/cm), and same voxel size based on different pairs of FoV and matrix size (150 mm @ 512 vs. 300 mm @ 1024). 5 Precision Matrix White Paper Why a larger matrix size is not always the best choice Three experiments demonstrated that larger matrix sizes Last but not least, all applications that are applied for either outperform the standard 512 matrix in resolution postprocessing and viewing of reconstructed images have and coverage or at least maintain its level of quality. to support larger matrix sizes. One needs to doublecheck Given the potential benefits, why not replace the the compatibility of the affected applications in advance, standard matrix size in every case and perform all image before applying a 1024 matrix. reconstructions with the largest matrix size available? In a nutshell: Whereas the reconstruction kernel and FoV Where there is light, there is shadow, and the advantages have to be chosen based on clinical needs, the matrix mentioned are countered by certain potential drawbacks. size needs to be large enough to cover the spatial As an example, CT images increase in size by a factor of 4 resolution of the kernel and simultaneously be as small when using a 1024 matrix instead of the standard 512. as possible to limit demands on computational and Still, assuming a use of 1024 matrix in only 10% of cases, storage resources. In addition, the matrix size has to be 786 in 40% of cases, and the standard 512 otherwise, supported by the respective postprocessing and viewing the storage capacity requirements on PACS and acquisition applications. workplaces increase by 76% and double the image storage capacity required. Another related problem is the fact that the reconstruction effort increases up to a factor of 4 when using a 1024 matrix. The actual factor will however be lower because operations in the projection space are not affected by the matrix size. Nevertheless, a noticeable increase in reconstruction effort has to be expected and needs to be offset with commensurate computational power in order to limit the impact on clinical workflow. When a parameter selection induces a mismatch between FoV size and reconstruction kernel, a larger matrix size will increase the spatial resolution. But image noise level will increase simultaneously. In this case, the previous configuration should be checked. Was it intentional and, in fact, limited by standard matrix size? If not, it is obvious the reconstruction kernel can be adapted and does not need a larger matrix size with its previously mentioned drawbacks. 6 Precision Matrix White Paper Which matrix sizes are available with Precision Matrix and how to select them? With the introduction of Precision Matrix, the user can next to the previously available iBHC and iMAR options. now modify the matrix size to values beyond the standard Matrix sizes can be independently chosen for each 512. Possible choices are: 1024, 768, 512, and 256*. A single image reconstruction. This applies to axial as well respective combo box becomes part of the reconstruction as 3D reconstructions using the same configuration tab via “Advanced reconstruction options” (see Fig. 4), option. Smith, John Precision Matrix Total mAs: O Recon job 1 2 3 4 5 6 8 Series description Hip 0.6 Br59 3 Slice 0.6 mm ] Advanced reconstruction options ADMIRE Strength Matrix size 768 Kernel Br59 iBHC None Auto FAST Window Pelvis iMAR Hip implants 512 768 1024 HD FoV FoV 380 mm 2: Image order Craniocaudal Center X 0 mm 2 Increment 0.4 mm 2 Overview Center Y 0 mm :2 No. of images 251 Mirroring None Comments Extended CT scale Matrix size 768 Routine Scan Recon Auto Tasking Fig. 4: Matrix size accessible via advanced reconstruction options of reconstruction tab. For oblique reconstructions via 3D Recon, there is a It is possible to limit the selection by setting an upper second opportunity to alter matrix size: the 3D Graphical limit for matrix size within the Examination Configuration Reconstruction Planning (GRP) controls bar (see Fig. 5). “Workflow” subtask card. The matrix size can also be Here, all sizes are available as square as well as non- pre-configured within the Scan Protocol Assistant, just as square options. The largest aspect ratio of non-square users are accustomed to from all the other scan and matrixes is 1:8 for 256 and 1:4 for all other matrix sizes. reconstruction parameters. Thus, CT reconstructions with a matrix up to 1024 x 4096 are now possible. ELLLE Matrix Size 1024 Non-square Planning base Hip RTD Fig. 5: Matrix size accessible via 3D GRP as a square as well as non-square option. *Available in cardio protocols only 7 Precision Matrix White Paper And what if I need support to select the right matrix size? In addition to the manual selection of matrix sizes up to adapted to patient size in every single case, e.g., via FAST 3D, 1024, Precision Matrix also offers “Automode”. In order and different organ characteristics include optimized to activate this mode, the radiographer just has to select settings for imaging different body parts, to name just a “Auto”. But why does Precision Matrix include Automode few examples. Almost each individual clinical case has its and how does it work? own setting. And the matrix size depends on the above- mentioned parameters and more. The optimal choice There are so many degrees of freedom when it comes to may require verification for every patient. It will be reconstruction parameters. A large number of difficult for radiographers to choose the right matrix size reconstruction kernels are provided to cover the huge for every clinical question and patient as well as to variety of clinical tasks; the FoV can be automatically establish consistency among all staff members (see Fig 6). Parameters to be Potential pitfalls considered without Automode Scenario 1: matrix size too large • Scan protocol • • Field of view No additional image visualization • Reconstruction kernel improvement • Reconstruction effort too great • Reconstruction method • Matrix size Image size too big • Scenario 2: • Image size Automode off matrix size not large enough • Reconstruction effort • Potential image visualization improvement missed • Manual entry Potential additional reconstruction Correlated parameters needed (e.g., with a different FoV) Fig. 6: Parameters to be taken into account and potential pitfalls when matrix size is entered manually. For this reason, Precision Matrix offers Automode, which effort and image data size. In other words, it selects the takes the burden from the radiographer. This mode proper matrix size in order to allow for the requested considers different variables like sharpness of reconstruction spatial resolution at the lowest costs. By default, the list kernel, FoV size, reconstruction effort, amount of image of available matrix sizes includes the values 512, 768, data, and a list of available matrix sizes (see Fig. 7). and 1024. It is possible to limit the selection by setting Automode provides as output the best matrix size that an upper limit for matrix size within the Examination maintains the sharpness of the reconstruction kernel Configuration “Workflow” subtask card. This setting will within the final CT image while minimizing reconstruction be valid for both Automode and manual selection. 8 Precision Matrix White Paper The spatial properties of the reconstruction kernel, chosen optimal matrix size. Two examples are shown in Fig. 8, by the user, are converted into a minimum required which shows the threshold’s dependency on requested voxel size, potentially taking further parameters into account voxel size and the optimal matrix size for respective FoV like organ characteristics or reconstruction method. ranges. Depending on the requested minimum voxel size, one gets several thresholds. FoV sizes below such a threshold include voxel sizes smaller than or equal to the requested one. For sizes above the threshold, a larger matrix size is needed. The FoV size, selected by the user, is then compared with those thresholds in order to get the Parameters to be Benefits with considered Automode Scan protocol • Image visualization optimized • Field of view • Matrix size optimized • Reconstruction kernel • Reconstruction effort optimized • Reconstruction method • Image size optimized • • Number of reconstructions optimized Automode on Fig. 7: Automode considers several variables to select the proper configuration. 9 Precision Matrix White Paper Let’s take another look at the experiments from the same level of detail as 1024 matrix, but with a lower previous section. The standard 512 matrix was not computational and storage burden (see Fig. 9). In sufficient for the first experiment and image details that Experiment 2, Automode selects the standard 512 matrix were visible with 1024 matrix were lost. What happens because it is already sufficient for smooth kernels. when we apply Automode? The mode recognizes that the Different FoV sizes can lead to different matrix sizes, as standard matrix size is insufficient but selects 768 and not shown in Experiment 3. There, Automode selects 512 for 1024. The reason: 768 matrix is sufficient to cover the 150 mm FoV and 768 for 300 mm FoV. Voxel Size (mm) FoV (mm) 512 768 1024 0,9 0,8 0,7 0,6 0,5 0,4 Requested voxel size 0,3 0,2 0,1 Threshold 512→768 Threshold 768→1024 0,0 50 200 350 500 0,9 0,8 0,7 0,6 Requested voxel size 0,5 0,4 0,3 0,2 0,1 Threshold 512→768 0,0 50 200 350 500 Fig. 8: Optimal matrix size for each requested voxel size. 512 matrix 786 matrix 1024 matrix (Automode) Fig. 9: Matrix size of 786 is the right choice, which Automode made as well. 10 Precision Matrix White Paper Additional factors to be considered in high-resolution imaging Maximum resolution can only be achieved when everything structures compared to standard pitch breath-hold fits together, with or without Precision Matrix. Therefore, acquisition for detection of pulmonary embolism [14]. it is necessary to pay particular attention to acquisition settings for high-resolution imaging. Besides spatial High spatial resolution is always accompanied by a high resolution, temporal resolution is one of the main level of image noise that nonetheless should not contributors particularly in chest imaging. Limitations in affect diagnostic confidence. In addition to iterative temporal resolution can deteriorate image quality by reconstruction methods to reduce noise, one should artifacts induced through breathing or cardiac motion. always exploit the full potential on the acquisition side. Ultra-fast data acquisition is an essential technique to This includes patient size-dependent tube voltage reduce motion artifacts especially when patients are not selection such as provided by CARE kV and 10 kV Steps, able to hold their breath. Turbo Flash mode on Dual Source CT or spectra dedicated to high-contrast imaging such scanners, for example, increases diagnostic confidence as those available on SOMATOM® CT scanners with and improves assessability of vascular and bronchial Tin Filter. 11 Precision Matrix White Paper Clinical cases Precision Matrix with matrix sizes up to 1024 can improve A clinical example from chest imaging is shown in Fig. 10: clinical workflow where sharp kernels are required and a 42-year-old female with pulmonary fibrosis and a large FoV has to be covered. What are clinical questions emphysema in the lower lobes. Here, Turbo Flash mode and examples that combine these two requirements of acquisition was conducted with a total scan time of high spatial resolution in a large region of interest? 616 ms to avoid any motion artifacts. In addition, For bilateral hip replacements, it is important to cover Tin Filter was applied to achieve the low radiation dose the entire pelvis and to apply bone kernels, e.g., to level of 2.02 mGy. For this case, axial views that share all measure acetabular cup placement or assess osteolysis reconstruction parameters except the matrix size are when the possibility of revision arthroplasty needs to be compared to each other (see Fig. 11). The section evaluated. The same applies for acute care where covering the entire chest (top row) does not reveal any whole-body bone CT scans are conducted to detect missed difference. But the details in the zoomed version from bone injuries in polytrauma patients. Sharp kernels and the same image (bottom row) reveal clear differences a large FoV including the entire rib cage are used in visible in resolution. With the standard 512 matrix, an high-resolution chest imaging. These are just some additional reconstruction and dataset may be required examples, and they become even more challenging in with a dedicated smaller FoV to reacquire the level of obese patients. detail that the 1024 matrix already provides. SOMATOM Force Scan time: 616 ms Scan length: 315 mm Sn150 kV CTDIvol: 2.02 mGy DLP: 76.63 mGy cm Fig. 10: Acquisition parameters and coronal MPR in a sample chest imaging case. Courtesy of Hôpital Albert Calmette, Lille, France. Another example is shown in (Fig. 12). An 81-year-old visible at the edges of bony and metal structures due to with bilateral hip replacement; taken from a chest- the limited resolving power of the standard 512 matrix abdominal-pelvic scan (120 kV, scan time 4.9 s, scan for such a large FoV including both hip implants. The range 579 mm, CTDIvol 10.84 mGy, DLP 678.21 mGy cm). larger matrix size provided by Precision Matrix plus metal Here two kinds of artifacts are present in the default artifact reduction by iMAR can substantially improve reconstruction. Metal artifacts emerging from the hip image quality and allow evaluations on just a single implants deteriorate image quality and have to be reconstruction. Without Precision Matrix, dedicated FoVs addressed by a metal artifact reduction technique like are needed, e.g., one for each implant to get the same iMAR [15]. In addition, stair-step artifacts are clearly image quality. 12 Precision Matrix White Paper 512 matrix 1024 matrix 1 cm Fig. 11: Axial views for 512 matrix (left) and 1024 matrix (right). Reconstruction parameters: Bl64, ADMIRE level 3, and max. FoV. Courtesy of Hôpital Albert Calmette, Lille, France. Large matrix sizes provided by Precision Matrix can ease resulting FoV size. Also of interest are details that do clinical workflow and allow creating just a single dataset not change the diagnosis but increase confidence that covers the entire region of interest with the level through the clear presentation of the finest structures of detail needed independent of patient size and in every patient. Fig. 12: Coronal views for 512 matrix without iMAR (left) and 1024 matrix with iMAR (right). Reconstruction parameters: Br59, ADMIRE level 5, and 460 mm FoV. Courtesy of Hôpital Albert Calmette, Lille, France. 13 Precision Matrix White Paper Combine forces When it comes to high-resolution imaging, with Turbo Flash mode, exceptionally fast reconstruction SOMATOM® Force offers the latest technologies and with up to 70 images/s for iterative reconstruction at thus prominent characteristics: outstanding spatial standard matrix size, and personalized radiation dose resolution provided by VectronTM X-ray tubes with the reduction via Tin Filter or 10 kV Steps. This unique imaging smallest available focal spot size of 0.4 x 0.5 (IEC), chain enables sharp and contrast-rich images for every 3D anti-scatter grid, StellarInfinity detectors, unparalleled patient at high speed and low dose. The variety of and unbeaten native temporal resolution driven by Dual advanced technological features is now complemented Source technology and a rotation time of 250 ms, the by Precision Matrix to improve workflow performance fastest acquisition speed of up to 737 mm/s available and unlock the full power of SOMATOM® Force. SIEMENS . Healthineers SOMATOM Force 14 Precision Matrix White Paper [1] Lynch D. A, et al. High-resolution computed [12] Hata A, et al. Effect of matrix size on the tomography in idiopathic pulmonary fibrosis. image quality of ultra-high-resolution CT of the Am J Respir Crit Care Med. 2005; 172(4): 488‒493. lung: Comparison of 512 × 512, 1024 × 1024, [2] Meyer M, et al. Initial results of a new generation and 2048 × 2048. Acad Radiol. 2018; 25(7): dual source CT system using only an in-plane 869‒876. comb filter for ultra-high resolution temporal [13] Flohr T, Stiersdorfer K, Süss C, Schmidt B, bone imaging. Eur. Radiol. 2014; 25(1): 178‒185. Primak A. N, McCollough C. H. Novel [3] Kim C. R, Jeon J.Y. Radiation dose and ultrahigh resolution data acquisition and image conspicuity comparison between image reconstruction for multi-detector row conventional 120 kVp and 150 kVp with CT. Med Phys. 2007; 34(5): 1712‒1723. spectral beam shaping for temporal bone CT. [14] Martini K, Meier A, Higashigaito K, Eur J Radiol. 2018; 102: 68‒73. Saltybaeva N, Alkadhi H, Frauenfelder T. [4] Zhou W, et al. Comparison of a photon-counting- Prospective randomized comparison of detector CT with an energy-integrating-detector high-pitch CT at 80 kVp under free breathing CT for temporal bone imaging: A cadaveric with standard-pitch CT at 100 kVp under study. AJNR Am J Neuroradiol. 2018; 39(9): breath-hold for detection of pulmonary embolism. Acad Radiol. 2016; 23(11): 1733‒1738. 1335‒1341. [5] Wuest W, May M, Saake M, Brand M, Uder M, Lell M. Low-dose CT of the paranasal sinuses: [15] Kachelrieß M, Krauss A. Iterative metal Minimizing X-ray exposure with spectral artifact reduction (iMAR): Technical principles shaping. Eur Radiol. 2016; 26(11): 4155‒4161. and clinical results in radiation therapy. White Paper. Siemens Healthcare. 2015. [6] Chen C, et al. Quantitative imaging of peripheral trabecular bone microarchitecture using MDCT. Med Phys. 2018; 45(1): 236‒249. [7] Kazerooni E. A. High resolution CT of the lungs. AJR Am J Roentgenol. 2001; 177(3): 501‒519. [8] Flohr T, Schmidt B, Merz J, Aulbach P. SOMATOM Force ‒ Get two steps ahead with Dual Source CT. White Paper. Siemens Healthcare. 2018. [9] Grimes J, et al. The influence of focal spot blooming on high-contrast spatial resolution in CT imaging. Med Phys. 2015; 42(19): 6011‒6020. [10] Kachelriess M, Knaup M, Penssel C, Kalender W. A. Flying focal spot (FFS) in cone-beam CT. 2006; 53(3): 1238‒1247. [11] Rubert N, Szczykutowicz T, Ranallo F. Improvement in CT image resolution due to the use of focal spot deflection and increased sampling. J Appl Clin Med Phys. 2016; 17(3): 452‒466. 15 On account of certain regional limitations of sales rights and service availability, we cannot guarantee that all products/services/feaures included in this document are available through the Siemens Healthineers sales organization worldwide. Availability and packaging may vary by country and are subject to change without prior notice. The information in this document contains general descriptions of the technical options available and may not always apply in individual cases. Siemens Healthineers reserves the right to modify the design and specifications contained herein without prior notice. Please contact your local Siemens Healthineers sales representative for the most current information. In the interest of complying with legal requirements concerning the environmental compatibility of our products (protection of natural resources and waste conservation), we may recycle certain components where legally permissible. For recycled components we use the same extensive quality assurance measures as for factory-new components. Any technical data contained in this document may vary within defined tolerances. Original images always lose a certain amount of detail when reproduced. This product is pending 510(k) clearance, and is not yet commercially available in the United States. Siemens Healthineers Headquarters Legal Manufacturer Siemens Healthcare GmbH Siemens Healthcare GmbH Henkestr. 127 Henkestr. 127 91052 Erlangen, Germany 91052 Erlangen, Germany Phone: +49 9131 84-0 siemens-healthineers.com Published by Siemens Healthcare GmbH · Printed in Germany · 7203 0319 · ©Siemens Healthcare GmbH, 2019