Four-dimensional computed tomography: Image formation and clinical protocol
Isosurface renderings of a spherical object , CT scanned while periodically moving on a sliding table (, ). Top row: different artifacts obtained by standard axial CT scanning. Artifacts depend on the interplay (relative motion phase) between CT data acquisition and object motion; to measure different artifacts CT scans were started at different motion phases. Note that slices of the object can be imaged in mixed order. Top row, right image shows two disconnected parts of the sphere, the upper part consists of the top and the bottom of the sphere. Bottom row: left shows a CT scan of the static object. Other images show 3 positions of the sphere while moving as imaged with 4DCT (29 images per couch position reconstructed, see text for details). Only small residual artifacts on the surface remain, the object is imaged at different positions of the motion cycle.
Performance of the RPM-system for regular, sinusoidal motion (, ). Top: amplitude and phase recorded during 4DCT data acquisition, “beam on” is shown as solid lines, “beam off”—while couch is advancing—dotted. Bottom: left shows the excellent performance of the RPM-system in phase determination for regular motion; right illustrates the regularity of the motion, almost perfect relation between amplitude and phase (data plotted for complete acquisition interval, ).
Axial slices of a spherical object , 4DCT scanned while periodically moving on a sliding table (orthogonal to shown slices, , ). The color map was adjusted to emphasize image formation of axial slices for moving objects. Tube rotation during CT data acquisition results in an angular dependency of the imaged cross sections. Image reconstruction “averages” over angles of a full rotation , resulting in spiral-like images of the spherical object with decreasing reconstructed density from inside to outside.
Central cut through segmentation of objects on the moving phantom, imaged during periodic motion (, ). Top: severe artifacts obtained by standard axial CT scanning. Middle and bottom: 10 phases of the motion cycle as obtained by 4DCT scanning (reconstruction of 10 images per couch position). Motion sequence advances from left to right, middle to bottom. Horizontal lines were plotted to guide the eye. Imaging artifacts are significantly reduced, the full amplitude of the motion is captured.
Central sagittal slice of a spherical object displayed using different HU-thresholds. Top: CT scanned static; bottom: 4DCT scanned, periodic motion parallel and horizontal to shown slices (, ). Thresholds from left to right: HU to 2500 HU in increments of 500 HU. Threshold to obtain true volume of the static sphere in an axial CT scan is 1300 HU. 4DCT shows a more pronounced dependency on the threshold (see text for details).
Isosurface renderings of the phantom. Direction of motion is along a line through the spheres’ centers (, ). Isosurface values are HU (a, c, d) and (b). (a) Axial CT scan of static phantom; (b), (c) 0% phase of 4DCT scan; (d) 20% phase. Spiral shaped residual motion artifacts remain at the poles of the spheres. (b) The lower isosurface value results in more pronounced partial projection effects for the small, high-density objects. (c), (d) The higher isosurface value increases spiral shaped motion artifacts for low density objects, artifacts vary by phase (see text for details).
Motion of a sphere (, , ) measured by 4DCT (, images per couch position, phase tolerance ), measurement repeated 4 times. The sphere was automatically segmented using different thresholds. The sphere was moved between each of the static scans. The top plot shows the trajectory of the sphere’s centroid. Voxel size in the -direction is . Center of object motion does not show a significant dependency on the threshold used for automated segmentation. The bottom plot shows the corresponding relative volume differences. Due to partial projection effects the volume is significantly bigger for lower and smaller for higher thresholds. Volumetric differences show a distinct pattern, which corresponds to the absolute value of a sinusoid. For higher velocity, volumetric differences are increased by changing the threshold. The true threshold obtained from calibration measurements is HU. By setting the threshold to HU the volume of the sphere is segmented within and .
Relative volumetric overlap between registered static and moving spheres for different thresholds and 4DCT scanning parameters: motion period and number of images reconstructed per couch position ( , phase tolerance and for and , respectively). All data are normalized to the volume of the static CT scan. The left column shows results for a sphere of ( HU), the right for (1300 HU), respectively. Plots in the top row show the relative overlap; the middle row shows the relative volume of the moving sphere outside the static sphere’s volume. Plots in the bottom row show overlap between static and moving spheres after decreasing the radii by 10%. Data are shown for thresholds every 100 HU, red and green curves were shifted for displaying. For each threshold data for 10 motion phases are plotted, vertical lines indicate variability within a 4DCT series.
4DCT volumes at tidal end expiration and end inspiration (top) in comparison to two CT scans acquired during breath hold (bottom) at end expiration and inspiration. For breath hold scans, the patient was instructed to hold his/her breath at normal tidal volumes for fast helical CT scanning. Inhale and exhale respiratory states are color coded in yellow and blue respectively; matching high-density structures appear in grey when they overlap. This illustrates that breath hold CT scans do not necessarily represent patient anatomy during respiration.
Lung tumor, comparison of same sagittal slice. Top: helical scan, acquired during light breathing; middle: CT scans at end-expiration (left breath hold, right 4DCT volume); bottom: CT scans at end-inspiration (left breath hold, right 4DCT volume). 4DCT parameters: , , phase tolerance .
Respiratory motion measured on the abdominal surface for 2 patients, recorded with the RPM-system during 4DCT data acquisition. The top row shows respiratory amplitudes, data acquisition intervals indicated by solid lines; dashed lines indicate advancing of the CT couch between data acquisitions. The middle row shows the corresponding phases as assigned by the RPM-system. Due to the irregular respiratory trace of patient 2 (right column) the RPM-system encountered problems in assigning correct phases. It has to be pointed out that the RPM-system, originally designed for gating, considered most of the respiratory trace irregular; therefore a gated therapeutic beam would have been turned off. The bottom row shows a comparison between amplitudes and assigned phases. All data points are plotted .
Sagittal cut through a patient’s 4DCT data, both images show the same respiratory phase. The patient showed an irregular breathing pattern (see Fig. 11, right column). Retrospective sorting was based on respiratory phases assigned by the RPM-system for the top image, and according to amplitudes for the bottom image. Resorting according to motion amplitudes reduces the residual artifacts at the top of the diaphragm. Partial projection effects are visible in the lower part of the image at the diaphragm.
Different phases of liver motion as obtained by 4DCT scanning (, , phase tolerance ), end inhalation is shown on the left, end exhalation in the middle. The right image displays the same coronal cut for a standard helical CT scan. Note the motion artifacts resulting in a distorted image of the liver and tumor. Lines were added to guide the eye.
Coronal cuts of a thoracic cancer patient as obtained from 4DCT data acquisition. The top row shows 4 different phases of the breathing cycle, left to right corresponds from end inhalation to end exhalation. An outline, containing the tumor at all breathing phases, is shown in red. The bottom row shows different possible motion artifacts. Artifacts were simulated by randomly mixing different phases of 4DCT data at different couch positions. The images illustrate different artificial shapes of the tumor that could be imaged by standard CT scanning. Differences in artifacts depend on the interplay between CT scanning and respiratory motion.
Volume renderings of different 4DCT data sets (, , phase tolerance ): end inhalation (top), mid exhalation (middle), end exhalation (bottom). Posterior view of a lung tumor patient, the back of the patient rendered transparent. The rendering is based on HU values only, the color map was set to visualize tumor and structures in the lung.
Spheres in the motion phantom. Arrangement of the spheres on the moving platform, see Figs. 4 and 6.
Article metrics loading...
Full text loading...