:: rsna 2003 presentations
ImPACT was involved with a number of presentations and educational exhibition posters at the RSNA 2003 meeting in Chicago, which runs from the 30th November to the 5th December. Click for the meeting report, looking at what was new in CT scanning at RSNA 2003.
The presentations were:
Low contrast detail detectability measurements on multi-slice CT scanners
N Keat and S Edyvean, Sunday 30th November, 10:45-10:52, Room S404AB
A Visual Method for Demonstrating the Relative Performance of Cone Beam Reconstruction Algorithms
DJ Platten, C McCollough, N Keat, S Edyvean, Friday 5th December, 11:40-11:47, Room S401AB
The educational exhibit posters were:
Artefacts in CT: recognition and avoidance
JF Barrett, N Keat
A methodical approach for comparison of CT image quality relative to dose
S Edyvean, N Keat, MA Lewis, JF Barrett, DJ Platten
Managing radiation dose and image quality in multidetector CT of the abdomen: experience in a large teaching hospital
YH Kim, YH Auh, S Saini, N Keat, H Ji, PR Mueller, PR Ros
Low contrast detail detectability measurements on multi-slice CT scanners (view presentation)
N Keat and S Edyvean
Purpose: To investigate the detectability of low contrast details on multi-slice CT scanners using a Catphan phantom. Subjective assessments of images of this phantom were made for a selection of 4 and 16 slice scanners. Variations of detectability with scan parameters and inter- and intra- viewer variability were assessed. An evaluation of the scanners' performance under standardised exposure conditions was also performed.
Methods and Materials: The low contrast section of a Catphan 550 phantom (The Phantom Laboratory, Salem, NY) was scanned under a range of exposure conditions, on multi-slice CT scanners from GE Medical Systems (LightSpeed Plus, LightSpeed 16), Philips Medical Systems (Mx8000), Siemens Medical Solutions (Sensation 4, Sensation 16) and Toshiba Medical Systems (Aquilion Multi, Aquilion 16). This phantom contains nine circular objects from 2-15 mm diameter, in a uniform background, at each of three nominal contrast levels; 0.3%, 0.5% and 1.0%. The 0.3% objects were scored for visibility, with exposure characterised by the CTDI as measured on the surface of the phantom. 20 images were assessed for each exposure condition, and when making comparisons, images were presented to the viewer in a random order.
Results: Inter- and intra-operator variability was assessed. Four observers scored 80 images for the smallest visible object in the phantom. The mean value of the standard deviation of object scores for each image was 1.1 objects. There was complete agreement between observers on the smallest visible object in only six images (7.5%). One observer scored a group of 20 images on five separate occasions, greater than a month apart. The mean value of the standard deviation for each image was 1.4 objects. There was agreement on limiting object size in none of the images. Within one image assessment session, for a single observer, the results were more consistent, with detail visibility closely following the expected trend as mAs was increased. In addition, although absolute values of detail visibility varied, general trends were apparent in the results when the same images were assessed on multiple occasions, or by multiple observers.
Conclusion: Comparison of low contrast detectability performance is made very difficult by the inherent subjectivity of the standard methods used to assess it. Meaningful comparisons are best made by viewing the images that are being compared in a single session.
A Visual Method for Demonstrating the Relative Performance of Cone Beam Reconstruction Algorithms (view presentation)
DJ Platten, C McCollough, N Keat, S Edyvean
Purpose: To devise a method that visually demonstrates the relative performance of cone beam reconstruction algorithms.
Methods and Materials: Several commercially available multi-detector row CT systems were used to obtain helical CT images of a plastic funnel. The mean CT number within the plastic was approximately -150 HU. The maximum diameter of the funnel was 130 mm; the minimum diameter at the neck of the funnel was 20 mm, and the height of the funnel was 80 mm. The funnel neck was aligned at isocenter and held in place using an ionization chamber stand. The funnel was placed such that the diameter increased along the z axis. Scans were acquired at a variety of detector collimations (0.625, 0.75, 1, 1.25 and 1.5 mm) and images reconstructed at varying scan widths (0.625, 0.75, 1, 1.25, 2, and 3 mm). Images were reconstructed both with and without cone beam algorithms, as appropriate to the specific system. Maximum intensity projections (MIP) images were created in the transaxial and sagittal planes to form a collapsed end-on view and a side view of the funnel.
Results: The MIP images generally showed less severe artifacts for images reconstructed using cone beam algorithms. Images reconstructed from the same helical data set with and without the cone beam algorithm clearly demonstrated the benefits of using cone beam reconstructions. For data acquired with sixteen 0.75-mm detector rows, 2-mm wide scans were reconstructed using a cone-beam algorithm (AMPR) while 3-mm wide scans were reconstructed without a cone beam algorithm. The transaxial MIPS (end-on-view of the funnel) created from the 2-mm wide scans (1-mm recon increment) from data acquired at pitch values of 0.5, 1, and 1.5 all show fine concentric circles of uniform brightness, representing discrete, artifact-free axial images of the funnel at varying diameters along the z axis. The transaxial MIPS created from the 3-mm wide scans at the same recon increment and pitch values show discontinuous spirals of markedly varying brightness. The artifactual appearance increased as the pitch increased. The sagittal MIP images showed a smooth rendering of the side-view of the funnel for the 2-mm images, yet a discontinuous, artifactual appearance of the funnel for the 3-mm images.
Conclusion: The proposed testing method provides a clear visual indication of the ability of cone beam reconstruction algorithms to provide realistic 2- and 3-D renderings of high-contrast objects that vary dramatically along the z axis. This can provide a more intuitive understanding of cone beam reconstruction algorithms.
Artefacts in CT: recognition and avoidance (view poster)
JF Barrett, N Keat
CT systems are inherently more prone to artefacts than conventional radiography. They appear as streaks, rings or shaded areas which can seriously degrade the CT image, sometimes to the point of making it diagnostically unusable. They originate from a variety of sources: some are a consequence of the physical processes involved in x-ray attenuation and detection, some are patient-related, some are due to imperfections in scanner function and some are a result of the interpolation algorithm used in helical scanning. Certain types of artefact can be partially corrected for in the software, but good scanner design, careful patient positioning and optimum selection of scan parameters all contribute to the minimisation of artefacts in CT.
A methodical approach for comparison of CT image quality relative to dose (view poster)
S Edyvean, N Keat, MA Lewis, JF Barrett, DJ Platten
CT image quality is often quoted using standard imaging performance parameters: the standard deviation of pixel values in a uniform water phantom; data from the modulation transfer function; and the measured imaged slice width. Dose is quoted using a standard dosimetry parameter, the volume computed tomography dose index (CTDIvol). It is not always possible to make clear judgments in the comparison of scanners and scanning protocols due to the interdependence of these performance parameters. To overcome this, the dose and image quality values must be normalised and combined for easier interpretation and comparison. The relationship between noise and dose is well understood, as is that between noise and slice width. However the influence of the resolution algorithm on noise is less well defined. An organised methodical approach is needed to incorporate this factor into any image quality and dose relationship. A graphical approach for comparing dose and image quality is demonstrated. A single number can also be extracted from the data to summarise the relationship; this number is more abstract, but can be useful for image quality characterisation.
To demonstrate the methodology for an easy comparison of image quality and dose, by combining standard performance parameters, for different scanners or scan protocols. Data can be presented graphically for easy interpretation or as a single number for characterisation.
Managing radiation dose and image quality in multidetector CT of the abdomen: experience in a large teaching hospital
YH Kim, YH Auh, S Saini, N Keat, H Ji, PR Mueller, PR Ros
Patient radiation dose is determined by factors such as CT scanner design, scanning parameters, patient size and anatomic region under investigation. This exhibit describes elements of CT scanners which vary among vendors that affect radiation dose efficiency and image quality. These factors include pre-patient beam collimation, X-ray filter, automated dose modulation, detector configuration and reconstruction methods. This exhibit will describe the various scanning parameters and how modifications should be made based on patient size and clinical applications. The exhibit will also demonstrate how different scanning parameters impact patient radiation dose and image quality. Image quality in terms of noise, spatial resolution, and contrast resolution will be explained with illustrations from phantom experiments performed with different scanning parameters. Various scanning parameters such as tube current (mA) and potential (kVp), scan length and collimation, table speed and pitch, and scanning modes used in teaching hospitals in specific clinical indications such as CT colonography and urinary tract calculi will be described.
To identify the scanner design and acquisition parameters that affect patient dose. To correlate the changes in image quality resulting from parameter adjustments using a phantom. To demonstrate the varying scanning parameters used in different large teaching hospitals.