Factors Affecting Contrast-Detail Performance in Computed Tomography

Journal of Medical Imaging and Radiation Sciences

Journal of Medical Imaging and Radiation Sciences 44 (2013) 62-70

Journal de l’imagerie médicale et des sciences de la radiation

www.elsevier.com/locate/jmir

Factors Affecting Contrast-Detail Performance in Computed Tomography: A Review

Haney Alsleem, BSc, MSca* and Robert Davidson, PhD, MAppSc(MI), BBus, FIRb

aRMIT University, Bundoora, Victoria 3083, Australia b School of Dentistry and Health Sciences, Faculty of Science, Charles Sturt University, Wagga Wagga 2678, Australia

ABSTRACT

This article reports on recent research findings into the factors that influence the detectability performance of different systems of com-

puted tomography (CT) scanners. These systems include multidetec- tor CT (MDCT) of different slice numbers, dual-source CT (DSCT), and cone-beam CT (CBCT). The introduction of more sli-

ces for MDCT, DSCT, and the new technology of CBCT increases the need to optimize the image quality and to examine the potential reduction of radiation doses to the patient. Low-contrast detail de-

tectability is a method that has proven to be an appropriate evalua- tion method for this purpose. However, it is essential to recognize factors that affect detectability performance and understand how

these factors influence image quality and radiation dose. It is argued that deep understanding of the influences of these factors is the key to image quality optimization in terms of contrast-detail detectability and radiation dose reduction. The purpose of this article is, therefore,

to specify these factors and to explain their influence on detectability performance and hence on CT image quality. Further low-contrast detail studies are required to optimize imaging performance of differ-

ent CT systems and scanners.

* Corresponding author: Haney Alsleem, BSc, MSc, RMIT University,

Discipline of Medical Radiation, Bundoora, Victoria 3083, Australia.

E-mail address: haniims@hotmail.com (H. Alsleem).

1939-8654/$ – see front matter � 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jmir.2012.12.001

R�ESUM�E

Cet article porte sur les r�esultats de recherches r�ecentes sur les facteurs qui influencent le rendement de d�etection des scanners de diff�erents syst�emes de tomodensitom�etrie (TDM). Ces syst�emes comprennent les syst�emes de TDM �a d�etecteurs multiples avec des nombre de tranches diff�erents, la TDM �a double source et la TDM �a faisceaux coniques. L’introduction d’un plus grand nombre de tranches pour la TDM �a d�etecteurs multiples, la TDM �a double source et la TDM �a faisceaux coniques augmentent la n�ecessit�e d’optimiser la qualit�e de l’image et d’examiner la possibilit�e de diminuer la dose de radiation pour le patient. La d�etectabilit�e des d�etails �a faible contraste est une m�ethode d’�evaluation qui s’est av�er�ee appropri�ee �a cette fin. Cepend- ant, il est essentiel de reconnâıtre les facteurs qui influent sur le ren- dement de d�etectabilit�e et de comprendre comment ces facteurs influencent la qualit�e de l’image et le dosage de radiation. On croit qu’une compr�ehension approfondie de l’influence de ces facteurs ser- ait la cl�e de l’optimisation de la qualit�e de l’image en termes de d�etectabilit�e des d�etails �a faible contraste et de r�eduction de la dose de radiation. Le but de cet article est donc de d�eterminer ces facteurs et d’expliquer leur influence sur le rendement de d�etectabilit�e, et donc sur la qualit�e des images de TDM. Il faudra d’autres �etudes sur les d�etails �a faible contraste pour optimiser le rendement d’image- rie de diff�erents syst�emes et scanners de TDM.

Keywords: Low-contrast detail detectability; image quality; MDCT; CBCT

Mots-cl�es: D�etectabilit�e des d�etails �a faible contraste; qualit�e de l’image; TDM �a d�etecteurs multiples; TDM �a faisceaux coniques

Introduction

Computed tomography (CT) imaging technology is rapidly changing the conditions of image quality optimization and radiation dose reduction. Each CT system has its own specific image quality [1]. The introduction of multidetector CT (MDCT), using an increasing number of slices, dual-source

CT (DSCT), and cone-beam CT (CBCT) has enormously in- creased the range of examinations, which has in turn increased the number of CT examinations [2, 3]. To further add to this technological complexity, different technical applications and software are utilized in systems from different manufacturers, and various models of CT scanners utilize different algorith- mic software [1].

Several studies have shown that there is still misdiagnosis or loss of information in CT images, as some pathologic lesions and details are not detected by interpreters [4–6].

Figure 1. The detectability performance can be optimized by balancing

between the adjusted protocols parameters (milliampere-second [mAs], peak

kilovolt [kVp], slice thickness/pitch, and software processing) and tolerated

noise and artifacts while maintaining low radiation dose.

Although contrast and temporal resolutions have been signif- icantly improved by the current advanced technology of MDCT, the spatial resolution or in-plane spatial resolution has not improved. Therefore, there are still some limitations in the rate of detection and accurate assessment [7–9]. Fur- thermore, the highest radiation dose from medical imaging modalities is received from CT scans [10]. Thus, dose reduc- tion has become a very important goal in CT applications [11]. However, there are tradeoffs between image quality and dose; the higher the dose contributing to the image, the lower image noise, and hence, the better visualization of low-contrast structures. Detection of low-contrast details and lesions is primarily limited by noise, which can be re- duced by increasing radiation dose [12, 13]. Consequently, there is an imperative need for image quality optimization and radiation dose reduction for CT images.

Several methods are used to evaluate imaging performance and image quality. Detection quantum efficiency, receiver- operating characteristics, visual grading characteristics, and low-contrast detail (LCD) detectability are all commonly used methods [14, 15]. However, several authors state that LCD is the most appropriate method to optimize image qual- ity and to examine the potential of radiation dose reduction [16, 17].

Since the common task of diagnostic CT scan images is the visual detection of lesions, detectability performance is an im- portant measure of image quality [18]. The theory behind LCD implies that the detectability of details increases with the increasing size of objects or contrast between objects and their background [14, 19]. LCD is usually measured by using low-contrast detail phantoms that contain cylindrical objects of a range of different sizes and low-contrast levels [20, 21]. LCD phantom images are assessed subjectively by interpreter observation or objectively by measuring the contrast-to-noise ratio (CNR) [22]. LCD can also be used to compare and contrast the performance of different imaging systems [23]. LCD studies are also useful to examine image optimization and to assess the potential of dose reduction of imaging systems [17, 24]. However, recognizing and under- standing the factors that influence the detectability perfor- mance of different CT scan systems are fundamental concerns in effectively implementing this method.

The purpose of this review is to determine the factors influencing LCD performance of different CT systems and to explain their influences on image quality optimization.

Factors Affecting Low-Contrast Detail Performance in CT

Detectability performance of CT imaging systems is influ- enced by CT system specification, milliampere-second, peak kilovoltage, slice thickness, pitch, and beam collimation, as well as image processing and visualization. These factors should be adjusted to optimize image quality in terms of LCD performance by lowering image noise and maintaining lower radiation dose to the patient (Figure 1).

H. Alsleem and R. Davidson/Journal of Medical Ima

Scanner Systems

Each CT system and model has its own performance abil- ity according to its properties and specifications (Figure 2). The design criteria employed in CT systems fundamentally characterize the type of noise, which in turn affects the detect- ability performance of the produced images [25]. Image blur is largely determined by scanner specifications. The size of the sampling aperture is regulated by the focal spot size and the detector size; the size of the voxels is considered a blurring source [26]. Each system has limited gantry rotation time and area coverage, which influence image quality and lesion detectability [11, 27]. Accordingly, the influences of imaging factors on image quality vary across different systems, models, and manufacturers.

The following discussion will show that the detectability performance of different CT systems and scanners is not the same. Although the latest MDCT and other newer technolo- gies are suggested to have better image quality, they still have limitations that may affect the LCD.

Spiral CT (Single-Slice CT)

The introduction of spiral or helical CT in 1989 increased the advantages and the applications of CT systems [9]. Spiral CT involves continuous patient translation and continuous radiation exposure during the rotation of x-ray tube and data acquisition (Figure 2A). Therefore, a volume dataset is obtained in a relatively short period of time compared to single-slice CT [9]. However, in single-slice spiral CT, coro- nary artery imaging could only be possible by gating or under

ging and Radiation Sciences 44 (2013) 62-70 63

Figure 2. A, Spiral computed tomography (CT) single slice, helical CT scanner with single row detector (modified from Exxim Computing Corp). [39]. B, Multi-

detector CT, CT scanner with multiple row detectors (modified from Exxim Computing Corp). [39]. C, Dual-source CT, CT scanner with two x-ray tubes

(modified from Exxim Computing Corp). [39]. D, Cone-beam CT, CT scanner with flat panel detector (FPD) (modified from Exxim Computing Corp) [39].

ideal imaging conditions with slow heart rates, and even then they can have biphasic motion artifacts [28].

Multidetector CT

Figure 3. Sagittal reformatted computed tomography image obtained with

5-mm collimation and a 5-mm reconstruction interval shows stair-step

artifacts [30].

MDCT is a spiral CT scanner with more than one row of detectors; there may be 4, 16, 64, 256, or 320 detector rows that are able to generate many slices simultaneously and com- plete multiple scans in seconds or in a subsecond period (Figure 2B). Recent MDCT can also provide isotropic resolu- tion and reconstruct cross-sectional images in arbitrary planes [28, 29].

It is accepted that MDCT improves and enhances the im- age quality of CT studies. MDCT systems are better than single-slice CT because they have faster scanners and smaller detector element size, cover a larger area, and use enhanced reconstruction algorithms. Sensitivity and specificity of MDCT to detect pathologies, particularly cardiovascular diseases, are much higher than that of single-slice CT [28]. The entire chest of a patient can be scanned by multislice CT with 1-mm slices and within one breathhold. Spatial res- olution becomes much higher with MDCT [28, 29]. Hence, the accuracy of diagnostic applications is improved.

The development of more row detectors CT has reduced examination time and has improved image quality. Gantry ro- tation time is significantly reduced by using faster MDCT

64 H. Alsleem and R. Davidson/Journal of Medical Ima

scanners, and thinner slices are obtained by utilizing thinner detector rows. Shorter gantry rotation time improves tempo- ral resolution, which is essential to reduce the effects of mo- tion artifacts. Thinner slices improve spatial resolution, which minimizes the effects of partial volume average and cal- cium artifacts [28]. Stair-step artifacts (Figure 3) are almost eliminated with more detector rows scanners, particularly with 64 or more detector rows MDCT. Stair-step artifacts occur around the edges of structures in the volume or

ging and Radiation Sciences 44 (2013) 62-70

multiplanar reformatted images, in particular when wide col- limations and nonoverlapping scanning are used [30].

The 256–detector row MDCT is able to cover 128 mm of anatomy with 0.5 mm slices. In addition, the number of channels in the radial axis has been increased in this scanner. It is able to generate fine, isotropic resolution of structures. Cardiac imaging is especially difficult because of heart beating motion and tiny coronary arteries structures. The 256 MDCT scanner provides higher image quality and has higher poten- tial of radiation dose reduction than that of previous scanners. It ensures more accurate and quicker diagnoses [28]. The pos- sibility of fusing the images with CT angiography examina- tions allows morphologic and functional assessment [31].

The most recent development of MDCT includes an in- creased number of detector rows, up to 320. The gantry rota- tion time (350 milliseconds) is shorter than many 256 MDCT scanners; a 320 MDCT system achieves complete coverage of the heart within a single rotation without table movement, as this system can cover 160 mm of anatomy [28, 32]. Volumetric imaging of the entire heart is completed within one cardiac cycle [31]. This system enables assessment of smaller coronary vessels up to 1.5 mm and detection of small volume plaque [8]. The results of the study conducted by Khan et al [33] suggest that with the same image quality of 64 MDCT, 320 MDCT has the capability to significantly reduce radiation doses delivered to patients. The wider area coverage and faster gantry rotation offered by 320 MDCT improves the temporal resolution and avoids exposure- intensive overscanning [34]. The higher temporal resolution offers a potential to significant dose reduction and reduces the motion effects of heart structures on the image [34].

However, multiplanar and three-dimensional reformation approaches generate artifacts. Zebra artifacts appear as faint stripes, with an increased degree of noise (Figure 4) [30]. In- terpolation methods, which were developed with the intro- duction spiral scanning, also contribute to additional artifacts on images [35]. Because of the continuous tube and table motion, the projections processed in a spiral motion around the patient and did not lie in a single plane. Hence,

Figure 4. Computed tomography image obtained with helical shows zebra

artifacts [30].

H. Alsleem and R. Davidson/Journal of Medical Ima

interpolation reconstruction algorithms methods are used to generate projections in a single plane [11]. The effects of interpolation artifacts increases as pitch and the number of detector rows increase. Interpolation artifacts also cause inac- curacies in CT number assessment, which leads to misdiagno- sis [35].

Furthermore, one of the main disadvantages of the MDCT system is the use of wider beam collimation, which leads to image deterioration compared to sequential type scanners [35]. Wider collimation is required for more detector rows, and the x-ray beam, in turn, becomes cone shaped. The new image reconstruction technology using cone-beam algo- rithms creates negative effects on image quality called cone- beam artifacts. The greater the divergence of cone beams is, the larger the artifact effects [35]. Cone-beam artifacts occur because the data collected from each detector during gantry rotation do not correspond to the ideal flat plane but actually to volume contained between two cones. The artifacts caused by cone beam are similar to those caused by partial volume around off-axis structures. Cone-beam artifacts are less pro- nounced for the inner detector rows than for the outer rows [30].

In-plane spatial resolution is not improved by the current advanced technology of MDCT. Current reconstruction methods are focused on cross-plane spatial resolution, and the focus has not been on spatial resolution with the two- dimensional image plane [8, 9].

Dual-Source CT

DSCT utilizes two x-ray tubes that are arranged in a single gantry at a 90-degree offset. Although the two tubes and de- tectors are operated simultaneously, a quarter rotation of the gantry is sufficient to collect the data necessary for one image. So that a gantry rotation time of 330 milliseconds provides an effective scan time of 83 milliseconds in the centre of rotation, the exposure time can be reduced by a factor of two (Figure 2C) [36]. As a result, temporal resolution can be in- creased by the same factor [28, 36, 37]. However, DSCT has similar difficulties of recent MDCT because of the intrin- sic limitations of CT image reconstruction matrix and spatial resolution [38].

Flat-Panel Volume or CBCT

The most recent development in CT technology is CBCT. In CBCT (Figure 2D), flat-panel detectors instead of the mul- tidetector rows in MDCT are utilized [39]. CBCT is prom- ising for diagnostic and interventional clinical purposes because of the capability for high-spatial resolution volumetric imaging and dynamic CT scanning [40].

The wide coverage (z-axis) flat-panel detector allows for imaging of entire organs, such as an entire heart or brain, in one axial scan. Moreover, flat-panel detectors mostly con- sist of �200 mm detector element size. CBCT detector with a 150-mm element size can provide ultrahigh spatial resolu- tion up to 150 � 150 mm. Therefore, CBCT is superior in its spatial resolution compared with MDCT, which provides

ging and Radiation Sciences 44 (2013) 62-70 65

spatial resolution only up to approximately 400 mm in plane and approximately 500 mm in the z-axis direction. The two-dimensional flat panel detectors allow for imaging at any arbitrary angle. In comparison with MDCT, CBCT allows for thinner sections using similar radiation doses as MDCT [40]. However, the two processes are different in the sense that MDCT has superior contrast resolution com- pared with CBCT. In addition, MDCT also allows for a faster scanning time compared with CBCT, which has a slow cae- sium iodide scintillator that is utilized in the flat-panel detec- tor. Their use limits the projection acquisition time to 100 frames per second, compared with 900 to 1,200 projections during a single 0.5-second rotation for MDCT [40].

Tube Current, Tube Current-Time Product, and Radiation Dose

Image quality improvement requires reducing noise and in- creasing signal-to-noise ratio (SNR); however, this typically im- plies additional radiation to the patient.Doubling the signal, the radiation dose, is required to increase SNR by a factor of 1.4. The clinical situation determines the acceptable radiation dose, although the high-dose protocol is not recommended (Figure 5) [41]. Decreasing the milliampere-second reduces ra- diation dose, but at the same time, it increases image noise and reduces the contrast-to-noise ratio (CNR). The visibility of structures is negatively influenced by the reduction of x-ray quanta amount [42, 43]. Furthermore, spatial resolution prop- erties can also be affected by the radiation dose, and there is a tradeoff between them [44]. However, the acceptable level of tradeoff in image quality should be determined according to the diagnostic purpose and clinical task being performed [13].

LCD is improved by increasing themilliampere-second to re- duce the noise.However,milliampere-second should be adjusted to minimize radiation dose while maintaining optimum LCD.

Tube Potential

Subject contrast is increased with reducing kilovoltage. Low kilovoltage increases photoelectric interactions, which

Figure 5. The relationship between milliampere-second [mAs], noise, signal-

to-noise ratio (SNR) and contrast-to-noise ratio (CNR) and the low-contrast

detail (LCD) detectability is illustrated. Increasing milliampere-second re-

duces the noise and increases SNR and CNR and as a result LCD is im-

proved. However, increasing milliampere-second increases the radiation

dose to the patient.

66 H. Alsleem and R. Davidson/Journal of Medical Ima

improve the attenuation level, leading to image contrast en- hancement and improved detail visualization [13, 45]. How- ever, subject contrast must not be confused with visualized or displayed contrast, which can be altered on the monitor.

Godoy et al [46] suggest that, although the measured im- age noise was higher in the low-kilovolt images, the subjective quality of the image was higher for 80 kVp than for 140 kVp images. This result highlights the fact that SNR and CNR de- termine the image quality more than image noise. Further, Funama et al [43] found that low kilovolts can improve CNR, and their result suggests using 90 kVp rather than selecting 120 kVp. The noise with low-kilovoltage images did not cause a reduction in image quality due to the higher SNR and higher attenuations (Figure 6) [46–48].

Radiation dose to the patient is not linear with kilovoltage, but reducing the kilovoltage does reduce the amount of radi- ation when other exposure factors are fixed [13]. Funama et al [43] suggest that the radiation dose can be reduced by 29% without affecting the CNR by selecting 90 kVp instead of 120 kVp [43]. A study conducted by Zhang et al [49] sug- gested that selecting 100 kVp compared with 120 kVp in 320 MDCT can reduce the radiation dose to patients without degrading image quality.

A multireader study conducted by Ertl-Wagner et al [45] examined the impact of various kilovoltages, specifically 80, 120, and 140 kVp, while all other exposure factors were fixed on image quality of vessel delineation with cranial MDCT. Their study showed that the higher voltages were superior in terms of having the greatest effects for vessels close to bone and subsegmental arteries.

Lower kilovoltage techniques reduce the total energy flux if other exposure factors are not adjusted, which increases the image noise, leading to reduction in image quality and diag- nostic accuracy of CT [13, 45, 46, 50]. Artifacts, such as beam hardening, may also be increased with low-energy beams, such as 80 kVp [13]. Huda et al [50] suggest that im- age noise from low kilovoltage can be suppressed by utilizing a new adaptive filter.

Figure 6. The relationship between peak kilovoltage (kVp), signal-to-noise

ratio (SNR), contrast-to-noise ratio (CNR), noise, and the low-contrast detail

(LCD) detectability is illustrated. Appropriately lowering kilovoltage increases

photoelectric interaction (PEI) and the attenuation level (AAL), which leads

to an increase in SNR and CNR, and hence LCD performance is improved.

However, the noise level increases with excessively lowering kilovoltage or if

the other exposure factors are not adjusted.

ging and Radiation Sciences 44 (2013) 62-70

In addition, although the selection of low kilovoltage mini- mizes radiation to patients, kilovoltage should be selected de- pending on the patient’s cross-section diameter and should be adjusted according to the task. For example, while 80 kVp should be used in small children, 140 kVp should be used in obese patients [9].

According to the above discussion, the interdependence of image quality and radiation dose on kilovoltage is very com- plex. The kilovoltage should be optimized to be low enough to increase contrast resolution in order to improve LCD, but high enough to reduce the noise and minimize radiation. Patient size and examination purpose should also be consid- ered in kilovoltage selection to optimize LCD performance.

Pitch, Beam Collimation, and Slice Width

Thinner submillimetre slice thicknesses are routinely acquired when using MDCT and provide high-resolution isotropic image datasets. Therefore, through-plane, partial- volume averaging effects are minimized and image postpro- cessing, such as three-dimensional reconstruction, multiplanar reformatting software-assisted lesion detection, and quantifi- cation, is optimized [9, 51]. However, when thinner slices are used, noise will be increased. This is important because low-contrast detectability is degraded by noise [41]. Thus, exposure factors should be increased to reduce image noise. To do so, the radiation dose to patient will also be increased (Figure 7) [13, 41].

Radiation dose from slice thickness selection is also influ- enced by overranging and overbeaming. Overranging is when additional gantry rotations are automatically performed by the scanner to acquire enough data for image construction, where the rotation number increases with increasing collima- tion, when increasing section thickness in the primary recon- struction, and when increasing pitch [52, 53].

Overbeaming occurs when the actual profile beam collima- tion, which is determined by changing the active detectors number or their length in MDCT, widens larger than the nominal beam widths to keep uniform distribution of

Figure 7. The relationship between pitch, slice thickness, noise, and low-

contrast detail (LCD) detectability is illustrated. Selecting lower pitch allows

producing thinner image slices. Thinner image slices reduce the problem of

partial volume averaging and hence the LCD is improved. However, thinner

slices selection increase image noise, which in turn deteriorates LCD, if the

radiation dose is not increased.

H. Alsleem and R. Davidson/Journal of Medical Ima

radiation across the detector bank [53]. Overbeaming is explained as the resultant penumbra effect, which depends on the type of MDCT scanner. Overbeaming increases the ra- diation dose, but it may be reduced by selecting thicker slices or by utilizing more channels. For example, overbeaming effects are diminished much more with 16 MDCT scanners compared with 4 MDCT [53]. Therefore, there is a tradeoff between the advantages of nearly isotropic voxels, which influ- ence spatial resolution, and the disadvantages of image noise and radiation dose when determining the choice of section thickness [41].

Noise from thin slice thickness can be reduced to improve low-contrast lesion detectability by increasing radiation doses to increase SNR, using soft reconstruction kernels, applying various data filters, adjusting window and level settings, and utilizing sliding-thin-slab averaging. Using a sliding-thin- slab averaging algorithm with thin-section scanning during image reconstruction can reduce the effects of through- plane, partial-volume averaging and improve the detectability of low-contrast objects by the retrospective generation of thicker sections [41].

Thus, selecting slice thickness, according to the diagnostic purposes, is fundamental to acquiring higher LCD while maintaining desired spatial resolution and lower radiation dose.

Image Reconstruction, Processing, and Visualization

CT images are available in digital form, which can then be processed, manipulated, and modified directly by computer algorithms. Density values, histograms, and other tissue parameters can be determined at any time in digital CT im- ages. Different orientation views, such as coronal, sagittal, and oblique planes, can be reformed from the original axial images. Two-dimensional images can be reconstructed to three-dimensional displays. Virtual endoscopic views, interac- tive manipulation of image volumes, and four-dimensional animated studies are also possible applications in CT imaging. Automatic determination of specific tissues is now possible with advanced image processing approaches [9].

The recent implementation of iterative reconstruction techniques that are being used instead of filtered back- projection is a promising strategy for decreasing noise and ar- tifacts on CT images. Iterative reconstruction algorithms are statistical reconstruction methods that require higher compu- tational capabilities compared to analytical methods such as filtered back-projection. Iterative reconstruction methods consist of three main steps. First, the artificial raw data are cre- ated, then the artificial and measured raw data are compared, and then an updated image is computed, which is then back- projected to the current volumetric image. These steps are re- peated iteratively and form the iterative reconstruction loop. When the loop is terminated, the final volumetric image is produced [54].

Iterative reconstruction techniques demonstrate impressive improvements in noise reduction and image quality in com- parison with filtered back-projection [54–56]. Iterative

ging and Radiation Sciences 44 (2013) 62-70 67

reconstruction algorithms have the potential to reduce the radiation dose as they reduce image noise and various arti- facts. In addition, iterative methods avoid introducing new artifacts because of approximations, as more intuitive and na- tural ways of image reconstruction are represented by these methods. They also improve image quality, as they are more suited for dealing with missing data or irregular sam- pling. Iterative methods provide higher flexibility in the scan geometry, as many various trajectories are possible since no explicit expression of an inverse transform is needed [54].

Different iterative techniques are implemented in clinical CT. GE Healthcare started with adaptive statistical iterative reconstruction in 2008, followed by GE’s Veo technology, a more complex model-based iterative reconstruction method, in 2009. Siemens implemented image reconstruction in image space in 2009, and recently introduced is the sinogram- affirmed iterative reconstruction, a technique that works in both raw data and image space. Philips announced their iter- ative reconstruction, iDose, in 2009 [54, 56]. Iterative recon- struction techniques are performed either from the image data alone, such as iterative reconstruction in image space, Siemens Healthcare, from both the projection and image data such as adaptive statistical iterative reconstruction, sinogram-affirmed iterative reconstruction, and iDose, or from projection data alone. These techniques may be used either to reduce the amount of image noise in order to improve image quality in large patients or to reduce radiation doses in small or intermediate-sized patients while maintaining diagnostically adequate noise [56].

The detectability of low-contrast detail can be improved by selecting appropriate soft reconstruction kernels while main- taining low radiation doses to the patient [57, 58]. Conse- quently, diagnostic accuracy is improved [58]. Using the wrong filters for the reconstruction algorithm degrades image quality and reduces diagnostic reliability [59]. Unfortunately, there is no general recommendation that can be made for the optimum settings of reconstruction algorithms because their properties are not standardized and vary greatly between ven- dors and scanner types [58]. Generally, there is a tradeoff be- tween selection of a specific reconstruction algorithm and the desired spatial resolution and the tolerated image noise [12].

The window level and window width can be adjusted to display a CT image with appropriate contrast. Although the window level determines the centre CT number value dis- played by the range of grey scale, the window width deter- mines that range of CT numbers. The window level and window width settings dictate how the actual measurements of tissue attenuation are converted into a grey-scale image. They are adjusted according to the tissue properties and diag- nostic purposes. For example, to precisely visualize soft tis- sues, narrow width can be selected, and to accurately demonstrate the bone width, wide window widths can be used [60].

The factors of soft image display, such as monitor bright- ness, display function, resolution, room illumination, and image size, also affect the detectability of lesions [61]. High

68 H. Alsleem and R. Davidson/Journal of Medical Ima

display contrast is required to visualize low-contrast features. That can be achieved by increasing the contrast of the moni- tor and by reducing window width as far as possible without loss of diagnostic information of medical image [62].

The commonly available display devices are liquid-crystal display monitors (LCDM) and cathode-ray tube (CRT) mon- itors. LCDM are increasingly used in medical imaging depart- ments for their inherent advantages [63]. The dynamic range provided by LCDM is larger than that provided by CRT monitors. Higher small-spot contrast ratio is provided by LCDM more so than CRT monitors. However, when LCDM are visualized from different angulated views, the con- trast resolution is extremely decreased. This is considered the main limitation of LCDM [63]. Low-contrast detail detect- ability can be improved using high-resolution LCDM and by utilizing the interactive adjustment of brightness and con- trast of digital images [64].

From the above discussion, utilizing correct image recon- struction algorithms and image processing applications im- proves the performance of LCD while maintaining the desired spatial and contrast resolution and the tolerated noise. Display monitors and visualizing conditions are also essential to acquire higher detectability performance.

Conclusion

The effects of low-contrast detail performance of different CT scanner systems have been discussed in this article. The impact level of the factors of contrast detail detectability on image quality is complex and does not exactly match from one type of scanner to another or from one unit to another. These factors are the ultimate key to optimizing image quality in terms of detail detectability, while utilizing lower doses.

Although the performance detectability within CT is in- herent to the system type and unit specification and cannot be controlled by radiographers, radiographers play an essential role to improve system performance and image quality by ef- fectively controlling and adjusting exposure factors. It is rec- ommended that radiographers have a greater understanding of the various CT scanners systems in order to improve the image quality while lowering radiation dose to the patient. Further studies of contrast-detail performance are required to further enhance the understanding of the influences of the exposure factors on image quality and radiation dose.

Acknowledgment

We acknowledge the cooperation of Julia Barrett, Radio- graphics, and Exxim Computing Corp., who gave us permis- sion to use their figures.

References

[1] Ledenius, K., Gustavsson, M., Johansson, S., Stalhammar, F.,

Wiklund, I. L.-M., & Thilander-Klang, A. (2009). Effect of tube cur-

rent on diagnostic image quality in paediatric cerebral multidetector

CT images. Br J Radiol 82, 313–320. [2] Fishman, E. (2007). Hot topics in CT. Appl Radiol 36, 4–9.

ging and Radiation Sciences 44 (2013) 62-70

[3] Kato, Y., Nair, S., & Sano, H., et al. (2002). Multi-slice 3D-CTA–an

improvement over single slice helical CTA for cerebral aneurysms.

Acta Neurochir 144, 715–722. [4] Peldschus, K., Herzog, P., Wood, S., Cheema, J., Costello, P., &

Schoepf, U. (2005). Computer-aided diagnosis as a second reader. Chest 128, 1517.

[5] Imai, K., Ikeda, M., Enchi, Y., & Niimi, T. (2009). Statistical charac-

teristics of streak artifacts on CT images: relationship between streak ar-

tifacts and mA s values. Med Phys 36, 492–499. [6] Miller, J. M., Rochitte, C. E., & Dewey, M., et al. (2010). Quantitative

and diagnostic accuracy of 64-MDCTA for segmental coronary artery

stenosis detection: results from the core-64 multicenter international

study. JACC 55, E757–E759. [7] Sun, J., Zhang, Z., & Lu, B., et al. (2008). Identification and quantifi-

cation of coronary atherosclerotic plaques: a comparison of 64-MDCT

and intravascular ultrasound. AJR Am J Roentgenol 190, 748–754. [8] Paul, N., Blobel, J., Kashani, H., Rice, M., & Ursani, A. (2010). Quan-

tification of arterial plaque and lumen density with MDCT. Med Phys 37, 4227–4237.

[9] Kalender, W., & Khadivi, K. (2011). Computed tomography: funda-

mentals, system technology, image quality, applications, (3rd ed.).

Erlangen: Publicis.

[10] Hayton A., Wallace A., Edmonds K., & Tingey D. (2010). Application

of the European DOSE DATAMED methodology and reference doses

for the estimate of Australian MDCT effective dose (mSv). In Proceed- ings of the Third European IRPA Congress 2010. Helsinki: ARPANSA.

[11] Mahesh, M. (2009). MDCT physics: the basics–technology, image

quality and radiation dose. Philadelphia: Lippincott Williams &

Wilkins.

[12] Goldman, L. W. (2007). Principles of CT: radiation dose and image

quality. J Nucl Med Technol 35, 213–225. [13] Seibert, J. A. (2004). Tradeoffs between image quality and dose. Pediatr

Radiol 34, 183–195. [14] Bath, M. (2010). Evaluating imaging systems: practical applications.

Radiat Prot Dosimetry 139, 26–36. [15] Cowen, A., Kengyelics, S., & Davies, A. (2008). Solid-state, flat-panel,

digital radiography detectors and their physical imaging characteristics.

Clin Radiol 63, 487–498. [16] Baker, M. E., Dong, F., & Primak, A., et al. (2012). Contrast-to-noise

ratio and low-contrast object resolution on full-and low-dose MDCT:

SAFIRE versus filtered back projection in a low-contrast object phantom

and in the liver. AJR Am J Roentgenol 199, 8–18. [17] Alsleem, H., & Davidson, R. (2012). Quality parameters and assess-

ment methods of digital radiography images. Radiographer 59, 46–55. [18] Wunderlich, A., & Noo, F. (2008). Image covariance and lesion detect-

ability in direct fan-beam x-ray computed tomography. Phys Med Biol 53, 2471–2493.

[19] Davidson, R. (2007). Radiographic contrast-enhancement masks in dig-

ital radiography. Thesis, University of Sydney. Available at: http://hdl.

handle.net/2123/1932. Accessed December 20, 2012.

[20] Suess, C., Kalender, W. A., & Coman, J. M. (1999). New low-contrast

resolution phantoms for computed tomography. Med Phys 26, 296–302.

[21] Zarb, F., Rainford, L., & McEntee, M. F. (2010). Image quality assess-

ment tools for optimization of CT images. Radiography 16, 147–153. [22] Verdun, F. R., Denys, A., Valley, J. F., Schnyder, P., & Meuli, R. A.

(2002). Detection of low-contrast objects: experimental comparison of

single–and multi–detector row CT with a phantom. Radiology 223, 426–431.

[23] Hernandez-Giron, I., Geleijns, J., Calzado, A., & Veldkamp, W.

(2011). Automated assessment of low contrast sensitivity for CT systems

using a model observer. Med Phys 38, S25. [24] Hamer, O. W., V€olk, M., Zorger, N., Feuerbach, S., & Strotzer, M.

(2003). Amorphous silicon, flat-panel, x-ray detector versus storage

phosphor-based computed radiography: contrast-detail phantom study

at different tube voltages and detector entrance doses. Invest Radiol 38, 212–220.

H. Alsleem and R. Davidson/Journal of Medical Ima

[25] Faulkner, K., & Moores, B. (1984). Noise and contrast detection in

computed tomography images. Phys Med Biol 29, 329–339. [26] Hsieh, J. (2009). Computed tomography: principles, design, artifacts,

and recent advances, (2nd ed.). Hobokin, NJ: SPIE.

[27] Seeram, E. (2009). Computed tomography: physical principles, clinical

applications and quality assurance, (3rd ed). New York: Saunders/

Elsevier.

[28] Hurlock, G. S., Higashino, H., & Mochizuki, T. (2009). History of

cardiac computed tomography: single to 320-detector row multislice

computed tomography. Int J Cardiovasc Imaging 25(Suppl 1), 31–42. [29] Bardo, E., & Brown, P. (2008). Cardiac multidetector computed to-

mography: basic physics of image acquisition and clinical applications.

Curr Cardiol Rev 4, 231–243. [30] Barrett, F., & Keat, N. (2004). Artifacts in CT: recognition and avoid-

ance. Radiographics 24, 1679–1691. [31] Hsiao, E. M., Rybicki, F. J., & Steigner, M. (2010). CT coronary

angiography: 256-slice and 320-detector row scanners. Curr Cardiol Rep 12, 68–75.

[32] Kitajima, K., Maeda, T., & Ohno, Y., et al. (2011). Capability of ab-

dominal 320-detector row CT for small vasculature assessment com-

pared with that of 64-detector row CT. Eur J Radiol 80, 219–223. [33] Khan, A., Khosa, F., Nasir, K., Yassin, A., & Clouse, M. E. (2011).

Comparison of radiation dose and image quality: 320-MDCT versus

64-MDCT coronary angiography. AJR Am J Roentgenol 197, 163–168. [34] van der Wall, E., de Graaf, F., van Velzen, J., Jukema, J., Bax, J., &

Schuijf, J. (2012). 320-Row CT: does beat-to-beat motion of the coro-

nary arteries affect image quality? Int J Cardiovasc Imaging 28, 147–151. [35] Romans, L. E. (2010). Computed tomography for technologists: a com-

prehensive text. Philadelphia: Lippincott Williams & Wilkins.

[36] Achenbach, S., Anders, K., & Kalender, W. A. (2008). Dual-source car-

diac computed tomography: image quality and dose considerations. Eur Radiol 18, 1188–1198.

[37] Flohr, G., McCollough, H., & Bruder, H., et al. (2006). First perfor-

mance evaluation of a dual-source CT (DSCT) system. Eur Radiol 16, 256–268.

[38] Barreto, M., Schoenhagen, P., & Nair, A., et al. (2008). Potential of

dual-energy computed tomography to characterize atherosclerotic pla-

que: ex vivo assessment of human coronary arteries in comparison to

histology. J Cardiovasc Comput Tomogr 2, 234–242. [39] EXXIM Computing Corporation. Cone-beam CT. Available at: http://

www.exxim-cc.com/cone-beam_ct.html. Accessed December 20, 2012.

[40] Gupta, R., Cheung, A. C., & Bartling, S. H., et al. (2008). Flat-panel

volume CT: fundamental principles, technology, and applications. Ra- diographics 28, 2009–2022.

[41] von Falck, C., Galanski, M., & Shin, H. (2010). Informatics in radiol-

ogy: sliding-thin-slab averaging for improved depiction of low-contrast

lesions with radiation dose savings at thin-section CT. Radiographics 30, 317–326.

[42] Toth, T. L. (2012). Image quality in CT: challenges and perspectives.

In: D. Tack, M. K. Kalra, & P. A. Gevenois (Eds.), Radiation dose from multidetector CT, (2nd ed.). (pp. 81–100) Berlin Heidelberg: Springer.

[43] Funama, Y., Awai, K., & Nakayama, Y., et al. (2005). Radiation dose

reduction without degradation of low-contrast detectability at abdomi-

nal multisection CT with a low–tube voltage technique: phantom study.

Radiology 237, 905–910. [44] €Ozg€un, A., Rollv�en, E., Blomqvist, L., Bremmer, S., Odh, R., &

Fransson, A. (2005). Polyp detection with MDCT: a phantom-based

evaluation of the impact of dose and spatial resolution. AJR Am J Roent- genol 184, 1181–1188.

[45] Ertl-Wagner, B. B., Hoffmann, R.-T., & Bruning, R., et al. (2004).

Multi–detector row CT angiography of the brain at various kilovoltage

settings. Radiology 231, 528–535. [46] Godoy, M. C. B., Heller, S. L., & Naidich, D. P., et al. (2010). Dual-

energy MDCT: comparison of pulmonary artery enhancement on ded-

icated CT pulmonary angiography, routine and low contrast volume

studies. Eur J Radiol 79, E11–E17.

ging and Radiation Sciences 44 (2013) 62-70 69

[47] Schindera, S. T., Nelson, R. C.,&Mukundan, S., et al. (2008). Hypervas-

cular liver tumors: low tube voltage, high tube current multi–detector row

CT for enhanced detectiondphantom study. Radiology 246, 125–132. [48] Marin, D., Nelson, R. C., & Barnhart, H., et al. (2010). Detection of

pancreatic tumors, image quality, and radiation dose during the pancre-

atic parenchymal phase: effect of a low-tube-voltage, high-tube-current

CT techniquedpreliminary results. Radiology 256, 450–459. [49] Zhang, C., Zhang, Z., Yan, Z., Xu, L., Yu, W., & Wang, R. (2011).

320-row CT coronary angiography: effect of 100-kV tube voltages on

image quality, contrast volume, and radiation dose. Int J Cardiovasc Imaging 27, 1059–1068.

[50] Huda, W., Scalzetti, E. M., & Levin, G. (2000). Technique factors and

image quality as functions of patient weight at abdominal CT. Radiology 217, 430–435.

[51] Rubin, G. D. (2003). 3-D imaging with MDCT. Eur J Radiol 45(S1), S37–S41.

[52] van der Molen, A. J., & Geleijns, J. (2007). Overranging in multisection

CT: quantification and relative contribution to dosedcomparison of four 16-section CT scanners. Radiology 242, 208–216.

[53] Theocharopoulos, N., Damilakis, J., Perisinakis, K., &

Gourtsoyiannis, N. (2007). Energy imparted-based estimates of the

effect of z overscanning on adult and pediatric patient effective doses

from multi-slice computed tomography. Med Phys 34, 1139–1152. [54] Beister, M., Kolditz, D., & Kalender, W. A. (2012). Iterative recon-

struction methods in x-ray CT. Phys Med 28, 94–108. [55] Winklehner, A., Karlo, C., & Puippe, G., et al. (2011). Raw data-based

iterative reconstruction in body CTA: evaluation of radiation dose

saving potential. Eur Radiol 21, 1–6.

70 H. Alsleem and R. Davidson/Journal of Medical Ima

[56] Marin, D., Nelson, R. C., Rubin, G. D., & Schindera, S. T. (2011).

Body CT: technical advances for improving safety. AJR Am J Roentgenol 197, 33–41.

[57] Yoo, S., Kim, Y., & Hammoud, R., et al. (2006). A quality assurance

program for the on-board imager. Med Phys 33, 4431–4447. [58] Bissonnette, P., Moseley, J., & Jaffray, A. (2008). A quality assurance

program for image quality of cone-beam CT guidance in radiation ther-

apy. Med Phys 35, 1807–1815. [59] Flohr, T., Stierstorfer, K., Ulzheimer, S., Bruder, H., Primak, A., &

McCollough, C. (2005). Image reconstruction and image quality evalu-

ation for a 64-slice CT scanner with z-flying focal spot. Med Phys 32, 2536–2547.

[60] Barnes, J. (1992). Characteristics and control of contrast in CT. Radio- graphics 12, 825–837.

[61] Yamaguchi, M., Fujita, H., Bessho, Y., Inoue, T., Asai, Y., &

Murase, K. (2010). Investigation of optimal display size for detect-

ing ground-glass opacity on high resolution computed tomography

using a new digital contrast-detail phantom. Eur J Radiol 80, 845–850.

[62] Warren, R. (1984). Detectability of low-contrast features in computed

tomography. Phys Med Biol 29, 1–13. [63] Samei, E., Ranger, N. T., & Delong, D. M. (2008). A compara-

tive contrast-detail study of five medical displays. Med Phys 35, 1358–1364.

[64] Bacher, K., Smeets, P., & De Hauwere, A., et al. (2006). Image quality

performance of liquid crystal display systems: influence of display reso-

lution, magnification and window settings on contrast-detail detection.

Eur J Radiol 58, 471–479.

ging and Radiation Sciences 44 (2013) 62-70