Procedures for Evaluation of Slice Thickness in Medical Imaging Systems

The main goal of a medical imaging system is to produce images to provide more accurate and timely diagnoses (Torfeh et al., 2007). In particular, Computed Tomography (CT), Magnetic Resonance Imaging (MRI), and Ultrasound (US) are resourceful tools in medical practice, and in many cases a life saving resource when rapid decisions are needed in the emergency room (Rehani et al., 2000). The above diagnostic techniques are based on the evaluation of high resolution images from technologically sophisticated equipment. Individual images are obtained by using several different electronic components and considerable amounts of data processing, which affect the quality of images produced and, consequently, make the diagnostic process more complicated (Torfeh et al., 2007). In this context, to guarantee a consistent image quality over the lifetime of the diagnostic radiology equipment and to ensure safe and accurate operation of the process as a whole, it is necessary to establish and actively maintain regular and adequate Quality Assurance (QA) procedures. This is significant for computer-aided imaging systems, such as CT, MRI and US. The QA procedure should include periodic tests to ensure accurate target and critical structure localization (Mutic et al., 2003). Such tests are referred to as Quality Controls (QCs). They hold a key role within the QA procedure because they enable complete evaluation of system status and image quality (Chen et al., 2004; Vermiglio et al., 2006). Importantly, QCs permit the identification of image quality degradation before it affects patient scans, and of the source of possible equipment malfunction, pointing to preventive or immediate maintenance requirements. Thus, image QC is crucial to ensure a safe and efficient diagnosis and treatment of diseases (Rampado et al., 2006; Torfeh et al., 2007). For this reason, periodic QCs have been recommended by manufacturers and medical physicists’ organizations to test the performance of medical imaging systems. Protocols for QCs and QA in medical imaging systems have been produced by several professional groups (AAPM – NEMA) (Goodsitt et al., 1998). This highlights the extensive role of QA programs, including QC testing, preventive maintenance, etc. (Rampado et al., 2006). In any clinical imaging study, it is important to have accurate confirmation of several physical characteristics of the medical imaging device. In particular, the slice thickness


Introduction
The main goal of a medical imaging system is to produce images to provide more accurate and timely diagnoses (Torfeh et al., 2007). In particular, Computed Tomography (CT), Magnetic Resonance Imaging (MRI), and Ultrasound (US) are resourceful tools in medical practice, and in many cases a life saving resource when rapid decisions are needed in the emergency room (Rehani et al., 2000). The above diagnostic techniques are based on the evaluation of high resolution images from technologically sophisticated equipment. Individual images are obtained by using several different electronic components and considerable amounts of data processing, which affect the quality of images produced and, consequently, make the diagnostic process more complicated (Torfeh et al., 2007). In this context, to guarantee a consistent image quality over the lifetime of the diagnostic radiology equipment and to ensure safe and accurate operation of the process as a whole, it is necessary to establish and actively maintain regular and adequate Quality Assurance (QA) procedures. This is significant for computer-aided imaging systems, such as CT, MRI and US. The QA procedure should include periodic tests to ensure accurate target and critical structure localization (Mutic et al., 2003). Such tests are referred to as Quality Controls (QCs). They hold a key role within the QA procedure because they enable complete evaluation of system status and image quality (Chen et al., 2004;Vermiglio et al., 2006). Importantly, QCs permit the identification of image quality degradation before it affects patient scans, and of the source of possible equipment malfunction, pointing to preventive or immediate maintenance requirements. Thus, image QC is crucial to ensure a safe and efficient diagnosis and treatment of diseases (Rampado et al., 2006;Torfeh et al., 2007). For this reason, periodic QCs have been recommended by manufacturers and medical physicists' organizations to test the performance of medical imaging systems. Protocols for QCs and QA in medical imaging systems have been produced by several professional groups (AAPM -NEMA) (Goodsitt et al., 1998). This highlights the extensive role of QA programs, including QC testing, preventive maintenance, etc. (Rampado et al., 2006). In any clinical imaging study, it is important to have accurate confirmation of several physical characteristics of the medical imaging device. In particular, the slice thickness

CT scanners
The slice thickness is evaluated by measuring the width (FWHM) of the image of one or more Aluminium ramps at the intersection of the ramp(s) with the scan plane (CEI, 1998), as illustrated schematically in Fig. 1. The sensitivity profile measures the system response to an attenuating impulse as a function of the z-axis position, through the slice plane. The sensitivity profile is a function of pre-and post-patient collimation, and appears as a blurred square wave. The FWHM of this blurred square is defined as the nominal slice width. The slice thickness of a CT scanner is determined by focal spot geometry as well as pre-patient and detector collimation and alignment. An customary method of monitoring equipment performance is to measure the parameters of interest using test objects. For example, Goodenough et al. proposed an approximate measure of beam width by using a series of small beads positioned across the beam width. However, this phantom is difficult to use for a precise quantitative measure because of the uncertainty on the alignment of the beads with respect to the beam. Indeed, the beam width can be measured directly from a beam profile plot only if care is taken to ensure that the Aluminium piece is oriented at 45 degrees across the width of beam (Judy et al., 1977, as cited in Goodenough et al., 1977.  (Philips, 1997).
A typical performance phantom (Philips, 1997) uses Aluminium plates slanted 26.565 degrees and which are across each other, as shown in Fig. 2. With X-rays irradiated to the Aluminium plates, the axial length of each Aluminium can be measured in CT image. With this method, it is possible to obtain an accurate measurement even if the intersection of the Aluminium plates is not aligned with the X-ray beam, by averaging the two measurements L a and L b . The slice thickness L is calculated as follows: CT image of the section dedicated to the measurements of slice thickness of a typical performance phantom: (a) layout of the measurement methodology; (b) schematic of resulting image with the centres of the four bars appearing as black spots (Philips, 1997).
Another performance phantom used for CT scanners (Fig. 3)  To determine the slice thickness, the image is displayed at the recommended window level and width. The number of visible holes (representing air-filled holes) is counted. Holes that appear black in the image represent a full millimetre slice thickness. Holes that appear grey count as fractions of a millimetre; two equally grey holes count as a single 1 mm slice thickness.

MRI scanners
The technique of MRI differs from X-ray CT in many ways, but one of the most interesting is perhaps that the slice is not determined primarily by the geometry of the scanning apparatus but rather by electronic factors, namely the spectrum of radio frequency pulse and the nature of the slice selection gradient (Mc Robbie et al., 1986). The slice profile and width of a 2D imaging technique such as MRI is a very important feature of its performance. Fig. 4. High signal ramp phantoms: (a) A typical slice-thickness phantom consisting of a set of crossed thin ramps. A ramp crossing angle of 90° yields an angle of 45° between the ramp and the image plane. The phantom length (L) should be greater than twice the maximum slice thickness. An alignment rod placed between the two ramps defines the point where the two ramps cross. When the slice is properly aligned through the intersection of the ramps the images of the ramps and of the rod will be aligned. (b) The slice sensitivity profile is directly proportional to the image intensity profiles if the image plane is perpendicular to the alignment rod (Price et al., 1990).
Since the signal obtained is directly proportional to the thickness of the slice, an inaccurate slice width can lead to a reduced Signal-to-Noise Ratio (SNR) (Lerski, 1992). Partial volume effects can significantly alter sensitivity and specificity. Quantitative measurements such as relaxation time T 1 and T 2 values, are also greatly influenced by slice thickness. Inaccuracies in slice thickness may result in inter-slice interference during multi-slice acquisitions, and invalid SNR measurements (Och et al., 1992). The slice profile, ideally rectangular, may contain side lobes which can produce very confusing effects (Lerski, 1992). In addition, gradient field nonuniformity, radio frequency field nonuniformity, nonuniform static magnetic field, noncoplanar slice selection pulses between excitation and readout, T R /T 1 ratio (where T R represents the repetition time), and radio frequency pulse shape and stimulated echoes can also affect the slice thickness accuracy (Price et al., 1990).
A variety of phantoms have been designed to evaluate slice thickness. All are some variation of an inclined surface. These may include wedges, ramps, spirals, or steps. A typically used phantom is the crossed high signal ramps. High signal ramp (HSR) phantoms generally consist of opposing ramp pairs oriented at a fixed angle  (Fig. 4). The HSR's should be thin (ideally infinitesimally thin) to quantify the slice profile accurately. In general, the thickness of a (90°) HSR oriented at 45° respect to the image plane should be < 20% of the slice profile FWHM (i.e., for 5-mm slice it is necessary to use a l-mm ramp) to obtain a measurement with < 20% error. The FWHM is the width of the slice profile (SP) at one-half of the maximum value. In this case, the SP should be obtained for each ramp. The FWHM then becomes where a and b refer to the FWHM of the intensity profiles measured for ramp 1 and ramp 2, respectively. Note that for =90° then Eq. 2 is simplified to: Example of CT scanner image obtained from the assessment of the EUROSPIN phantom (Lerski, 1992).
The EUROSPIN test phantom contains two sets of structures that may be used for slice profile and width measurement. Pairs of angled plates are used to obtain a direct measurement. The additional pairs of wedges are used to calibrate especially thin slices. Typical examples of images obtained for slice width are presented in Fig. 5(a and b). The dark bands on the left hand side of Fig. 5a represent a projection of the slice profile from the angled plates; the shaded region on the right hand side of the Fig. 5b represents the projection of the profile at the wedge.

US scanners
Ultrasound image resolution depends on beam width in the scan and elevation (section thickness) planes (Skolnick, 1991). On US scanners slice thickness evaluation or elevational resolution is useful to understand some of the problems due to partial volume effect. Section thickness is significantly more complicated to check. This characteristic of the ultrasound beam depends on the focusing effect in the elevation direction, which is perpendicular to the scanning plane (Richard, 1999). In linear, curved linear and phased array sector probes, focus is controlled electronically, but in the elevation plane it is determined mechanically by the curvature of the crystals. The beam in the scan plane can be sharply focused only in a narrow focal range. Thus, beam profiles in the scan plane are not indicative of beam profiles in the elevation plane. As with lateral and axial resolution, elevational resolution can be measured indirectly with anechoic spherical objects or cylindrical plug phantoms. Slice thickness focusing can also be evaluated qualitatively by scanning the anechoic cylindrical objects in an ultrasound QC test phantom with the scan plane along the lengths of the cylinders (e.g., perpendicular to the usual scan direction). Quantitative assessment can be achieved by using an "inclined plane" phantom ( Fig. 6) (Goodsitt et al., 1998). The methodology used with the inclined plane phantom consists in obtaining the elevation beam profile, finding the depth where the image is narrowest. By focusing on that plane, the thickness of the image is measured at the focal plane. This technique enables measurement of the elevation dimension of the beam only at a single depth. To determine the entire profile in the elevation plane, the probe must be moved horizontally along the surface of the phantom to make a series of measurements, with the beam intersecting the inclined plane at different depths (Skolnick, 1991). Slice thickness focal range, thickness and depth should be recorded on the US unit for each commonly used transducer. Any significant variation from those reference values may indicate a detachment of the focusing lens.

The novel procedure
A novel and versatile methodology is proposed here to determine the slice thickness accuracy using a novel phantom. The methodology can be applied to any image scanning technique, including CT, MRI and US scanning. The methodology consists of two steps: 1) acquisition of images of the phantom; 2) image elaboration by using the dedicated LabViewbased software. To test the proposed procedure and to obtain detailed information about the www.intechopen.com Modern Approaches To Quality Control 390 quality of the obtained results, the acquired and processed images were compared with those obtained by elaborating the same phantom images using commercial software following already validated procedures . The novel proposed dedicated phantom consists of a poli-methyl-methacrilate (PMMA) empty box (14.0 cm x 7.5 cm x 7.0 cm) diagonally cut by a septum at 26 degrees (Fig. 7). The PMMA septum is 2.0 mm thick and it divides the box into two sections, thus reproducing both single and double wedges. The two sections can be filled with the same fluid or with fluids of different densities. In particular, to determine the slice thickness accuracy the PMMA box was filled with two different fluids (water and air) for assessment in a CT scanner. To perform the same assessment with an MRI scanner, water was replaced with a CuSO 4 5H 2 O+H 2 SO 4 +1ml/l antialga (ARQUAD) liquid solution (T 1 =300 ms, T 2 =280ms). For US systems the upper wedge was filled with ultrasound gel, as conductive medium. In addition, a spirit level is used to verify the planarity of the phantom with respect to the beam and the patient couch. The test procedure followed for both CT and MRI devices consists of four steps: 1. Placing the slice thickness accuracy phantom in the scanner head holder. 2. Adjusting level position of the phantom if necessary. 3. Moving/positioning the phantom in the gantry aperture. 4. Scanning the phantom with a single slice using the desired slice width available The phantom images were acquired using standard Head and Body protocols, shown in Table 1 and Table 2 for CT and MRI medical devices, respectively. After, the phantom images were acquired, elaborated and analyzed, they were stored and/or transmitted to a printer. The stored ones were further transferred to a dedicated workstation, whereas the printed ones were acquired by the same workstation using a VIDAR Scanner, for the next elaboration. The methodology developed here for data elaboration, utilizes a purposely developed LabView-based slice thickness measurement software. LabView is a graphical programming language that uses icons instead of lines of text to create applications. In contrast to textbased programming languages, where instructions determine program execution, LabView uses data flow programming, where the flow of data determines execution (National Instruments [NI], 2003). The software is compatible with both non-standard and standard image formats (BMP, TIFF, JPEG, JPEG2000, PNG, and AIPD) (Vermiglio et al., 2008). To evaluate the slice width the FWHM of the wedge, expressed in pixels, is measured and calibrated with respect to the effective length of the PMMA box, expressed in mm. The result is displayed in real-time at the user interface, known as the software Front Panel (Fig.  8), with the advantage of providing a complete set of data with a user-friendly interface. By plotting the radiation profile obtained as a system response to an attenuating impulse, as a function of position (z axis), that is through the slice plane, it is possible to estimate the slice thickness accuracy of the acquired image utilising the developed software. This is referred to as the sensitivity profile. Gaussian smoothing is applied to smooth out the sensitivity profile and permit a clearer estimate of the desired FWHM. To evaluate the slice thickness , in real time, the software utilises the following equation: (4) To test the proposed procedure, results are compared with those obtained by elaborating the same phantom images using commercial software, in particular Image-Pro Plus software from Media Cybernetics (Sansotta et al., 2002;Testagrossa et al., 2006). The measurements presented here have been conducted on several CT and MRI devices in an extended study from 2006 to 2010. In addition, a statistical analysis was conducted on the resulting datasets to further validate the proposed methodology. The chosen statistical method is the variance analysis, through Fisher's exact test (F-test), to assess if a significant difference exists between datasets obtained following different procedures (C.A. Markowski & E.P. Markowski, 1990). The F-test is useful when the aim of the study is the evaluation of the precision of a measurements technique. In fact, the variance analysis consists in the factorisation of the total variance into a set of partial variances corresponding to different and estimated variations. The statistical analysis was conducted both on CT and MRI datasets. For CT scanners three different datasets of slice thickness measurements were considered. The first dataset of 16 measurements was done on a reference (RF) value of 10 mm The second dataset of 14 measurements, on a RF value of 5 mm. The third dataset of 10 measurements, on a RF value of 2 mm. The data was obtained using two different procedures. For MRI scanners slice thickness measurements were done on a RF value of 10 mm. In this case three different procedures were compared and 24 measurements in total were obtained on different MRI systems.

Results
The slice thickness measurements results using the novel proposed methodology and the statistical data analysis are presented in this section for CT, MRI and US systems.

CT medical devices
A two-dimensional image of the wedge of the dedicated phantom of Fig. 7 was acquired using the Standard Head Protocol described in Table 1. The X-Ray image of the phantom is presented in Fig. 9. In this case the line profile tool was available and the related trend was shown in the same figure. The results obtained by measuring the slice thickness accuracy with different CT scanners and by employing the in-house developed LabView (LV) program and the commercial Image Pro Plus (IPP) software are compared with the corresponding 10 mm RF value in  Table 3. Slice thickness accuracy results obtained with LV and IPP for a 10 mm RF value, using different types of CT scanners.
The data of Table 3 are presented in graphical form in Fig. 10. The slice thickness accuracy results (blue: IPP; red: LV) and deviation from the RF value (blue: IPP; red: LV) obtained using the IPP and LV procedures are presented for the 10 mm RF value. From the results presented in Table 3, it is observed that the mean values calculated for both the IPP and LV procedures, are comparable. However, the standard deviation obtained from the two different procedures is considerably different, with the LV procedure providing a narrower deviation, and hence more accurate results of the performed measurements.
In Table 4 Also for the 5 mm RF value case, the mean values obtained from LV and IPP datasets are comparable, with a slightly better estimate for the IPP procedure. However, the standard deviations obtained from the two procedures are significantly different. As in the previous case, the LV procedure provides a narrower deviation, thus enabling a more accurate measurement.
The data of Table 4 are reported in graphical form in Fig 11, where the slice thickness accuracy results obtained with the two procedures (blue: IPP; red: LV) and their deviation from the 5 mm RF value are presented.

www.intechopen.com
Procedures for Evaluation of Slice Thickness in Medical Imaging Systems 397 Finally, in Table 5, the results obtained by measuring the slice thickness accuracy by employing the IPP and the LV procedures are compared with the corresponding 2 mm RF value and mean values and standard deviation are also indicated. From the analysis of the data of Table 5, it can be observed that the mean value calculated by the LV dataset is significantly closer to the RF value than that calculated from the IPP dataset. This further supports the validity of the proposed technique.  Table 5. Slice thickness accuracy results obtained with LV and IPP for a 2 mm RF value, using different types of CT scanners.

RF
The data of Table 5 are represented in graphical form in Fig 12, where the slice thickness accuracy results obtained using IPP and LV (blue: IPP; red: LV) and their deviation from the 2 mm RF value (blue: IPP; red: LV) are presented. Statistical analysis conducted on the datasets shown in Tables 3-5 yielded the F-values reported in Table 6. Such F-values indicate that there is no significant statistical variation between the two different procedures, thus validating the methodology.
RF Values (mm) F C F T 10 0.011 4.54 5 0.332 4.67 2 4.18 5.12 Table 6. F values calculated for different RF values (F C ). These F-values were compared to the tabulated ones (F T ) for the P=0.05 confidence level.

MRI medical devices
Two-dimensional images of the wedge of the dedicated phantom were acquired using the Scan Head Protocol reported in Table 2. A typical MRI image of the PMMA box and the corresponding line profile calculated with the in-built MRI software are shown in Fig. 13. The preliminary results obtained using the IPP and LV procedures discussed above applied to MRI are reported in Table 7 for a RF value of 10 mm. In this case the slice thickness accuracy measured using an in-built MRI software (CS) is also included. The corresponding mean value and standard deviation of the three datasets are also reported. Also in this case, the mean values calculated from the three different datasets of measurements are comparable between them. However, whereas the standard deviations obtained from CS and LV procedures are comparable, those obtained from IPP and LV are significantly different. The LV procedure provides a narrower deviation with respect to that obtained with IPP, which gives evidence for a more accurate measurement. The slice thickness accuracy data of Table 7 are presented in graphical form in Fig. 14(a), where the slice thickness accuracy determined directly at the equipment console and that measured using the IPP and the LV procedures are compared. In Fig. 14  The F-test is then used to verify that the three datasets are comparable (Table 8). As evident from the data of Table 8, the F values calculated for the three sets of data are significantly smaller than that tabulated (F T ) for a P=0.05 confidence level in all cases. Therefore, there is no significant statistical difference between the three different procedures, thus validating the novel LV procedure proposed here also for MRI imaging.
Datasets F C F T CS-IPP 0.836 4.54 CS-LV 0.00165 IPP-LV 0.513 Table 8. Results of the F-test calculated comparing the CS, IPP and LV datasets (F C ). These Fvalues were compared to the tabulated ones (F T ) for the P=0.05 confidence level.

US scanners
The novel PMMA phantom proposed here (Fig. 7) was utilised also to test the elevation resolution in ultrasound systems. Preliminary measurements have been conducted on US probes to evaluate if it is possible to measure the beam width in the elevation plane. To correctly determine the elevation beam profile, it was necessary to slightly modify the phantom. In particular, the septum position was modified and, for the beam width evaluation, the inclined plane was oriented 28 degrees to the top and bottom surface, so the probe intersects the inclined plane at 28 degrees. The upper side of the inclined surface was filled with ultrasound gel, as conductive medium. The echoes reflected from the inclined plane are displayed as a horizontal band. Because the beam intersects the plane at 28 degrees, the beam width in the elevation plane can be calculated as follows: The above technique enables the measurement of the elevation dimension of the beam at a single depth only. To determine the entire profile in the elevation plane, the probe must be moved horizontally along the surface of the phantom to make a series of measurements, with the beam intersecting the inclined plane at different depths. With the use of the novel phantom described here, the resolution in the elevation plane is completely independent from the lateral resolution in the scanning plane. The new phantom proposed here is easy to use and does not require any additional equipment and the results are immediately displayed on the screen of the dedicated PC.

Conclusion
Slice thickness represents an important parameter to be monitored in CT, MRI an US. Partial volume effects can significantly alter sensitivity and specificity. Especially for MRI, quantitative measurements such as T 1 and T 2 , are also greatly influenced by the accuracy of the slice thickness. Inaccuracies in the measurement of this parameter may result in interslice interference in multi-slice acquisitions, leading to invalid SNR measurements. In addition, for the US scanners, significant differences in image resolution can occur because of variations in the beam width of the elevation plane. So, it is important to know the profiles of the elevation planes of various probes to choose the probe that offers resolution in the elevation plane that is most appropriate for the particular clinical application. Moreover, slice thickness accuracy is an important element also in the QC program for CT scanner, because quantitative CT analysis is heavily dependent on accuracy of slice thickness. Therefore, to provide an adaptable, reliable and robust procedure to evaluate the slice thickness accuracy a novel, dedicated phantom and corresponding LabView-based procedure have been proposed here, up to date applied to both CT and MRI devices. The new PMMA box proposed here to be associated with the dedicated software enables an innovative, accurate, easily applicable and automated determination of this parameter. The carried on studies have utilised the 2 mm septum because the z axis dimensions for rotating anode x-ray tube foci are typically less than this value, but is our intention reduce the septum width, in order to evaluate the slice thickness also for the helicoidal CT scanners, that can reach slice widths smaller than 2 mm. The accuracy (standard deviation) obtained with the novel procedure proposed here is significantly higher than that obtained with other procedures (e.g., in-built software, IPP). This new slice thickness accuracy procedure is proposed as an alternative to the commonly adopted ones, which are typically complicated by the use of ad-hoc software and phantoms distributed by manufacturers and specific to the medical equipment. The proposed method employs a novel universal phantom, coupled with a dedicated LabView-based software that can be used on any CT and MRI scanner in a quick, simple and reproducible manner. The readiness and applicability of the proposed procedure has been validated by quantitative tests using several different medical devices and procedures. In all cases the results obtained using the novel proposed procedure were statistically compatible with other commonly used procedures, but provided a very immediate determination of the slice thickness for both CT and MRI equipments, thus confirming the flexibility of the described method, its simplicity and reproducibility as an efficient tool for quick inspections. The same procedure should be suitable also for determining elevation accuracy on US scanner: in fact, preliminary results confirmed that the novel phantom, opportunely modified, when coupled with the LabView dedicated software, allowed measurements of the section thickness CEI EN 61223-2-6. (1997). Evaluation and routine testing in medical imaging departments. Phantoms for performance evaluation and quality assurance of CT scanners.