DIN 6868-161-2013 Image quality assurance in diagnostic X-ray departments Part 161 R V acceptance testing of dental radiographic equipment for digital cone-beam computed tomography.pdf

1、January 2013DEUTSCHE NORM English price group 14No part of this translation may be reproduced without prior permission ofDIN Deutsches Institut fr Normung e. V., Berlin. Beuth Verlag GmbH, 10772 Berlin, Germany,has the exclusive right of sale for German Standards (DIN-Normen).ICS 11.040.50; 37.040.2

2、5!%L 5. 4.3.7.2 Test Using the reconstruction of a homogeneous slice of the PMMA TEST PHANTOM, five equally sized fields from the corresponding data set (each approximately 2 to 3 % of the imaged phantom surface area) are to be defined. One field is to be in the centre of the TEST PHANTOM surface ar

3、ea, the other four fields are to be equidistant from the centre and at a distance of at least one-half of the imaged PHANTOMS radius. The arithmetic mean of the PIXEL VALUES in each field is to be recorded: Hc(centre), Hl(left), Hr(right), Ht(top) and Hb(bottom). NOTE 1 One possible distribution of

4、the fields is shown in Annex E.3. To calculate the large-area HOMOGENEITY INDEX, the maximum difference between the means for fields Hc, Hl, Hr, Htand Hb, and the average of the five field means, HM, is to be compared to the basic contrast maxmin (see Annex B) according to Equation (4): = maxminMax|

5、cM|,|bM|,|lM|,|rM|,|tM|(4) where: H is the HOMOGENEITY INDEX; maxis the average of the PIXEL VALUES in the row mmax(see Annex B); is the average of the PIXEL VALUES in the row mmin(see Annex B); Hcis the arithmetic mean of the PIXEL VALUES in the central field; Hlis the arithmetic mean of the PIXEL

6、VALUES in the left field; Hris the arithmetic mean of the PIXEL VALUES in the right field; Htis the arithmetic mean of the PIXEL VALUES in the top field; Hbis the arithmetic mean of the PIXEL VALUES in the bottom field; HMis the average of the values Hc, Hl, Hr, Htand Hb. DIN 6868-161:2013-01 15 NOT

7、E 2 By comparing the differences to the basic displayed contrast, the HOMOGENEITY INDEX remains independent of the chosen gray-level scaling for the tested systems. A larger value of H corresponds to better large-area homogeneity. 4.3.8 ARTEFACTS 4.3.8.1 Requirement No ARTEFACTS shall appear that in

8、terfere significantly with the diagnosis. Examples of possible ARTEFACTS are shown in Annex C. 4.3.8.2 Test The entire data set in the axial view (along the z-axis) shall be examined. NOTE Depending on the reconstruction algorithm, certain ARTEFACTS are unavoidable. The appearance of new ARTEFACTS c

9、an indicate malfunctions. DIN 6868-161:2013-01 16 Annex A (normative) Simplified determination of the MODULATION TRANSFER BEHAVIOUR The basis for the determination of the MODULATION TRANSFER BEHAVIOUR is a rectangular region of interest (ROI) selected in an axial slice of the reconstructed data set

10、of the TEST PHANTOM. The sides of this ROI shall be parallel to the pixel rows and columns. The ROI shall display only areas of the TEST PHANTOM containing PVC and air. The transition between these materials shall be depicted as an edge running parallel to one of the ROI sides. The parallelism of th

11、e edges can be optimized by repositioning the TEST PHANTOM and repeating the image acquisition. The number of PIXELS along the edge of the ROI shall correspond to 5 mm in the TEST PHANTOM (tolerance: 1 PIXEL). The number of PIXELS perpendicular to and on each side of the edge within the ROI shall co

12、rrespond to at least 3 mm in the TEST PHANTOM. NOTE 1 An example of an appropriate ROI is given in Annex E. The calculation of the MODULATION TRANSFER BEHAVIOUR shall be carried out using the following steps: a) Data acquisition Using the row (column) containing the PVC/air edge as the reference dir

13、ection, the PIXEL VALUES of all consecutive rows (columns) parallel to the edge and within the ROI are to be arithmetically averaged. The resulting mean values of the rows (columns) are to be numbered consecutively and recorded (1,2, M3, . . . ,n). b) Differentiation Adjacent mean values are to be s

14、ubtracted from each other to produce a series of differences 1, 2, 3, . . . ,1according to Equation A.1 m1mmMMD =+(A.1) where Dmis the difference of the average of the PIXEL VALUES from consecutive rows/columns; Mm+1is the average of the PIXEL VALUES in the pixel row/column m+1; Mmis the average of

15、the PIXEL VALUES in the pixel row/column m. c) Restriction to the transition region The differences 1,2,3, . . . ,n1contain one value, k, that is larger in absolute value than all the others. Beginning with this value, a symmetrical, adjacent data band, in which the differences still have the same s

16、ign, is to be determined: kl,kl+1 ,k, ,k+l1,k+l. DIN 6868-161:2013-01 17 NOTE 2 This constraint serves to separate the image of the edge from the relatively smaller differences that arise from the Heel effect and from noise. d) Fourier transformations If the number of difference values from step c)

17、is not already a power of 2, the set of values shall be padded to a power of 2 by adding zero values around the data band. The resulting series shall then be transformed using a discrete Fourier transform. The Fourier coefficients shall be normalized to their maximum value, and the first half of the

18、 resulting coefficients 0, 1, recorded. Similarly, the set of arithmetically symmetrized values 12|kl+ k+l|,12|kl+1+ k+l1|, , |k|, . . . ,12|kl+1+ k+l1|,12|kl+ k+l| to a power of two, discrete Fourier transformed and the resulting coefficients normalized to their maximum value. The resulting transfe

19、r coefficients 0, 1, mare to be provided for the spatial frequency range from 0 to the NYQUIST FREQUENCY . e) Averaging The arithmetic means p=12(p+ p) are to be recorded and assigned to their corresponding SPATIAL FREQUENCIES p= nm( 0,1, , m). NOTE 3 The pvalues serve as sampling points for the MOD

20、ULATION TRANSFER BEHAVIOUR in the spatial frequency range between 0 and the NYQUIST FREQUENCY . f) Graphical representation The value pairs (p;p) are to be plotted on a graph and joined with straight lines. The abscissa shows the values of the SPATIAL FREQUENCIES (p) on a linear scale. The ordinate

21、shows the values of the transfer factors (p) on a linear scale. DIN 6868-161:2013-01 18 Annex B (normative) Calculation of the CONTRAST-TO-NOISE INDEX The basis for the calculation of the CONTRAST-TO-NOISE INDEX is a rectangular region of interest (ROI) selected in an axial slice of the reconstructe

22、d data set of the TEST PHANTOM. The sides of this ROI shall be parallel to the PIXEL rows and columns. The ROI shall display only areas of the TEST PHANTOM containing PVC and PMMA, and the transition between these materials shall be depicted as an edge running parallel to one of the ROI sides. The n

23、umber of PIXELS along the edge within the ROI shall correspond to 10 mm in the TEST PHANTOM (tolerance: 1 PIXEL). The number of PIXELS perpendicular to and on each side of the edge within the ROI shall correspond to at least 3 mm in the TEST PHANTOM. The calculation of the CONTRAST-TO-NOISE INDEX is

24、 to be carried out using the following steps: a) Data acquisition The PIXEL VALUES in the ROI are to be read out in consecutive rows (columns) parallel to the PVC/PMMA edge. The mean values (1,2,3, . . . ,n) and standard deviations (1,2,3, . . . ,n) of the rows (columns) shall be numbered and be rec

25、orded. b) First differentiation The first differentiation uses a rolling mean over neighbouring values. The differences 5,6, . . .,44are calculated following Equation B.1. m=15(m+4+ m+3+ m+2+ m+1+ m) 14(m4+ m3+ m2+ m1) (B.1) where Pmis the mean of the PIXEL VALUES in the row (column) m; mis the diff

26、erence between the mean of Pm+4, Pm+3, Pm+2, Pm+1and Pmand the mean of Pm-4, Pm-3, Pm-2and Pm-1. NOTE The rolling mean serves to localize the PVC/PMMA edge reliably. c) Second differentiation The differences 6,7, . . ., 6are to be calculated according to Equation B.2. m= m+1m(B.2) where mis the diff

27、erence between m+1and m. From the series of P values, the indices maxand minof the largest and smallest values, mmaxand mmin, respectively, shall be determined. DIN 6868-161:2013-01 19 d) CONTRAST-TO-NOISE INDEX The CONTRAST-TO-NOISE INDEX is to be calculated according to Equation B.3. ( )2m2mmmminm

28、axminmax21SSPPCN)+= (B.3) where CNI is the CONTRAST-TO-NOISE INDEX; mmaxis the average of the PIXEL VALUES in the pixel row (column) mmax; mminis the average of the PIXEL VALUES in the pixel row (column) mmin; mmax2is the VARIANCE of the PIXEL VALUES in the pixel row (column) mmax; mmin2is the VARIA

29、NCE of the PIXEL VALUES in the pixel row (column) mmin. NOTE The PVC/PMMA edge generates two extreme values of the curvature in the series of averaged PIXEL VALUES (1,2,3, . . . ,). These are found during the second differentiation. The difference between the corresponding PIXEL VALUES, mmaxand mmin

30、, serves as measure for the CONTRAST in the image. The rolling mean in the first differentiation serves to stabilize the result. DIN 6868-161:2013-01 20 Annex C (informative) Possible ARTEFACTS in DENTAL DIGITAL CONE-BEAM COMPUTED TOMOGRAPHY (CBCT) C.1. General ARTEFACTS appear due to discrepancies

31、between the assumptions made in the calculation of the discrete three-dimensional volume data set and the actual, continuous object being imaged, or, rather, the physical measurement procedure on which the calculation is based. The following essential ARTEFACTS can be distinguished for DENTAL CBCT:

32、beam-hardening ARTEFACTS; extinction ARTEFACTS; partial volume effect and exponential edge-gradient effect (EEGE); aliasing ARTEFACTS; ring ARTEFACTS; ARTEFACTS caused by geometrical errors (for example, motion artefacts). C.2 Beam-hardening ARTEFACTS Beam-hardening ARTEFACTS are common in X-ray ima

33、ging due to the polychromatic nature of the spectra produced by X-RAY TUBES. The beams spectrum changes along its path due to differential absorption inside the materials of the object. Lower energy X-RADIATION (longer wavelengths) are absorbed preferentially inside the irradiated material, whereas

34、higher energy X-RADIATION (shorter wavelength) passes through the object without significant interactions, and is recorded on the detector. As a result, the radiation measured at the detector contains relatively more high-energy radiation compared to the radiation spectrum emitted by the X-RAY TUBE.

35、 In effect, the absorbing materials inside the object, especially those with high mass-attenuation coefficients (e.g., metals), act as a filter. When back-projected for the 3D reconstruction, these relatively higher energy areas can lead to typical striped artefacts in the reconstruction in the dire

36、ction of the absorption (radiation beam path), see Figure C.1. Although striped artefacts can also have other sources (see also C.3. extinction artefacts, for example), beam hardening is their main cause 17. DIN 6868-161:2013-01 21 Figure C.1 Typical striped beam-hardening ARTEFACTS in the direction

37、 of the beam path, caused by metal implants (titanium) in DENTAL CBCT C.3 Extinction ARTEFACTS Extinction artefacts appear specifically in 3D X-ray imaging when an object absorbs (nearly) all radiation completely, resulting in an insufficient signal measured at the detector 18. In DENTAL DIGITAL CON

38、E-BEAM TOMOGRAPHY, this is the case for very dense substances such as gold inlays or X-ray-opaque gutta-percha. During reconstruction, incorrect values are back-projected, which, in turn, leads to typical striped artefacts in the direction of the radiation (or absorption) beam path. Due to their com

39、mon physical origin, extinction artefacts appear in combination with beam-hardening artefacts (Figure C.2). Figure C.2 Extinction ARTEFACTS in the direction of the beam path due to very dense gold restorations in the teeth DIN 6868-161:2013-01 22 C.4 Partial volume effect and exponential-edge-gradie

40、nt effect (EEGE) These two ARTEFACTS are caused by the assumption of line integrals during the reconstruction, when, in fact, finite-sized “rays” are actually measured whose true dimensions are determined by the acquisition geometry, the pixel pitch of the detector and the focal spot size. Intensity

41、 errors are generated by absorption variations within a measured “X-ray” 19 (Figure C.3). Partial volume effects, such as the exponential-edge-gradient effect, typically lead to streak ARTEFACTS. Dark streaks surrounded by light ones (arrows) appear along sharp, high-contrast edge structures. Figure

42、 C.3 Intensity errors It has been shown that these effects always reduce the calculated intensity value 19, and that, therefore, streak ARTEFACTS appear in the volume tangentially to sharp edges 17. C.5 Aliasing ARTEFACTS Aliasing ARTEFACTS appear when the Nyquist theorem which states that the sampl

43、ing frequency must be at least twice as high as the highest object frequency to be sampled is not satisfied. One primary reason for too low a sampling frequency in DENTAL DIGITAL CONE-BEAM TOMOGRAPHY is the cone-shaped divergence of the RADIATION BEAM 20. This means that the number of “X-rays” that

44、traverse each VOXEL decreases linearly as one moves from the RADIATION SOURCE to the detector, effectively reducing the sampling rate nearer to the detector. In addition, an inaccurate interpolation between VOXEL and PIXEL VALUES during the back-projection can lead to further aliasing ARTEFACTS 21.

45、Subtle line patterns, so-called Moir patterns (Figure C.4), are the visual result of these aliasing effects. Aliasing ARTEFACTS can be reduced by better interpolation procedures during the reconstruction process 21, and by appropriate post-processing. DIN 6868-161:2013-01 23 Figure C.4 Subtle line p

46、atterns (Moir patterns) C.6 Ring ARTEFACTS Ring ARTEFACTS appear as subtle, circular patterns, arranged concentrically around the image of the rotation axis (Figure C.5). They are recognizable in those reconstruction slices that are oriented parallel to the radiation beam path (usually axial slices)

47、. Ring ARTEFACTS are attributed to the inhomogeneity of the flat-panel detector, combined with the VOXEL discretization 18, 22, 23, 24. They increase with increasing spatial resolution 23. The rebinning from polar to Cartesian coordinates in direct-inversion-based reconstruction 25 and the resulting

48、 sampling ARTEFACTS are likely to be the reason for their existence. They can be avoided or at least reduced, by means of calibration 18 or image post-processing 23, 24. In a broader sense, ring ARTEFACTS can also be classified as sampling ARTEFACTS. Figure C.5 Ring ARTEFACTS DIN 6868-161:2013-01 24 C.7 ARTEFACTS caused by geometrical errors (e.g., motion ARTEFACTS) Deviations of the actual acquisition geometry from the geometry assumed for the reconstruction process lead to an incorrect back-proj