Manual
43 Pages
![8.0 Alpha = 0.0001, a priori diff of means = 0] 0.5 mAs 0.71 mAs 1.4Totals mAs for 0.71 mA mAs 4 2.00.71 mAs 0.71 mAs 5 0.71 mAs 6 0.71 mAs 7 0.71 mAs 2 0.71 mAs 3 Diameter (mm) Diameter (mm) 6.3 8.0 5.0 4.06.3 5.0 3.2 4.0 2.5 3.2 2.0 2.5 1.6 1.32.0 1.6 1.0 1.3 0.8 1.0 0.6 0.8 0.5 0.6 0.4 0.5 0.3 0.4 0.30.3 0.4 0.5 0.6 0.8 1.0 1.3 1.6 2.0 2.5 3.2 4.05.0 6.3 8.0 Depth (mm) 2 By Roeland van der Burght, Marianne Floor, Martin Thijssen1 and 0.3 0.4Remko 0.5 0.6 0.8 1.0 1.3Bijkerk 1.6 2.0 2.5 3.2 4.05.0 6.3 8.0 Depth (mm) IQFInv Scores Alpha = 0.0001, a priori diff of means = 0] Contrast Detail Score Diagram Alpha = 0.0001, a priori diff of means = 0] 6.3 6 5.0 4.0 5 3.2 2.5 4 2.0 IQFInv Diameter (mm) 7 8.0 1.6 3 1.3 1.0 2 0.8 0.6 1 0.5 0 2. 4 1. As m As m As m As m 71 0. 5 0. 0.4 0 0.3 0.3 0.4 0.5 0.6 0.8 1.0 1.3 1.6 2.0 2.5 3.2 4.05.0 6.3 8.0 Depth (mm) Contrast Detail Curve Contrast Detail Curve Alpha = 0.0001, a priori diff of means = 0] Alpha = 0.0001, a priori diff of means = 0] 8.0 Diameter (mm) 2.0 1.6 0.5 mAs 0.71 mAs 1.4 mAs 2.0 mAs 80 70 60 2.0 50 1.6 1.3 1.3 1.0 1.0 0.8 0.8 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 % Diameter (mm) 5.0 4.0 2.5 Alpha = 0.0001, a priori diff of means = 0] 8.0 Totals for 0.71 mAs 6.3mAs 4 0.71 0.71 5.0mAs 5 0.71 4.0mAs 6 0.71 mAs 7 3.2 0.71 mAs 2 2.5mAs 3 0.71 6.3 3.2 Total Detected (%) 40 30 20 10 60 6 3.2 2.5 Artinis Medical Systems Einsteinweg 17 6662 PW Elst The Netherlands www.artinis.com 50 5 0.8 0.6 0.5 0.4 0.3 4 CDQuality 2.0 IQFInv 3 30 20 1 10 0 0 Alpha = 0.0001, a priori diff of means = 0] 1,2 Dept. of Radiology, Radboud University Nijmegen, the Netherlands 80 70 60 50 40 30 20 Rev1402 - © 2014 Artinis Medical Systems As m As m Total Detected (%) 0 2. 4 1. As m As m As m As m As m As m Tel: + 31 481 350 980; Fax: + 31 842 105 702 71 0. 5 0. 0 2. 4 1. 71 0. 5 0. 0.3 0.4 0.5 0.6 0.8 1.0 1.3 1.6 2.0 2.5 3.2 4.05.0 6.3 8.0 Depth (mm) 40 2 % Diameter (mm) As m 7 5.0 4.0 1.0 0 2. Alpha = 0.0001, a priori diff of means = 0] 70 6.3 1.3 As m CDQuality Scores 8.0 1.6 4 1. Alpha = 0.0001, a priori diff of means = 0] As m IQFInv Scores Alpha = 0.0001, a priori diff of means = 0] 71 0. Contrast Detail Score Diagram 0 As m 0.3 0.4 0.5 0.6 0.8 1.0 1.3 1.6 2.0 2.5 3.2 4.05.0 6.3 8.0 Depth (mm) 5 0. 0.3 0.4 0.5 0.6 0.8 1.0 1.3 1.6 2.0 2.5 3.2 4.05.0 6.3 8.0 Depth (mm)](https://public.bioclinicalservices.com.au/library/representations/proxy/eyJfcmFpbHMiOnsiZGF0YSI6MTIzMTMsInB1ciI6ImJsb2JfaWQifX0=--3bc62eee96fa0429dc053da605354e6dce654aea/eyJfcmFpbHMiOnsiZGF0YSI6eyJyZXNpemUiOiIzMjB4In0sInB1ciI6InZhcmlhdGlvbiJ9fQ==--4214356078d1f110e449cda512a9432562da3e70/cdrad_2-0_phantom___analyser_manual_ver_2-1_feb_201.pdf.png)
Preview
Page 1
8.0
Alpha = 0.0001, a priori diff of means = 0]
0.5 mAs 0.71 mAs 1.4Totals mAs for 0.71 mA mAs 4 2.00.71 mAs 0.71 mAs 5 0.71 mAs 6 0.71 mAs 7 0.71 mAs 2 0.71 mAs 3
Diameter (mm) Diameter (mm)
6.3 8.0 5.0 4.06.3
5.0 3.2 4.0 2.5 3.2 2.0 2.5 1.6 1.32.0 1.6 1.0 1.3 0.8 1.0 0.6 0.8 0.5 0.6 0.4 0.5 0.3 0.4 0.30.3 0.4 0.5 0.6 0.8 1.0 1.3 1.6 2.0 2.5 3.2 4.05.0 6.3 8.0 Depth (mm)
2 By Roeland van der Burght, Marianne Floor, Martin Thijssen1 and 0.3 0.4Remko 0.5 0.6 0.8 1.0 1.3Bijkerk 1.6 2.0 2.5 3.2 4.05.0 6.3 8.0 Depth (mm)
IQFInv Scores Alpha = 0.0001, a priori diff of means = 0]
Contrast Detail Score Diagram Alpha = 0.0001, a priori diff of means = 0]
6.3 6 5.0 4.0 5 3.2 2.5 4 2.0
IQFInv
Diameter (mm)
7 8.0
1.6 3 1.3 1.0 2 0.8 0.6 1 0.5
0 2.
4 1.
As m
As m
As m
As m
71 0.
5 0.
0.4 0 0.3
0.3 0.4 0.5 0.6 0.8 1.0 1.3 1.6 2.0 2.5 3.2 4.05.0 6.3 8.0 Depth (mm)
Contrast Detail Curve
Contrast Detail Curve
Alpha = 0.0001, a priori diff of means = 0]
Alpha = 0.0001, a priori diff of means = 0]
8.0
Diameter (mm)
2.0 1.6
0.5 mAs 0.71 mAs 1.4 mAs 2.0 mAs
80 70 60
2.0
50
1.6
1.3
1.3
1.0
1.0
0.8
0.8
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
%
Diameter (mm)
5.0 4.0
2.5
Alpha = 0.0001, a priori diff of means = 0]
8.0 Totals for 0.71 mAs 6.3mAs 4 0.71 0.71 5.0mAs 5 0.71 4.0mAs 6 0.71 mAs 7 3.2 0.71 mAs 2 2.5mAs 3 0.71
6.3
3.2
Total Detected (%)
40 30 20 10
60 6
3.2 2.5
Artinis Medical Systems Einsteinweg 17 6662 PW Elst The Netherlands www.artinis.com
50
5
0.8 0.6 0.5 0.4 0.3
4
CDQuality
2.0 IQFInv
3
30
20
1
10
0
0
Alpha = 0.0001, a priori diff of means = 0]
1,2 Dept. of Radiology, Radboud University Nijmegen, the Netherlands 80 70 60 50 40 30 20
Rev1402 - © 2014 Artinis Medical Systems
As m
As m
Total Detected (%)
0 2.
4 1.
As m
As m
As m
As m
As m
As m
Tel: + 31 481 350 980; Fax: + 31 842 105 702
71 0.
5 0.
0 2.
4 1.
71 0.
5 0.
0.3 0.4 0.5 0.6 0.8 1.0 1.3 1.6 2.0 2.5 3.2 4.05.0 6.3 8.0 Depth (mm)
40
2
%
Diameter (mm)
As m
7
5.0 4.0
1.0
0 2.
Alpha = 0.0001, a priori diff of means = 0] 70
6.3
1.3
As m
CDQuality Scores
8.0
1.6
4 1.
Alpha = 0.0001, a priori diff of means = 0]
As m
IQFInv Scores
Alpha = 0.0001, a priori diff of means = 0]
71 0.
Contrast Detail Score Diagram
0
As m
0.3 0.4 0.5 0.6 0.8 1.0 1.3 1.6 2.0 2.5 3.2 4.05.0 6.3 8.0 Depth (mm)
5 0.
0.3 0.4 0.5 0.6 0.8 1.0 1.3 1.6 2.0 2.5 3.2 4.05.0 6.3 8.0 Depth (mm)
Contents 1. 2. 3. 4. 5.
Introduction ... 1 Description of the phantom... 1 Directions for use of the phantom ... 3 Phantom handling ... 3 Evaluation of the phantom image ... 3 5.1. Correction scheme ... 4 5.2. Correction examples ... 4 5.3. Presentation of the results ... 6 5.3.1. By using formula’s ... 6 5.3.2. The Contrast-Detail curve ... 7 6. CDRAD Analyser ... 9 6.1. Introduction ... 9 6.2. Resolving the phantom position ... 9 6.3. Resolving the centre position of the 225 contrast-detail combinations ... 9 6.4. Determination of the Phantom version (drill pattern) ... 10 6.5. Determination of the background signal ... 10 6.6. Determination of the spot signal ... 10 6.7. True/False allocation ... 11 6.8. Computation of the Contrast Detail curve ... 11 6.9. Multivariable contrast detail curve... 12 6.10. IQFInv and Total detected ... 13 7. CDRAD Phantom exposure... 13 8. Group analysis ... 14 8.1. Introduction ... 14 8.2. Defining Groups ... 14 8.3. Adding DICOM files to a Group ... 15 8.4. Group results ... 19 8.4.1. Computation of the Overall Contrast Detail curve ... 19 8.4.2. The IQFInv group result ... 19 9. Comparing different Groups of images ... 20 10. CDRAD Analyser commands ... 21 10.1. Program Menu ... 21 10.2. Right mouse button ... 26 10.3. Toolbar ... 26 11. Instructions and problem solving ... 27 12. System requirements and installation ... 28 12.1. System requirements, operating system and file types... 28 12.2. Installation and the use of the USB Key ... 28 13. CDRAD Analyser versions ... 29 13.1. Improvements implemented in version 1.1 ... 29 13.2. Improvements implemented in version 2.1 ... 29 14. Literature ... 29 Index ... 30 Appendices ... 31 A1. Score form of the CDRAD phantom ... 32 A2. Evaluation form of the CDRAD phantom ... 33 Software License Agreement ... 38 15. Warranty Policy Artinis Medical Systems. ... 40
1.
Introduction
Most definitions of image quality in radiology are based on characterizing the psychical properties of the imaging chain. However, medical diagnosis is not made by the image alone also the perception by the observer is crucit should be monitored. A test of the observer’s perception is possible with so called Contrast-Detail (CD) phantoms. With a CD-phantom it is possible to quantify the visibility of details at various contrasts, as observed by the radiologist. The CDRAD 2.0 phantom can be used within the entire range of diagnostic imaging systems, such as fluoroscopy and digital subtraction angiography. For mammography the CDMAM 3.4 phantom has been developed as a specific tool. The CDRAD phantom has been adapted from the Burger phantom1. With the phantom, the "threshold contrast" as a function of object diameter can be determined, and plotted in a Contrast-Detail curve. The phantom has been designed by the project team: Quality Assurance in Radiology Department of Radiology University Medical Center Nijmegen P.O.BOX 9101 NL-6500 HB NIJMEGEN The Netherlands 1988 - 1992 The use of digital systems makes it possible to automatically evaluate the images made from the CDRAD phantom. This enhances the use of this phantom as the time consuming task of manual evaluation is reduced significantly. By using the knowledge of different academic institutes 2,3 the CDRAD Analyser program was developed. In this manual, a description of the CDRAD-phantom is given in chapter 2, directions for use are presented in chapter 3 and the evaluation of the X-ray image is discussed in chapter 4. Chapter 5-12 describe the CDRAD Analyser software. Some literature is given in Chapter 13.
2.
Description of the phantom
The CDRAD phantom consists of a Plexiglas tablet (square 265 x 265 mm) with a thickness of 10 mm. The tablet contains cylindrical holes of exact diameter and depth (tolerances: 0.03 mm). In the tablet a grid (line pattern) has been engraved, which was treated with led-containing paint. The Xray image will show 225 squares arranged in 15 columns and 15 rows. In each square either one or two spots are present, being the images of the holes. The first three rows show only one spot, while the other rows have two identical spots in each square, one in the middle and one in a randomly chosen corner, to allow verification of the detection of each object. Easily recognisable patterns have been avoided. Figure 2.1 shows a schematic representation of the phantom. Within a row the hole-diameter is constant, with exponentially increasing depth, and within a column the hole-depth is constant, with exponentially increasing diameter (Table 1).
1
Figure 2.1. Schematic representation of the CDRAD-phantom Column
Depth [mm]
Row
Diameter [mm]
1
0.3
1
0.3
2
0.4
2
0.4
3
0.5
3
0.5
4
0.6
4
0.6
5
0.8
5
0.8
6
1
6
1
7
1.3
7
1.3
8
1.6
8
1.6
9
2
9
2
10
2.5
10
2.5
11
3.2
11
3.2
12
4
12
4
13
5
13
5
14
6.3
14
6.3
15
8
15
8
Table 1. Depth and diameter of the holes within the phantom
2
3.
Directions for use of the phantom
To make an X-ray image, the CDRAD-phantom should be positioned on the patient table over the cassette and automatic exposure control (AEC), in combination with one or more Plexiglas plates to fit the average patient thickness and examinations for that X-ray machine. A loaded cassette should be put in the table. Choices have to be made with regard to the exposure technique: -
Tube potential Focal spot size With or without grid Manual or automatic exposure
The density of the image has to be checked after the film has been processed. In a series of CDimages, all images should approximately have the same densities in a reference-position on the film (approximately 1.5 OD). Possible measurements with the phantom are: -
Comparison of image quality with various film-screen combinations. Determination of the optimum background density, by variation of this density. Determination of optimum exposure technique, e.g. by variation of tube potential. Comparison of image quality at various object thicknesses, by variation of the amount of Plexiglas at a fixed density. Influence of filtering by variation of the added filter thickness
For different simulated patient thicknesses different Plexiglas thickness should be added, preferably symmetrical over and under the phantom.
4.
Phantom handling
Store and use the phantom and its belongings at room temperature (15o-25o) and at normal humidity protected for fluid and moisture, dust, etc. preferable in the delivered case. Handle all products with care. PMMA scratches easily which might give that the phantom or plates is useless for evaluation. Clean the materials with non-aggressive general cleaner. The phantom cannot be used in MRI systems or in the neighbourhood of other magnetic materials.
Send back to manufacturer
5.
Evaluation of the phantom image
For best results the X-ray image of the CDRAD-phantom needs to be evaluated by at least 3 experienced observers. To increase statistics three independent images made at the same settings can be evaluated. The "Score form CDRAD-phantom" (see Appendix 1) can be used for this purpose. The image should be evaluated in the area where the holes are just visible, by indication of the location where the non-central holes are seen. At least 3 fields, covering at least one non-visible choice, must be observed in each column or row, in order to comply with the suggested correction
3
scheme, which is described in paragraph 5.1. The indicated positions of the eccentric holes have to be compared to the true hole-positions in the phantom, for which the "Evaluation form CDRAD-phantom" can be used (see Appendix 2). To evaluate the observations certain rules have to be applied, taking into account the 4 nearest neighbours (defined by a common vertice) of the field under examination. The evaluation of a particular field will always refer to the original observations for the nearest neighbours. Examples of the correction scheme are discussed in paragraph 5.2.
5.1. Correction scheme In the correction scheme, there are three possibilities for each observation: -
T: the eccentric hole was indicated at the true position F: the eccentric hole was indicated at a false position N: the eccentric hole was not indicated at all
The two main rules within the correction scheme are: 1. A True needs 2 or more correctly indicated nearest neighbours to remain a True. 2. A False or Not indicated hole will be considered as True when it has 3 or 4 correctly indicated nearest neighbours. Exceptions on the two main rules are: 1. A True which has only 2 nearest neighbours (at the edges of the phantom) needs only 1 correctly indicated nearest neighbour to remain True. 2. A False or Not indicated hole which has only 2 nearest neighbours will be regarded True if both nearest neighbours are correctly indicated.
5.2. Correction examples Six examples of the correction scheme are discussed below.
T N N N
T T* F* N
T T T N
T T T T
Example 1: The common situation. T* remains T because of its 2 correctly indicated nearest neighbours. F* remains F because it has only 2 correctly indicated nearest neighbours.
T N N N
T T* N N
T F* T* N
T T T T
Example 2: F* is considered T because it has more than 2 correctly indicated nearest neighbours. Both T*'s however have only 1 correctly identified nearest neighbour, and thus are considered to be F's.
4
T* N N N
T N N N
T T N N
T T T N
Example 3: T* remains T because it has 1 out of 2 correctly indicated nearest neighbours.
5
F* T* N N
T T N N
T T T N
T T T T
Example 4: F* will be considered as a T because of its 2 out of 2 correctly indicated nearest neighbours. T* will be considered as an F because it has only 1 correctly indicated nearest neighbour.
F* N N N
T* N T* N
T T N N
T T T N
Example 5: F* remains an F, because it has only 1 out of 2 correctly indicated nearest neighbours. Both T*'s are considered as F's because they have none respectively 1 correctly indicated nearest neighbour.
T* T N N
F* T T N
T T T T
T T T T
Example 6: T* remains T because it has 1 out of 2 correctly indicated nearest neighbours. F* will be considered as a T because of 3 correctly indicated nearest neighbours.
5.3. Presentation of the results 5.3.1. By using formula’s The curve through the threshold fields is called the Contrast-Detail curve4. The image quality can be expressed in a figure by calculation of the ratio of correctly identified hole-positions to the total number of squares (formula 1).
Correct observation ratio =
Correct observations x 100% Total number of squares
(1)
Another method to quantify image quality is called the Image Quality Figure (IQF)-method 5,6, which is defined in formula 2. 15
IQF = C i Di,th (2) i=1
where Di,th denotes the threshold (th) diameter in contrast-column Ci. Summation over all contrastcolumns yields the IQF.
6
For calculation purposes two extra rules apply: 1. A completely invisible column will result in a Di,th of 10.00 mm (for a hole depth between 0.3 and 8 mm). 2. A completely visible column will result in a Di,th of 0.3 mm (for a hole depth between 0.3 and 8 mm). Image quality increases with an increasing number of correctly identified hole-positions. In this case the IQF will become smaller because the values of diameter and depth of the threshold-holes are smaller. If another approach, giving an increasing figure for increasing image quality, is required formula 3 should be used.
IQFinv =
100
i 1CiDi ,th 15
(3)
Figure 5.1. Contrast detail lines of monitor image (--●---) and the hard copy image (---▲---) of the same DSA (Digital Subtraction Angiography) equipment 5.3.2. The Contrast-Detail curve The results can be presented in a graph, in which the hole-depth is plotted against the holediameter. The curve through the threshold fields is called the Contrast-Detail curve 4-6.
7
For comparison of the imaging performance of different systems, phantom images are made under identical conditions and evaluated by the same observer and at the same time. The better system will produce an image in which smaller contrasts and details are visible: producing a shift of the CDcurve to the lower left part of the image (Figure 5.1). Comparison of the performance of several observers is also possible. The better performing observer produces a CD-curve more to the lower left part of the image.
8
6.
CDRAD Analyser
6.1. Introduction The CDRAD Analyser analyses the images and applies a statistical method in order to determine whether a certain contrast-detail combination is detected or not. This statistical method uses the average pixel signal value and standard deviation of both the image of the contrast-detail combination under evaluation and its background pixels. In order to correctly determine these two variables it is necessary to correctly locate the positions of all the 225 different contrast-detail combinations of the phantom image. The program first determines the position of the contrast-detail combinations. Subsequently, by using a statistical method, it is determined whether a contrast-detail combination can be significantly detected or not. After this the program shows the results to the user. This chapter describes how these steps are performed. The evaluation of a single CDRAD image can be divided into seven steps: 1. Resolving of the position of the phantom 2. Resolving of the centre of the 225 contrast-detail combinations of the phantom 3. Determination of the phantom drill pattern type 4. Determination of the background signal 5. Determination of the relevant contrast-detail signal 6. True/False allocation 7. Computation of the Contrast detail curve Please read this chapter carefully as it provides the necessary background information for a correct use of the program.
6.2. Resolving the phantom position For automatic detection of the phantom position the program determines the four corners of the surrounding grid. First, it assumes that only the phantom is illuminated. This results in an almost black background. The program uses this information in order to determine the phantom edges. Subsequently, the phantom outline is determined by a search algorithm, which detects for all four sides the first lead engraved outline. Once these lines are determined, the four phantom corners and phantom position are known. If the program is not able to detect the four phantom corners automatically the user has to indicate the four phantom corners manually (see paragraph 10.1).
6.3. Resolving the centre position of the 225 contrast-detail combinations The CDRAD phantom consists of a Plexiglas tablet (square 265 x 265 mm). In the tablet a line pattern has been engraved, which was treated with lead-containing paint. The X-ray image will show 225 squares arranged in 15 columns and 15 rows. In each square either one or two spots are present, being the projections of the relevant contrast-detail combination. The position of the central spot (contrast-detail combination) in each of the 225 squares is accurately determined. This is needed in order to determine the pixel values of the images of both the centre and the second (corner) contrast-detail combination. The algorithm first determines the centre of the relevant square and then the centre of the central spot.
9
The following algorithm is used for this purpose: - Step 1: The four phantom corners are used in order to determine the approximate centre of each separate square. - Step 2: By using this centre the four sides of the square under evaluation are determined. Subsequently the new centre is calculated by using the middle of all four sides of the square under evaluation. - Step 3: The X-ray image is the projection of the phantom on the detector plate, which gives a deviation between the projection of engraved lines (on the surface of the phantom) and the image of the contrast-detail combination under evaluation. This deviation depends on the source-bucky distance of the relevant phantom image. Therefore a projection correction is executed on the centre found in step 2.
6.4. Determination of the Phantom version (drill pattern) For an accurate evaluation of the phantom it is of importance to take into account the projection of the second (corner) spot. This provides the underlying statistics with additional information, thus enhancing the result of the analysis. During manual evaluation one tends to memorize the positions of the corner spots. To avoid this Artinis delivers four different versions (drill patterns) of the CDRAD phantom. These versions differ in the location of the second spot. The CDRAD Analyser program is able to automatically determine the version of the phantom under evaluation. However it is also possible to manually prescribe one of the five drill patterns (see paragraph 8.1). The latter is advised for images where only some spots in the right upper corner are visible. For the automatic determination the program uses that part of the phantom where the spots are best depicted (e.g. depth 8 mm and diameter 4 mm). Once the drill pattern of the phantom is known the position of the second spot is calculated by using the specific version information. The used version of the phantom has to be either: - Drill pattern 1 (most common used) - Mirrored Drill pattern 1 (no longer available) - Drill pattern 2 - Drill pattern 3 - Drill pattern 4 For other versions of the phantom the software will not work.
6.5. Determination of the background signal The background signal is calculated using the red areas of Figure 6.1. As can be seen from Figure 6.1 the location of the two background areas depends on the location of the second spot (the white areas). The white and red areas form always a square. This method diminishes the influence of the heel effect; it is the same for both the spot signal as for the background signal. From these red areas the average value μbackground and the standard deviation σbackground are calculated.
6.6. Determination of the spot signal Paragraph 6.3 describes how the centre of the central spot is determined. For spots which have a nominal diameter of less than 3 pixels, an area of 3 x 3 pixels is taken in order to determine the spot signal. For the other spots all the pixels within the real spot diameter are taken. From both the pixels of the central spot and the second spot the average value μsignal and the standard deviation σsignal are calculated.
10
Figure 6.1. Determination of the background signal. White areas are the signal spots, the red areas are used for background signal determination.
6.7. True/False allocation After the mentioned steps are performed for each square within the phantom the information from Table 2 is available. Using the Welch Satterthwhaite method (Student t-test with Welch correction) the program determines if the contrast-detail combination in a certain square is positively seen. Signal
Background
Average value
μsignal
μbackground
Standard deviation
σsignal
σbackground
Number of observations
nsignal
nbackground
Table 2. Information for statistic tests Two statistical values can be adjusted by the user: - Alpha level of significance (Alpha) - A priori difference of means (APD) The program tests if the average signal level in a certain square is greater than the average background level plus an “a priory difference of means”. All squares where a significant difference between the signal and the background is found are marked with a red dot. The level of significance is defined by Alpha. The APD is set relatively to the image depth. For example if images with an image depth of 16 bits are compared with images with an image depth of 12 bits, an APD of 16 for the 16 bits images is the same as an APD of 1 for the 12 bits images.
6.8. Computation of the Contrast Detail curve The program comes up with the Contrast Detail diagram showing a red dot in all squares where the
11
contrast-detail combination under evaluation was detected (Figure 6.2). Within the same figure the CD curve (classic IQF score curve) is shown.
Figure 6.2. Detected Contrast-Detail combinations with classic IQF score curve The results are presented in a graph, in which the hole-depth is plotted against the hole-diameter. The curve through the threshold fields is called the Contrast-Detail curve 4, 5, 6. For the calculation of the Contrast-Detail curve a model based interpolation scheme 7 to fit a curve through the data is used. The curve is fitted through the data by using a least square procedure for each depth independently. A threshold at 50% detected contrast-detail combination is used for computing contrast detail curves.
6.9. Multivariable contrast detail curve Figure 6.3 shows the contrast detail curve in a simple model. The model shows that the contrast and the detail are detectable to a certain level. The detection is limited by the contrast (the Cline), by the detail (D-line) and by the combination of both (CD-line). The transition between the high-score and low-score region is not as sharp as in the model, especially not when comparing many files (see Figure 8.11). Instead the classic used IQF score curve as shown in Figure 6.2, we developed a multivariable CD-curve, which takes into account the slopes of the transitions. The multivariable CD-curve is defined as
where - f(x,y) is the CD-curve
12
-
a is the location of the C-line b is the location of the D-line c is the intercept of the CD-line of the y-axis f,g,h are the slopes of the transitions from high to low for the C-, D- and CD-line α is the angle of the CD-line with the x-axis
Switch between the classic IQF score curve and the multivariable curve in the settings (paragraph 10.1 and Figure 10.2).
Figure 6.3. Left: Theoretical boundaries of the contrast detail curve. The C-line is the boundary defining the minimal detectable contrast, the D-line is the boundary defining the minimal detectable detail. Right: Multivariable contrast detial curve in CDRAD Analyser, compare with Figure 6.2
6.10.
IQFInv and Total detected
Besides the Contrast Detail Curve, also the IQFInv and the Total detected (%) are calculated over the group. The IQFInv is calculated according to formula 3, using the contrast values determined by the contrast detail curve as input values. The Total detected (%) is as the correct observated ratio and calculated as in formula 1.
7.
CDRAD Phantom exposure
For a correct use of the program the CDRAD image has to fulfill the following criteria: When taking the image the CDRAD phantom has to be placed with its engraved line pattern and textures placed faced upward as seen from the X-ray tube. This means that the textures have to be readable when looking from the X-ray tube position to the phantom (View Position: AP). See step 3 in paragraph 6.3. - The source to bucky distance must be at least 30 cm. - For automatic detection of the phantom outline during exposure only the phantom should be illuminated. This results in an almost black background.
13
-
8.
The position of the phantom has to be such that the objects with the largest details (diameter of 8 mm) are on the top side of the image. Place the middle of phantom under the middle of the x-ray source. The phantom should not be rotated more than plus or minus 5 degrees from the above described position. The exposure within a cell of the CDRAD matrix should not deviate too much. The used version of the phantom has to be either: o Drill pattern 1 o Mirrored Drill pattern 1 o Drill pattern 2 o Drill pattern 3 o Drill pattern 4 For other versions of the phantom the software works inaccurately.
Group analysis
8.1. Introduction As for optimal quality control it is of importance to use more than one image. This will reduce the influence of noise on the results. The CDRAD Analyser allows to merge the results of more images into one so-called Group result. In order to do so CDRAD images have to be allocated to a certain Group. This chapter describes how Groups are defined and how CDRAD Images can be allocated to a certain Group.
8.2. Defining Groups A group can be defined (Figure 8.1) by either: - a right mouse click on the Project name - a right mouse click on one of the already existing Groups Rename the group by a right mouse click of the related Group and select “rename”. Remove a Group by selection of “remove”.
Figure 8.1. Defining groups (left) and defining, renaming and removing groups (right)
14
8.3. Adding DICOM files to a Group Once a Group has been defined CDRAD DICOM files can be added (Figure 8.2). DICOM files have the extinction “.DCM”. DICOM files can be added by either: - a right mouse click on the related Group. - a right mouse click on one of the already available DICOM files in a Group. - Add quickly images by dragging and dropping from a Windows folder to a group (Figure 8.3). Already added DICOM files can be removed or renamed (Figure 8.2). The program accepts DICOM format Monochrome 1 and 2, bitmap, jpeg and tiff files. Monochrome 2 files are inverted by the program. BE AWARE THAT DIFFERENT FILE FORMATS HAVE DIFFERENT CONTRAST DETAIL QUALITY PROPERTIES.
Figure 8.2. Adding DICOM files to a group Use only one type of files per group! Add more files to the group to get better results. Small detection deviations or bad pixels will be averaged out by using many pictures. Make new groups to compare different settings, file types or systems in one project. If you have problems to open a file, check the following: -
Is the file a DICOM file? The extension should be “.DCM”. Is the file compressed? Compressed files have to be decompressed before they can be analysed. Use a decompressing program to decompress the files.
Note: DICOM files can be dragged from one Group to another Group. Once a DICOM file has been added, the screen of Figure 8.4 will appear.
15
Figure 8.3. Drag and drop files to a group
Figure 8.4. Display of a single not analysed CDRAD image.
16
Figure 8.5. Analysed CDRAD image with the Contrast Detail Diagram After the red exclamation mark in the toolbar is activated the active CDRAD image will be analysed. Use Menu Analysis Batch run to analyse all images in the group. On the screen the following (Figure 8.5) will be displayed. Move the cursor over the image to show the related depth and diameter. The properties of an image file can be viewed by right mouse clicking at a file and selecting “properties” (Figure 8.6).
Figure 8.6. Image file properties – file properties The pixel size of the file is shown at the second tab of this window (Figure 8.7). This window gives the ability to use manual defined pixel size. In the case manual marking of the grid is necessary, it is often due to the projection and the pixel size has to be adjusted.
17
Figure 8.7. Image file properties - pixel size The third tab is for gray scale interpretation (Figure 8.7). Files with an inverted gray scale can be analysed by selection of “reverse gray scale interpretation”.
Figure 8.8. Image file properties - gray scale interpretation Often the properties of all files in a group have to be adjusted in the same way. Right click at the group and select properties (Figure 8.9). The set pixel size of the group is shown at the second tab of this window (Figure 6.10). This window gives the ability to use manual defined pixel size. We advise to analyse first one file and take the settings of this file to the group. After setting the properties of the group, the grids of the next files might be automatically detected without the need for manual grid marking.
18