CT image quality and factors affecting it by T.R.B.

All the image parameters of the CT images are discussed in brief.

1. CT IMAGE QUALITY THUMBA RAJ BARUWAL B.Sc. MIT Year Roll Number :162  

2. CONTENT Introduction Spatial resolution Slice sensitivity profile Reconstruction algorithm Contrast resolution Temporal resolution References

3. Introduction Image quality is a basic concept that applies to all types of images including photographic and video images as well as a wide variety of images produced for medical purposes, At the most fundamental level, image quality is a comparison of the image to the actual object. In CT, image quality is directly related to its usefulness in providing an accurate diagnosis. The usefulness of an image can only be assessed on a case-by-case basis. Image quality relates to how well the image represents the object scanned. However, the true test of the quality of a specific image is whether it serves the purpose for which it was acquired.

4. Image Quality Measurement Number of methods are available for measuring CT image quality, and principal characteristics are numerically assigned. These include; Spatial resolution Contrast resolution Temporal resolution Noise Linearity and Uniformity Depending on the diagnostic task, these factors interact to determine the ability to perceive low contrast structures and the visibility of details. These tools help make possible comparison of one imaging system to another or the same system over time.

5. Spatial Resolution Spatial resolution is another term used for detail resolution. Spatial resolution is the ability to distinguish two small high contrast objects located very close to each other under noise free condition. The spatial resolution of a CT image can be described in two dimensions. Resolution in the x-y direction is called in-plane resolution, whereas resolution in the z-direction is called longitudinal resolution.

6. Spatial Resolution Measurement Spatial resolution of an image can be measured in two ways; Direct method(by counting the line pair) By analyzing the spread of information within the system (known as MTF)

7. Spatial Resolution Measurement: Direct Method Using a line pairs phantom(made of acrylic and has closely spaced metal stripes) The phantom is scanned and the number of stripes are counted Line pair(line + space) If 20 lines are visible in 1-cm section of an image of the phantom, the spatial resolution is reported as 20 line pairs per centimeter( lp /cm) How frequently an object will fit into a given space is known as spatial frequency CTP714 30 Line Pair High Resolution Module

8. Spatial Resolution Measurement: Direct Method Image of a Catphan high-resolution insert

9. Spatial Resolution and spatial Frequency The absolute object size that can be resolved is equal to one half the reciprocal of the spatial frequency at the limiting resolution. The number of line pairs per unit length is called the spatial frequency.

10. Spatial Resolution Measurement: Modulation Transfer function Most common method Is a graphical representation of a system’s capability of passing information to the observer The MTF is the ratio of the accuracy of the image compared with the actual object scanned The MTF scale is from 0 to 1 If the image reproduced the object exactly, the MTF of the system would have a value of 1 If the image is blank and contain no information about the object, the MTF would be 0.

11. Modulation Transfer Function In graphic form, MTF( Y-axis) is plotted against the spatial frequency( object size X-axis) Shows that as the size of the object increases, the MTF also increases i.e. as the size of the object increases, it can be more accurately portrayed on the image

12. Modulation Transfer Function Comparison Comparison of MTF of two different CT systems. Limiting resolution is the spatial frequency possible on a given CT system at an MTF equal to 0.1. In this example scanner A will be better able to reproduce small objects than scanner B.

13. Modulation Transfer Function comparison An MTF curve that is higher at low spatial frequencies indicates better contrast resolution. Imaging system A has better contrast resolution, imaging system B has better spatial resolution.

14. Spatial Resolution Comparison Compared with conventional radiography, CT has significantly worse spatial resolution. The limiting spatial frequency for screen film radiography is about 8 lp /mm, for digital radiography it is about 4 lp /mm; the limiting resolution for CT is approximately 1.5 lp /mm. It is contrast resolution that distinguishes CT from other clinical modalities.

15. Factors Affecting Spatial Resolution Matrix Size, Display Field of View, Pixel Size Slice thickness Focal spot size Reconstruction algorithm Pitch Patient motion Sampling: detector size(aperture) and sample spacing( detector pitch)

16. Pixel size, FOV and matrix size Smaller the pixel size, better is the spatial resolution Transverse (in plane X-Y) resolution depends on pixel size DFOV is defined by the user based on anatomy to be displayed-DFOV<SFOV

17. Pixel size, FOV and matrix size DFOV determines how much raw data will be used to reconstruct the image. Changing the DFOV will also alter the size of the image on the screen. Increasing the DFOV increases the size of each pixel in the image. The pixel size reflects how much patient data is contained within each square. A larger pixel will include more patient data. The information contained in each pixel is averaged, so that one density number, or Hounsfield unit (HU), is assigned to each pixel. If an object is smaller than a pixel, its density will be averaged with the density of other tissues contained in the pixel, creating a less accurate image.

18. Sampling Theorem Patients are not composed of uniform density squares that can be trusted to fall neatly into separate pixels. Instead, an actual object may not lie entirely within a pixel. Sampling theorem (also called the Nyquist sampling theorem), used in telecommunication and many types of signal-processing applications, provides insight into the parameters necessary to accurately depict an object in a reconstructed image. This theorem accounts for the element of random chance in the creation of a CT image.

19. SAMPLING THEOREM can be roughly summarized by the following statement: because an object may not lie entirely within a pixel, the pixel dimension should be half the size of the object to increase the likelihood of that object being resolved. Nyquist criteria states that resolving N lp /cm requires measurement of at least 2 × N samples per cm.

20. SAMPLING THEOREM Random chance plays a role in whether a small object will be seen on the reconstructed image. In (A), (B), and (C) the object to be displayed is the same size as the pixel. The three figures show different scenarios as to how the object could be reconstructed, each resulting in a different level of volume averaging. In (D) and (E), a smaller pixel size is used, and the scenarios regarding the likelihood of volume averaging improve

21. SLICE THICKNESS All CT examinations are performed by obtaining data for a series of slices through a designated area of interest. In general, thinner slices produce sharper images because to create an image the system must flatten the scan thickness (a volume) into two dimensions (a flat image). The thicker the slice, the more flattening is necessary. The raw data are segmented in the longitudinal (head/ foot) direction by the slice thickness, representing a volume in the patient. Each segmented portion of data is called a voxel. The voxel size also plays a role in volume averaging, and hence affects spatial resolution.

22. SLICE THICKNESS The matrix divides data into squares with an x and y dimension. The operator’s selection of slice thickness accounts for the z axis segmentation and results in a rectangular solid. If the slice thickness is set equal to the pixel width, the voxel can have equal lengths on all sides (i.e., be a cube). All of the tissue within the voxel is averaged together to produce one CT number. A 2-mm object is contained in a 5-mm slice, resulting in significant volume averaging on the image. B. Slice thickness is decreased to 2.5 mm and volume averaging is significantly decreased.

23. SLICE THICKNESS MDCT scanning results in isotropic (or near-isotropic) voxels. When the imaging voxel is equal in size in all dimensions there is no loss of information when data are reformatted in a different plane, This is particularly important for imaging small vascular structures. An isotropic voxel ensures that there is no data loss with either multiplanar reformation (MPR) or volume rendering (VR)

24. Isotropic Spatial Resolution It is the resolution where the cross-plane resolution(Z) match that of the in-plane(X-Y) Since in most CT scans the pixel length is considerably smaller than the slice thickness, the reformatted scan can have an unusual appearance or stair-step artifact Modern MDCT for body imaging has isotropic resolution Advantages Creates MPR images with the same spatial resolution as the original sections Avoids the need for direct coronal scanning, reduces dose and scanning time

25. Slice sensitivity profile Is a graph that describes the slice thickness, Describe the effective slice thickness of an image and to what extent anatomy within that slice contribute to the signal. Conventional step and Shoot/Sequential CT , SSP Rectangular and width is equal to section width, but in spiral scan they are extended and more peaked.

26. Slice sensitivity profile Due to continuous movement of patient through gantry the data are displaced along z axis causing widening of slice sensitivity profile and appears more peaked. FWHM of SSP gives Effective slice thickness.

27. Slice sensitivity profile Increase in pitch widens SSP, Increase in pitch means movement is greater than the collimation , so there is increased slice distortion and increase in effective slice thickness Pitch greater than one- high speed mode , less than one- high resolution mode, 360˚ linear interpolation algorithm also widens slice sensitivity profile (replaced by 180˚ linear interpolation algorithm) SSP in 180˚ linear interpolation is reduced so allowed imaging at pitch greater than 1.

28. Focal Spot Size The focal spot size affects the image quality but the effect is minimal As in any X-ray imaging procedure, larger focal spot causes more geometric un-sharpness in the image and reduce spatial resolution Although focal spot size affects CT spatial resolution, CT resolution is generally limited by the size of detector measurements( referred to as the aperture size) and by the spacing of detector measurements used to reconstruct the image Increase in effective focal spot size decreases SR because of increased penumbra causing more geometric un-sharpness in the image Z-flying focal spot increases cross plane resolution by obtaining two overlapping slices for each detector row

29. FOCAL SPOT SIZE The focal spot size affects image quality, but the effect is minimal. As in any x-ray imaging procedure, larger focal spots cause more geometric un-sharpness in the image and reduce spatial resolution. CT system run at very high mA typically and this can increase the size of the x-ray, focus (Focal spot blooming)

30. RECONSTRUCTION ALGORITHM The appropriate reconstruction algorithm depends on which parts of the data should be enhanced or suppressed to optimize the image for diagnosis. Some will “smooth” the data more heavily, by reducing the difference between adjacent pixels. This can help to reduce the appearance of artifacts (such as those that result from dental fillings) but do so at the cost of reduced spatial resolution. Conversely, some filters accentuate the difference between neighboring pixels to optimize spatial resolution, but must make sacrifices in low-contrast resolution, These filters are most often used when there are great extremes of tissue density and when optimal low-contrast resolution is not necessary.

31. RECONSTRUCTION ALGORITHM Image (A) was reconstructed using a standard filter. The same data were reconstructed with a bone filter in image (B). Notice the increased spatial resolution.

32. Pitch Increasing the pitch reduces resolution in the image Lower pitch is slower( decrease temporal resolution) Higher pitch has more slice broadening( decreased spatial resolution) but radiation dose decreases Pitch< 1 implies oversampling Pitch>2 implies skipped data Can reconstruct an infinite number of images provided helical raw data is present

33. Detector Pitch (Spacing) In addition to small aperture, closely spaced measurements are required for good resolution For fixed FOV, as detector pitch decreases, number of rays increases and give better resolution When the view data exhibit a sequence of higher and lower attenuations, the image also exhibit bars and spaces that are separate but fewer than those actually in the test object. Such image is said to be aliased Because of insufficient sampling, the higher spatial frequency test pattern appears in the alias of low frequency pattern and thus is not truly resolved.

34. Sampling (Detector Width and Detector Spacing) Detector aperture size w idth of active detector in CT detector array Spatial resolution improves significantly in longitudinal direction as the detector aperture size decreases The thinner the detector, higher the spatial resolution and that has been the driving force in MDCT In plane (X-Y) spatial resolution is not affected by aperture size

35. Temporal Resolution The ability to resolve fast moving object in the displayed CT image Refers to how quickly data are acquired. Reported in ms. Gantry rotation speed of 330 ms of a specific 64-slice detector ( Somatom Sensation 64, Siemens Medical Solution, Germany) reports the temporal resolution as 83 to 165 ms Good temporal resolution avoids motion artifacts and motion induced blurring of the image

36. Temporal Resolution Controlled by gantry rotation speed, the number of detector channels in the system and with the speed with which the system can record changing signals A good temporal resolution in CT is realized by fast data acquisition(fast rotation of the X-ray tube) Can be improved further by using dedicated reconstruction algorithms(cardiac CT with a segmented reconstruction) or by using a dual source CT scanner

37. Factors affecting Temporal Resolution Gantry rotation time : decrease in gantry rotation time increases temporal resolution Number of detector channels: increases in detector channel in z direction increases temporal resolution Reconstruction method: single segment reconstruction method has less temporal resolution than multi-segment reconstruction method

38. Contrast Resolution The ability to distinguish one soft tissue from another without regard for size or shape is called contrast resolution CT is far superior in detecting low contrast differences because of : Scatter rejection by pre pt. and pre detector collimators Its consideration of the contribution of attenuation coefficient not only by atomic number difference but also by mass density difference Radiography can discriminate a density difference of approximately 5% where as CT scan detect density differences from 0.25% to 0.5% depending on the scanner

39. Contrast Resolution

40. Contrast Resolution

41. Contrast Resolution Low contrast resolution can be measured with phantom that contain low contrast objects of different sizes and with a small difference in density ( typically from 4 to 10 HU) from the background

42. Contrast Resolution 1 % contrast difference corresponds to a difference of 10 HU The ability to image low-contrast object with CT is limited by the size and uniformity of the object and by the noise of the system The contrast between structure and surrounding is only detectable if it is 3-5 times greater than the noise in the image Because the difference between object and background is small, noise plays an important role in low–contrast resolution Artificially increased by adding a contrast medium such as Iodine Increased photon energy reduces contrast for high atomic number lesions which contain Iodine than soft tissues

43. CONTRAST RESOLUTION Low-contrast resolution can be measured with phantoms that contain low-contrast objects of different sizes. The low-contrast performance or low-contrast detectability (LCD) of the scanner is typically defined as the smallest object that can be visualized at a given contrast level and dose There are three sets of disks with contrasts of 0.3%, 0.5%, and 1.0%, and the sizes of the disks are 2 mm, 3 mm, 4 mm, 5 mm, 6 mm, 7 mm, 8 mm, 9 mm, and 15 mm

44. FACTORS AFFECTING CONTRAST RESOLUTION Many factors affect contrast resolution. Many of these factors influence contrast through their relationship to image noise. mAs /dose Pixel size Slice thickness Gantry Rotation Speed Reconstruction Algorithm Patient size

45. mAs /Dose The mAs selected for scanning directly influences the number of x-ray photons used to produce the CT image, thereby affecting the SNR and the contrast resolution. Doubling the mAs of the study increases the SNR by 40%. Therefore, if the initial image was degraded by quantum noise then doubling the mAs will improve the contrast resolution of repeat scans. The dose increases linearly with mAs per scan. It follows that increasing mAs , will improve contrast resolution, but at the cost of a higher radiation dose to the patient.

46. mAs and Contrast resolution Illustration of the impact of noise to low-contrast detectability. A, Acquired with 200 mA and standard algorithm. B, Acquired with 50 mA and standard algorithm

47. Pixel Size Keeping all other scan parameters the same, as pixel size decreases, the number of detected x-ray photons per pixel will decrease. Fewer photons per pixel results in an increase in noise and a subsequent decrease in contrast resolution,

48. Slice Thickness The slice thickness has a linear effect on the number of x-ray photons available to produce the image—a 5-mm slice will have twice the number of photons as a 2.5-mm slice. Because thicker slices allow more photons to reach the detectors they have a better SNR and appear less noisy.

49. Slice Thickness and Contrast Resolution llustration of the impact of slice thickness, at 3.75 mm (top) and 7.5 mm (bottom).

50. Gantry Rotation Speed The x-ray source and detector array are moving relative to the stationary patient, both in the angular dimension and along the z-dimension for helical acquisition. Faster the gantry rotation speed lesser the contrast resolution.

51. Reconstruction Algorithm Bone algorithms produce lower contrast resolution (but better spatial resolution), whereas soft tissue algorithms improve contrast resolution at the expense of spatial resolution.

52. Reconstruction Filter Mathematical filter applied during reconstruction removes the blur from images Affects spatial resolution but requires tradeoffs depending on clinical needs Sharp – high spatial resolution but yields greater image noise Soft or smooth- reduces image noise but also degrades spatial resolution

53. Patient size For the same x-ray technique, larger patients attenuate more x-rays photons, leaving fewer to reach the detectors. This reduces SNR, increases noise, and results in lower contrast resolution.

54. WINDOWING Windowing is a technique of adjusting contrast and brightness of any digital image by manipulating parameters called window width and Window level. The Window is the range of CT numbers that will be displayed with the different shades of gray ,ranging from black to white. Center value of window width is called window level. In CT image , tissues are divided into 3 categories according to window setting Tissues with CT numbers lower than the lower window setting appear black and tissues with CT numbers greater than the upper window setting appear white. Tissues with CT numbers within the lower and upper levels appear in different shades of gray.

55. Window width and Contrast Resolution. Wider Window width setting decrease image contrast.

56. Noise It is the local statistical fluctuation in the CT numbers of individual picture elements of a homogenous ROI Even if we image a perfectly uniform object(a water filled object) there is still a variation in the Hounsfield Units about a mean. This is due to noise. The SD measurement of an ROI of a known uniform phantom will indicate the degree of noise in an image. The smaller the SD, the less the noise and the better contrast resolution capability Noise degrades the image by degrading low contrast resolution and introducing uncertainty in the HU of the images

57. Noise Cont.. CT no. are the average values i.e. pixel have a range of values greater than or less than CT no, these variations of pixel value represent image noise where , is each CT value, is mean CT value and n is the no. of CT values averaged    

58. Noise Cont.. The major type of noise include quantum mottle, structural mottle and electronic mottle The dominant source of image mottle is quantum mottle Characterized by a grainy appearance of the image Reducing the mAs is expected to increase the noise( measured standard deviation) by 1/√ mAs . Therefore, if the mAs is reduced by ½ then noise should increase by √2 =1.414 ( 4 0 % increase)

59. Noise Source The first source is the quantum noise determined by the x-ray flux or the number of detected x-ray photons. It is influenced by the scanning techniques (e.g., x-ray tube voltage, tube current, slice thickness, scan speed, helical pitch), the scanner efficiency (e.g., detector quantum efficiency, detector geometrical efficiency, amber-penumbra ratio), and patient (e.g., patient size, amount of bones and soft tissues in the scanning plane). The second source that influences the noise performance is the inherent physical limitations of the system. These include the electronic noise in the detector photodiode, the electronic noise in the data acquisition system (DAS), scattered radiation , parameters.

60. Noise

61. Linearity Refers to the relationship between CT numbers and the linear attenuation values of the scanned object at a designated KVp value CT no. should be consistently same for a particular tissue. Eg. for water=0 To check linearity , calibration should be done frequently by catphan or 5 pin performance test phantom Each of the 5 pins are made up of different plastic material having known physical and x-ray attenuation properties (polyethylene, polystyrene, Nylon, Lexon , plexiglas & Water)

62. Linearity Cont.. Graph of CT no. vs linear attenuation co-efficient should be straight line Deviation from linearity should not exceed +/- 5 HU over specific range (soft tissue or bone) Linearity is typically measured semi-annually CT linearity is acceptable if a graph of average CT number vs linear attenuation coefficient is a straight line that passes through 0 for water

63. Uniformity The CT no. measurement should not change with the location of selected region of interest (ROI) or with the phantom position relative to the iso-center of the scanner This characteristic of CT system is known as spatial uniformity It is most commonly measured using a water phantom For uniformity measurement, there should be no more than + 2 HU variation from an ROI placed at the center of the water phantom to those placed at the periphery These tests should be performed on a weekly basis

64. SUMMARY Image accuracy may also be referred to as image fidelity Slice thickness plays an important role in volume averaging, thereby affecting spatial resolution in the image. Image quality is closely linked to radiation dose. Improvement in image quality most often comes at a cost of increased radiation dose. At same KV and mAs , number of detected photons varies linearly with slice thickness. Thinner slice provides higher spatial resolution but increased image noise. Thicker Slice provides higher contrast resolution but poor spatial resolution

65. REFERENCES Computed tomography for technologists: A comprehensive text, Computed tomography : physical principles, clinical applications, and quality control / dr. Euclid seeram,fourth edition, Radiologic Science for Technologists: Physics, Biology, and Protection Stewart Carlyle Bushong , The essential physics of medical imaging JERROLD T. Bushberg Christensen's physics of diagnostic radiology,

66. THANK YOU