PRODUCT 4.06 IMAGE MANAGEMENT

Transcription

1 IT EDUCTRA TELEMATICS APPLICATIONS PROGRAMME Sector: Healthcare PRODUCT 4.06 IMAGE MANAGEMENT Arie HASMAN This Product Section outlines the different methods of generating and analysing images, where each type is specifically utilised and the benefits and constraints to their use in particular circumstances. This Product addresses highly technical topics about which a non-specialist will only need to know the skeleton which is included in Product 2.02 of Workset 2, Diagnostic Methods. Within this overall workset are other complementary detailed Products relating to: the use of databases, spreadsheets and word processing in health (Product 4.01) use of multimedia in health (Product 4.02) electronic communications (Product 4.03) Internet use for health (Product 4.04) Image Management and signal processing (Product 4.06) Computer-aided diagnosis of the patient (Product 4.08) Reference Sources used in Health Product 4.12) Telemedicine (Product 4.13) Virtual Reality use in Health (Product 4.19) Expert systems and artificial intelligence in health (Product 4.24) In addition, the following graphics and pictorial What is a slides are particularly relevant to this Workset : multi-media in use (Product 3.02) Internet use for health (Product 3.04) image management and signal processing (Product 3.06) computer-aided diagnostic support (Product 3.08) strategic planning and modelling information (Product 3.14) virtual reality in health (Product 3.19) expert systems and artificial intelligence in health (Product 3.24) December 98 1

2 The user is referred to the listed SUGGESTIONS and Examples Products of Collection 1, for worked demonstrations and case studies relevant to diagnostic methods, notably: statistics file transfers networking and communications simulation and modelling expert systems image management and signal processing This Product reflects the full spectrum of Some details of... the imaging tools and techniques in health at this point in time (September 1997). It is a rapidly developing field, so the IT EDUCTRA Product user is advised to browse the contemporary trade press to see where the concepts have been rolled-out into in their area. The figures in this contribution were provided by Philips and Siemens. The AIMS of this Product section are to: to provide an outline of the technical concepts of medical imaging for the nonspecialist identify the areas in which medical imaging contributes to healthcare to highlight good practice in using medical images The Product outlines the science behind each of the methods but the Further References should be followed up if specific technical detail is required. This Product is one of a series of Products developed for the IT EDUCTRA project of the European Commission Telematics Application Programme (Health Sector). The overall index of the IT EDUCTRA portfolio is described elsewhere. The LEARNING OBJECTIVES of this Product At the end of this Product, the user should be able to: outline the scientific principles of medical imaging give in-depth descriptions of the use of images within healthcare delivery understand the impact of current medical imaging technology Keywords: ultrasound, radiology, image, scan, nuclear medicine, MRI, CT. December 98 2

3 The topics covered in this Product are: 1. Background and introduction 2. Ultrasound 2.1. Modes of operation 3. Radiology 4. Image generation by means of X-ray 5. Principle of digital subtraction angiography (DSA) 6. Computed tomography (CT) 6.1. Principles of computed tomography (CT) 7. Magnetic resonance imaging (MRI) 8. Nuclear medicine 9. Gamma camera 9.1. Histogram mode 9.2. List mode 9.3. Syncrhronised recordings 9.4. Processing of scintigrams 10. Two dimensional reconstruction 11. References for further reading 1. BACKGROUND AND INTRODUCTION Medical Imaging is a clinical service provided by radiologists. The medical imaging department of a hospital is usually lead by a radiologist. Images can be captured of all parts of the body. Dependent on the type of imaging, only certain tissues show up on the images. Traditionally, the images were taken and developed on hard copy like normal x-rays as a two-dimensional picture. Different forms of imaging can be discerned; ultrasound, x-rays, CT, MRI, nuclear medicine. In medical imaging, images of (parts) of organs are generated by means of (usually electro-magnetic (EM)) radiation. After image formation, the images have to be displayed for interpretation. The display medium can be the original carrier (for example traditional processed film) on which the image was formed or another carrier (for example a photograph printed from the original medium or a digital display on a computer). The organ under scrutiny may be generically referred to as an object in this Product. Not only electro-magnetic (EM) radiation is used for imaging, also ultrasound, light or electrical potentials. Visualisation methods can be used to produce images of phenomena that cannot be directly portrayed, such as the distribution of electrical potentials. In figure 1 the imaging chain, starting from the modalities up to December 98 3

4 the applications is shown. Another view of the imaging modalities is given in figure 2. In the following sections the principles of image formation by different types of EM radiation will be explained. First we will present ultrasound - an important example of non-em radiation - and after that radiology applications are discussed. Endoscopic imaging procedures, which use light, are not discussed in this product. Figure 1: The imaging chain in medicine (1) December 98 4

5 Figure 2: The imaging chain in medicine (2) 2. ULTRASOUND Ultrasound can be used to obtain both flow information and anatomical information. In figure 3 an ultrasound scanner is presented. Ultrasound is produced by piëzo-electric crystals that transfer electric energy into acoustic energy and vice-versa. In figure 4 a number of different transducers (containing the crystals) are shown. December 98 5

6 Figure 3: Example of an ultrasound device (echo scanner) These crystals can vibrate with frequencies in the order of 2 to 10 MHz, frequencies much higher than those present in audible sound (which are maximally 20 khz). The principles of ultrasound are similar to those of radar and sonar. Pulsed sound waves are emitted by the crystals and the strengths and arrival times of received echo's (which occur due to reflections at boundaries between tissues present in the path of the signal) are then measured. Because pulses and not a continuous wave are emitted the reflections (echoes) of the pulses can be detected. The time elapsed between the emission of a pulse and the detection of an echo is a measure of the distance between the source and the reflector that produced the echo. Since more reflectors (boundaries between tissues) may be present in the beam, more echoes may follow a single pulse. For the interpretation of the echo scan, it is assumed that the various boundaries are stationary during the time that the echoes are measured. December 98 6

7 Figure 4: Transducers used in ultrasound systems Echoes are only detected during a short interval, which interval therefore can be considered to have a duration almost equal to zero. Movements during this interval can therefore be disregarded. When a subsequent pulse is emitted some milliseconds later, the locations of the reflectors are detected again. If the reflectors have moved, this movement can be determined by comparing the elapsed times between two consecutive pulses and the corresponding echoes. A clinician is interested in the distance between the ultrasound source (the crystal) and possible reflectors (which could be a tumor, moving heart valves or similar entity). When suitably displayed, anatomical images of the investigated area can be obtained. It should be apparent that time is involved in two different ways in ultrasound. First the time elapsed between pulse and the corresponding echoes. Here the time is just a proxy for distance. This use of time should not be interpreted as representing time, but as representing depth. The time between the several pulses corresponds to the real elapsed time. The modes of operation are discussed in the following text. The detail of how the image is produced is described in the section about the principles of echo scanners. The distance between the source and the reflector can be calculated by dividing the elapsed time between the emission of the sound pulse and the detection of the echo coming from a reflecting boundary by twice the sound velocity (the distance is traversed twice, once by the original pulse and once by the echo). The accuracy of the distance measurement depends on the duration of the emitted pulse: the shorter this pulse the higher the accuracy. A pulse usually has a length in the order of one or two wavelengths. Reflectors that are positioned at least this distance apart can therefore be identified. The minimal distance along the axis of the beam between two reflectors that can still be identified is called the axial resolution. The cross-section of the beam emitted by the crystal source must also be small. The broader the beam, the more difficult it is to locate a reflector accurately (reflections from different parts of the boundary illuminated by the beam are now produced and a number of overlapping echoes are detected at the detector). December 98 7

8 However, the diameter of the crystal must be minimally ten to twenty times the wavelength of the signal emitted in order to obtain a directed beam MODES OF OPERATION There are four main ways of carrying out an ultrasound scan: Amplitude (A-mode) presents the intensity of the echo as a vertical excursion of the signal (similar to an EKG trace): the larger the vertical excursion the higher the intensity of the echo Brightness (B-mode): the intensity of the echo is presented as the brightness of a point corresponding the location of the reflector; the brighter the point, the higher the intensity of the corresponding echo Motion (M-mode) can be used for following moving structures and makes use of either the A- or B-mode. Compound (C-scan) allows re-positioning of the beam to produce more refined images Which of the different modes will be used depends on the application. Reflections (from boundaries between different tissues in the path of the beam) can be visualised in several ways. A distinction can be made between one and two-dimensional ways of visualisation. In a one-dimensional display, the information obtained from pulses going into the same direction are presented. In this case we have a fixed beam with a fixed direction. In a two-dimensional display, either the direction of the beam (which has to be narrow, as was explained earlier) changes or the beam itself is moved, so that we obtain information from different parts of the body. The following section describes details of each mode, which can be skipped if in-depth information is not required. The following modes are described: A-mode/B-mode M-mode C-mode Parallel and sector scan Principle of echo scanners A-mode/B-mode In the amplitude mode (A-mode) the strengths of the echoes resulting from a single pulse are displayed as a function of the time elapsed between pulse and each of the echoes. This elapsed time is equivalent to twice the distance between the reflector and the source. The amplitude of the echo is displayed along the vertical axis and the horizontal axis represents the time of arrival of the echo after emission of a pulse. This time is a measure of the location of the reflector. The December 98 8

9 image therefore contains a number of peaks and the location of the peak along the horizontal axis corresponds to the elapsed time (and therefore its depth) and its height represents the strength of the echo. Moving boundaries are visible as peaks that move horizontally along the time axis. Each separate pulse can give rise to a number of peaks (echoes) generated by reflectors in the body. The next pulse emitted somewhat later produces a similar image. If the reflector moved between the two pulses / in the mean time, the corresponding peak has shifted its position compared to its position in the previous pulse. The A-mode provides one-dimensional information about the location of the reflecting boundaries. The beam is fixed and only the echoes of those boundaries, which intersect the beam, are detected. In the B-mode (brightness) mode, the strength of an echo is represented as brightness. The B-mode is used in the M- and C-mode and in parallel and sector scans M-Mode The motion mode (M-mode) also produces a one-dimensional image, because here also the beam direction and location of the source are constant. This mode of ultrasound is used to visualise the movement of, for example, the heart valves. The motion mode (M-mode) is similar to the A-mode. In this case, the amplitude of each echo belonging to an individual pulse may be represented either as a vertical peak (A-mode) or as the brightness of the point located along the time axis corresponding to the arrival time of the echo (B-mode). The vertical axis is now used for displaying the echoes of subsequent pulses. In the M-mode of presentation, one co-ordinate therefore indicates the depth of the reflecting boundary within the patient and the other co-ordinate indicates the time elapsed after the first pulse, so that the movement of objects that are traversed by the ultrasound beam can be observed by following intense regions in the vertical direction (B-mode) or by following the peaks (A-mode) in the vertical direction. In this way, the movements of heart valves can be observed by following the corresponding points on the scan trace as they move along the time axis. When using more pulses, time therefore plays two roles. The time between the emission of a pulse and the corresponding echoes provides information about the depth of the reflectors. The subsequent following pulses similarly provide information about the depths of these reflectors, but these depths may have changed due to the movement of the reflectors. The time that elapses when more and more pulses are emitted therefore can be regarded as the real time. Although a certain time interval after each pulse is needed to detect echoes, this time interval is only used to determine the depth of the reflections. This time interval is small compared to the interval between pulses so that it can be assumed that the reflector does not move during the interval when the echoes are measured. Boundaries that are not perpendicular to the beam axis produce weak echoes. December 98 9

10 A-, B- and M-mode applications that have fixed beam directions therefore do not display all boundaries equally well. Some boundaries may hardly be detectable. This problem is solved by the C-mode C-scan The technology has been further developed to provide a so-called compound scan (C-scan) to compensate for the limitations identified previously: boundaries that are not perpendicular to the beam produce weak echoes. In a C-scan, the crystal can be moved so that the direction of the beam changes. The crystal is connected to a flexible arm, coupled to a fixed point of reference. The arm contains rotation transducers with which the position of the crystal and the direction of the beam always can be established. Since the crystals face different directions because of the movement of the crystal, a part of a certain boundary will probably be crossed by the beam in a direction perpendicular to it, thus producing a large echo. In this case a two-dimensional image is produced. Although the quality of a C-scan is high, a disadvantage is the time it takes to build up the image (of the order of many seconds) so that moving structures cannot be followed. Two-dimensional images can also be obtained by mechanically translating the beam or by using more crystals Parallel and sector scan. The movement of the crystal can be speeded up by mechanical means. In parallel scanning, the crystal is moved in such a way that the beam is continuously shifted along a certain direction until a whole area of interest is scanned. This procedure results in a parallel scan (figure 5). In another way of scanning, a socalled sector scan is obtained: each successive beam makes a small angle with the previous one (figure 6). In both cases, two-dimensional images are obtained in which the intensity of the image presents the amplitude of the echo (B-mode). Instead of working with a moving single crystal, a stationary probe containing a linear array of crystals (of the order of one hundred) can be used. When these crystals are electrically activated one after the other, a parallel scan can be produced. The data are usually stored in memory and can therefore be corrected and processed (distances between certain boundaries can for example be obtained or displacements of, for example blood vessels, can be measured). For the measurement of flow rates, use is made of the Doppler effect. When a continuous sound wave is reflected by a target that has a velocity component in the direction of the sound beam, the frequency of the reflected sound wave is higher (when the target approaches) or lower (when the target recedes from) the sound source. The shift in frequency (Doppler shift) is both proportional to the frequency December 98 10

11 of the incident beam and the velocity of the target. Since the frequency of the incident beam is known, the velocity of the target can be determined. In normal Doppler applications, a continuous stream of sound is generated by a transmitter and another transducer is necessary to detect the reflected sound waves. The distance of the target to the transmitter cannot be determined because of the continuous transmission of sound waves - as was explained earlier, distances can only be determined by measuring the time between the emission of a pulse and its echo. When a continuous wave is emitted, it is not possible to detect where the reflected beam comes from, since the reflected beam is also continuous and overlaps with the reflections of other boundaries. In modern equipment the Doppler shift can also be determined by pulsed sound. In that case only one transducer is necessary. By using sound pulses only the reflections from a small sample volume can be determined. By repeatedly repositioning the sample volume in a systematic way, a two-dimensional velocity distribution can be determined (flow map). Figure 5: Example of a flow map superimposed on an echo map (bifurcation of the aorta) December 98 11

12 Figure 6: Example of a sector scan (placenta previa) The flow map can be superimposed on the echo image (colour coded flow mapping). In this way flow information and anatomical information are presented simultaneously (see figure 5). Ultrasound applications have the advantage that they do not have harmful side effects. Therefore the echo-scan is an important technique for investigating the clinical condition of pregnant women and young children. Ultrasound is among others used for: the determination of the heart function brain investigations obstetric investigations eye investigations the determination of the perfusion of tissues the detection of tumors and cysts December 98 12

13 Principle of echo scanners In echo scanners, sound pulses are generated having frequencies in the order of a few megahertz (MHz). These pulses are absorbed, scattered or reflected within the patient. The reflections give rise to relatively strong echoes which can then be used to produce an image. Reflections occur at interfaces between media that are different with respect to density and/or sound velocity (sound is reflected at interfaces with differing acoustic impedances; the so-called acoustic impedance is equal to the product of sound velocity and density). Soft tissue and water have about equal density and sound velocity. Therefore most of the sound waves are transmitted through their interface without considerable reflection. At an interface between soft tissue at the one side and bone and air at the other a strong reflection is observed. Scattering instead of reflection takes place if the dimension of the object is small (relative to the wavelength of the incident radiation). The beam is then scattered in all directions and therefore the amplitude of the signal detected by the transducer is relatively small. The resolution of an echo scan (expressed as the smallest distance between two boundaries that can still be distinguished) is determined both by the wavelength of the sound waves and the duration of the emitted pulse. The pulse is usually a number of wavelengths long. In practice therefore, reflections from two points separated by a few wavelengths can be discriminated. The smaller the wavelength, the higher the resolution. Since the wavelength is inversely proportional to the frequency, the resolution is proportional to the frequency. Therefore structures which are close together (for example, in the eye) can only be visualised as separate structures using a high frequency, while they seem to be one structure when a lower frequency is used. The attenuation coefficient (expressing how much the beam is reduced in strength per cm of tissue because of scatter and absorption) is also proportional to the frequency (for soft tissue) or to the square of the frequency for other types of tissues. The depth of penetration of the sound waves is therefore inversely proportional to the frequency. The more the beam is attenuated (reduced in strength) the more difficult it is to measure the reflections of deeper located structures within the body (the signal-to-noise ratio becomes gradually smaller). Since resolution and penetration depth pose contradictory requirements, a compromise has to be made: deeper lying structures can only be visualised with relatively low frequencies with a concomitant lower resolution. Therefore deeper lying structures must be further apart in order to be able to distinguish them as separate structures. The type of tissue through which the signal is passing also influences the amount of the absorption of the beam. Air and bone are for example strong absorbers, attenuating the signal strongly, whereas muscle tissue and water hardly weaken the beam. December 98 13

14 At a frequency of 3 MHz (equivalent to a wavelength of 0.5 mm), depths up to 10 cm are visualised well with an axial resolution in the order of 1 mm. This means that structures that are further apart than 1mm can be visualised as separate entities. If they are closer than 1mm then they will be visualised as one structure. For investigations of the eye, a high resolution is needed because of the small dimensions of the eyeball. Frequencies between 5 and 13 MHz (wavelengths between 0.25 and mm) are used. For investigations of the brain, the sound beam first has to pass bone structures (for example - the tempora). In view of the high absorption capability of bone, the use of high frequencies in order to produce high-resolution images is infeasible. Therefore only low frequencies can be used and only a lower resolution image can be produced. 3 RADIOLOGY Roentgen discovered X-rays in X-rays have increasingly been used for diagnostic purposes. In this section, the ways in which computers have enhanced the capabilities of roentgen systems is discussed. The aspects are not addressed in a chronological order but rather by their increasing complexity of computer utilisation, starting with relatively simple applications (the same function but now supported by the computer) to complex ones (computers make new applications possible that cannot or are very difficult to be performed in a conventional traditional manual way). The relatively simple applications include computed radiology (CR) and digital subtraction angiography (DSA). The more complex applications include computed tomography (CT). Although not based on X-rays, a similar technology - magnetic resonance imaging (MRI) will also be discussed later : image generation by means of x-rays principle of Digital Subtraction Angiography (DSA) computed tomography (CT) 4. IMAGE GENERATION BY MEANS OF X-RAYS X-ray images are generated in a simple way. A beam of X-rays is generated by an X-ray tube. The cross-section of the beam is shaped by a diaphragm. The beam is directed at that part of a patient that is to be clinically investigated. At the opposite side of the patient, the transmitted X-rays are either absorbed by a fluoroscopic screen that transforms the X-rays into visible light or the X-rays expose a film (or film-screen combination). Also an image intensifier can be used which transforms an X-ray picture into a light image. December 98 14

15 The intensity of the beam is reduced as it passes through the various tissues in the body along a pathway between X-ray source and the receiver (film-screen combination or image intensifier). This process is called attenuation. Attenuation occurs because part of the X-rays are absorbed, but also because X-rays are scattered out of the beam. The more tissue that is traversed, the more the beam is attenuated. The attenuation depends on the type of tissue. The attenuation is determined both by the density of the tissue and its chemical composition. For example, lung tissue attenuates the X-rays much less then bone. Behind a rib the roentgen beam will have a much lower intensity than behind a space between the ribs. A silver halide film is too insensitive to X-rays to produce the potentially highresolution images that can be generated on such film by light. This low sensitivity was overcome by employing film-screen combinations as the receptor. These are sandwich-like structures with layers comprising one or two luminescent screens that transform X-rays into light and a silver halide film that is exposed by the light. The film is more sensitive to light than to X-rays. The X-ray picture is obtained after developing the silver halide film. December 98 15

16 Figure 7: Example of an image intensifier The intensity distribution of the transmitted beam can be transformed into an electrical signal with the help of an image intensifier (figure 7). A fluorescent screen in the image intensifier transforms the X-ray distribution into a charge distribution of electrons. These electrons are accelerated and then directed to a second smaller fluorescent screen on which an amplified light image appears. A TV or film camera then records this image. The TV signal can be used to digitise the image (figure 8). Images can also be generated on a storage phosphor plate instead of on a screen-film combination. The latent image stored on the phosphor plate can be read out by a laser scanner. No developing of films is necessary anymore and the data are directly in a digital format. Digital X-ray images obtained by this method are referred to as computed (CR) or digital radiology (DR) images. The images, once captured on film, can also be digitised with a laser film digitiser. December 98 16

17 Figure 8: Principle of image formation The above-mentioned techniques all produce so-called shadow images. On the film or TV-monitor we cannot discriminate the various structures or obstructions within the body that are successively passed by the beam. They all overlap on the resulting image. Also the contrast (differences in darkness and lightness in the image) is not always satisfactory to make, for example, blood vessels visible. The digital subtraction angiography (DSA) procedure makes use of shadow images but enhances the contrast and refines the image to highlight softer structures such as blood vessels, by subtracting bony structures out of the image. Computerised tomography (CT) enables the production of images of slices through the body. In this way neighbouring structures, that are not visible in shadow images, can be studied, because there is hardly any overlap within a slice, because of its small thickness. 5. PRINCIPLE OF DIGITAL SUBTRACTION ANGIOGRAPHY (DSA) It is usually not possible to visualise blood vessels in X-rays. Blood vessels often absorb as much radiation as the surrounding tissues and therefore cannot be discerned on X-ray images. By injecting contrast agents into the blood vessels, these vessels can be better visualised: the contrast medium contains iodine that absorbs X-rays relatively heavily. However, when bone is present, blood vessels still are difficult to perceive because the contrast differences around the blood vessels in that case are small compared to the contrast created by the bony structures. Since the eye is not able to detect contrast differences less then about December 98 17

18 3%, a procedure must be used that enhances contrast. Such a procedure is DSA. In the DSA procedure, a first X-ray shadow image (called pre-contrast image) is obtained with the help of an image intensifier and a TV camera. The signal coming from the TV camera is digitised and stored in the computer. Then contrast medium is injected into a blood vessel. Again a number of images are obtained and digitized. The pre-contrast image (also called the mask) is now digitally subtracted from the subsequent images. The resulting images will only contain structures that were not present in the mask. Since the contrast medium fills the vessels and the contrast medium was not present when the pre-contrast image was obtained, the resulting images will only show the vessels. Any bone structures, which would otherwise cloud the image and hide the vessels, are removed by the subtraction procedure. December 98 18

19 Figure 9: The principle of DSA The principle of contrast enhancement is presented in figure 9. In figure 9a the beam intensity along a line through the image in the pre-contrast image is presented. In figure 9b it is shown how the intensity distribution is altered after contrast medium is injected. Note the relatively small contrast differences that result. Usually these contrast differences are too small to be perceived. After subtraction of the mask, the intensity distribution shown is only due to the contrast medium in the vessels. The vessels now stand out as a result of the absorption caused by the contrast medium (fig 9c). The differences in intensity can be December 98 19

20 amplified in such a way that the eye is able to perceive the blood vessels in the image (figure 9d). In figure 10, a room with DSA equipment is shown. In figure 11, a DSA image is presented showing the bifurcation of the aorta. Note that on this image the spinal column and kidneys are hardly visible, due to the subtraction process. Figure 10: DSA equipment The quality of an image produced in this way deteriorates when the patient moves. The effects of such movements can be corrected to some extent through the use of suitable computer programs. These programs can shift or rotate the mask or a subsequent image in such a way that it maximally correlates one image with the next one so that movement effects / artifacts are suppressed. December 98 20

21 Figure 11: DSA image of the bifurcation of the aorta Since the contrast medium is injected in the form of a bolus in the first subtraction images, the proximal part of the vessel tree is usually visible, whereas in later subtraction images the distal part are more visible. With suitable computer software, containing enhancement functions, an image can be constructed that visualises the whole vessel tree. In addition to enhancing the images in this way, because the image is captured digitally in a computer, it is also possible to perform measurements on it. The dimensions of certain clinical details can be determined, such as: the extent of stenoses of vessels the movement of the heart muscle 6. COMPUTED TOMOGRAPHY Conventional roentgen images are projection (shadow) images that cannot reveal the real geometric distribution of tissues in organs. Moreover, organs that are situated behind each other are super-imposed in the image and a three dimensional volume is represented in two dimensions in the image. In order to obtain a three-dimensional impression, a number of images have to be available December 98 21

22 obtained from different angles. From these images, the radiologist or physician can form a mental picture of the three dimensional volume in which they are interested. Computed tomography (CT) is a means to obtain images that portray the real tissue distribution in a relatively thin single slice. Therefore, with CT a twodimensional image is obtained of a two-dimensional slice. Hounsfield introduced computed tomography in He was awarded the Nobel Prize in PRINCIPLES OF COMPUTED TOMOGRAPHY Consider a cross-section (with a thickness of the order of a few millimetres) through the body of a patient (referred to as a slice ). The cross-section can be divided into a large number of small squares (pixels) each with an area of the order of 1 mm 2. The idea behind CT is that if we can determine the absorption capability (also called absorption coefficient) of each pixel, we thereby will obtain information about the tissue distribution if the extent of absorption is characteristic of each tissue. In order to be able to measure the attenuation of a beam, a combination of an X-ray tube (source) and a detector is necessary. The absorbing slice is located in between them (see figure 12). The X-ray tube produces a pencil beam of known intensity and the detector measures the intensity of the resulting transmitted beam. The combination of X-ray tube and detector (mounted into a gantry) can both be translated and rotated. When a narrow beam of X-rays (a pencil beam) passes through the slice, each pixel traversed by the beam attenuates the beam to a certain extent. The amount of attenuation is determined by the tissue (in particular by its molecular composition and density) present in the pixel. The incident beam has a certain intensity at the start and the intensity of the finally transmitted beam will be smaller. Each pixel may attenuate the beam differently (due to the presence of different tissue). The intensity reduction of the beam is caused by all the pixels that the beam traversed. It cannot be deduced from one measurement of the intensity reduction how much each pixel separately attenuated the beam. Yet the determination of the attenuation by each pixel is the goal of the procedure since the idea, mentioned above, was that the extent of the absorption would be a characteristic of the tissue. Therefore it is expected that visualisation of the attenuation values as a function of location will produce a two-dimensional picture that will show the distribution of the tissues in the cross-section investigated. December 98 22

23 Figure 12: Principle of CT scanning Since one pencil beam only covers one row of pixels and since we need to determine the coefficient of all pixels, the beam is shifted over a distance equal to its width so that the attenuation coefficients of the pixels located in the next row can be taken into account. After measuring the transmitted intensity, the procedure is repeated by again shifting the beam and measuring the transmitted intensity until the total cross-section has been covered (see figure 12). Since, for each position of the beam, we only can measure the attenuation of the beam due to all the pixels passed, it is not possible to determine from these data what the individual attenuation coefficient of individual pixels is. For each position of the beam, only the total attenuation caused by all pixels traversed by the pencil beam can be determined. By plotting the distribution of this total attenuation as a function of the position of the beam, an intensity profile is obtained. Each point in the profile indicates how strongly the incident beam was attenuated by the row of pixels that was traversed by the beam in that position. It is known that a pencil beam is attenuated in an exponential way by the pixels encountered along the path. The attenuation due to a single pixel is determined by the length of the path traversed through the pixel and its attenuation coefficient. Assuming that the attenuation coefficient is constant over the whole pixel and representing this attenuation of pixel i by the attenuation coefficient m i, the intensity of the transmitted beam (I) can be related to the intensity of the incident beam (I o ) in the way presented in figure 13. If we take the natural logarithm of the ratio of I and I o we obtain the following relation for each beam position: N ln (I/I o ) = - d i m i i=1 In the formula d i is the length of the path that the beam traversed through each pixel i, and m i is the attenuation coefficient of the i th pixel and N is the total number of pixels in the beam. This produces an equation (see figure 13) relating one measured intensity ratio (left hand argument) to N unknown quantities. December 98 23

24 Figure 13: Attenuation of the beam in mathematical terms. These unknown quantities are the N attenuation coefficients of the pixels that were traversed by the beam. Since the direction of the beam is known, the length of the path - d i - traversed through each pixel can be calculated. A similar equation can be generated for each beam position (for each point of the intensity profile). The number of unknown quantities therefore only increases. After the beam has covered the whole cross-section, the total number of equations contains as many unknowns as there are pixels in the cross-section. Turning the gantry (holding the beam generating device) a small angle and repeating the earlier procedure results in another intensity profile. The attenuation coefficients of the pixels remain the same but the lengths of the paths traversed through each pixel change, because the direction of the beam changed. New equations with the same number of unknowns are therefore produced. By measuring the intensity profiles at enough angles it is possible, in principle, to obtain as many equations as there are unknowns. With the help of the computer the equations can be solved and the attenuation coefficient of each pixel in the cross-section determined. December 98 24

25 Figure 14: Example of a CT scan, with cross sections taken into 2 directions As can be seen from figure 14, a visualisation of the values of the attenuation coefficients by way of grey shades does indeed produce an anatomical image. In this case the CT consists of two parts of two mutually orthogonal slices. The slice in the upper part of figure 14 is the slice taken through one of the vertebra. More consecutive slices were obtained (not shown) and, from these slices, a slice perpendicular to them was calculated. The double line in the upper part of figure 14 shows the intersection. The slice in the lower part of figure 13 has the lowest resolution because the slices from which it was constructed all had a thickness of about half a centimetre. The process of CT as explained above is not actually used in practice because it would be too time consuming, but it provides a good insight in the principles of CT. One of the techniques that is used in practice to calculate the attenuation coefficients is back-projection. This technique can be used if intensity profiles are available covering the total cross-section at various angles. If we consider an individual profile, each point in the profile represents the amount of attenuation due to the pixels crossed by a line in the direction of the beam. In figure 15, the intensity profiles are shown that are obtained in a situation, where only one point absorber is present. Each profile shows a dip at the location where the beam passed December 98 25

26 through this absorber. If we have only one intensity profile we can not determine where on the path the absorber was located. We even cannot decide whether the absorption was due to a single point or due to a collective attenuating medium that was present over the whole path. The only inference we can make is that the attenuating medium was present only along one line in the cross-section, since the intensity profile showed a dip only at one point. Figure 15: Intensity profiles due to a point absorber and backprojection The back-projection method assumes that the absorbing medium is uniformly distributed over the line. Of course this assumption may be incorrect (as in this case where we have a point absorber) but errors resulting from this assumption can be corrected by computer program intervention. If several intensity profiles are available, obtained at different angles, a reconstructed image as shown in figure 15 can be produced. Instead of resembling the actual point attenuator, the reconstruction has a star-like distribution. By increasing the number of angles, the intensity in the centre will increase much faster than the intensity at the periphery. The more angles are used, the more the back-projected image becomes similar to the actual one. The back-projected image, however, always is a blurred version of the actual situation. This blurring effect can, to a certain extent, be corrected by using appropriate filtering techniques. This filtering results in a sharper image. December 98 26

27 A real cross-section through a patient can be considered as the sum of all cross-sections that each contain only one attenuating pixel. Therefore, the backprojection technique can also be applied to real patient cross-sections. It can be argued that, in the case of a point absorber, the location of this absorber can exactly be determined when two intensity profiles are available. The point absorber is located at the cross-section of two back-projected lines. The back-projection method however may not assume, that the attenuation is due to a single point absorber, because it has to be used in an environment where the real image contains a collection of many point absorbers. The back-projection technique can also be used in MRI (magnetic resonance imaging) and in SPECT (single photon emission computed tomography) as will be explained in subsequent sections. It appears that the attenuation coefficient is characteristic for the type of tissue (or more correctly, the chemical composition of the tissue) as is apparent from figure 14. When the values of the attenuation coefficients of the pixels are displayed on a monitor in the form of grey values, an anatomical image results that can directly be interpreted. Figure 16: Example of a CT scanner December 98 27

28 Since its introduction, CT equipment has gone through a number of generations: the first generation consisted of a combination of an X-ray tube generating a pencil beam and a detector for detecting the transmitted beam intensity. This combination of tube and detector was translated and rotated in order to cover the whole cross-section from several angles in second generation scanners, the single detector is replaced by multiple detectors and the X-ray tube produces a fan-shaped beam. The fan-shaped beam, however, still does not cover the whole patient cross-section and therefore, in second-generation scanners, translation and rotation remain necessary. Only the time for obtaining the intensity profiles was reduced significantly. in third generation scanners the detector array is so large that the transmission through the total cross-section of the patient can be measured simultaneously. Therefore no translation of the equipment is necessary. Only rotation of source and detector array is necessary to obtain enough profiles in fourth generation CT scanners there is a stationary detector array that covers 360 o. In this case only the X-ray tube rotates. In figure 16 an example of a CT scanner is shown. With the advent of X-ray tubes that can sustain an uninterrupted X-ray beam over a longer time period, it became possible to scan large volumes without having to wait in between. Before that time, the X-ray beam had to be interrupted because otherwise the tube would become too hot. By rotating the X-ray tube continuously and rapidly (making use of the so-called slip-ring power supply technology) and by slowly moving the patient table forward at the same time, in 50 seconds about 100 cm of anatomical length can be scanned. In this way, the X-ray beam traces a helical path over the patient (Figure 17). From the great amount of data that can be acquired in this way, cross-sectional slices can be reconstructed at any location. December 98 28

29 Figure 17: Principle of a helical scan 7. PRINCIPLES OF MAGNETIC RESONANCE IMAGING (MRI) Certain atomic nuclei (with an uneven number of protons and/or neutrons) rotate around their axis. This inherent rotation is also called spin. Because the atomic nuclei are charged, this rotation causes a current loop, and due to the current loop these nuclei behave like small magnets (see figure 18). This magnet has a certain direction and strength. It is called nuclear magnetic moment. December 98 29

30 Figure 18: Certain atomic nuclei behave like small magnets 1 H, 2 H, 13 C, 14 N, 15 N and 31 P are examples of biologically relevant isotopes, that possess a spin and therefore behave like a magnet. Under normal circumstances, the body is not magnetic in itself. This is due to the fact that the magnets point in all directions, so that the net magnetic strength (called the magnetisation) of the body is zero. When we place a collection of such nuclei in a strong magnetic field, the nuclei tend to align with the magnetic field, in the same way that a compass needle aligns itself to the earth s magnetic field. This alignment occurs because the nuclei prefer to be in a state with the lowest energy. At a temperature of 0 degrees Kelvin, all nuclei align themselves to the external field. At room temperature, the nuclei possess thermal energy which is large enough to enable their magnets to point into a direction opposite to the direction of the external field. This is comparable to a situation where a compass is shaken: the compass needle will now not always be aligned with the earth s magnetic field. The situation is different in that the nuclear magnetic moment, for reasons not discussed here, can only jump between two directions: parallel or anti-parallel to the external magnetic field. At room temperature the nuclei have such a high energy that there is only a small excess of nuclei pointing into the direction of the external magnetic field: in an external magnetic field with a strength of 0.1 Tesla for every million nuclei, one more points into the direction of the magnetic field than against it. The remaining nuclei have nuclear magnetic moments that in equal amounts point to both directions. Although one nucleus in a million is a very small number, a millilitre of water contains molecules, so that more than hydrogen atoms align themselves parallel to the magnetic field than in the opposite direction. These atoms together cause the magnetisation, because they all point into the same direction, whereas the magnetic fields of the other nuclei cancel out, because they point in equal amounts in both directions. When the external magnetic field strength increases, more nuclei will align themselves and therefore the December 98 30

31 magnetisation will increase (see figure 19). The magnetisation is characterised by its direction and strength. The direction of the magnetisation is parallel to the direction of the external magnetic field and its strength is both proportional to the strength of the external magnetic field and to the number of nuclei. Each nucleus will rotate around the external magnetic field as shown in figure 20. Figure 19: The strength of the magnetisation as a function of the external magnetic field and the temperature. Electromagnetic (EM) radiation is characterised by an electric and a magnetic component. The magnetic component can exert a force on the nuclei. When the magnetic component of the EM radiation has a direction perpendicular to the external magnetic field it can, under certain conditions, cause the magnetisation to progress around the direction of the external field in such a way that the angle between the direction of the magnetisation and that of the external field increases linearly with time. This only happens when the EM radiation has a certain frequency (called Larmor frequency). This frequency is proportional to the strength of the external magnetic field and to the so-called gyro-magnetic ratio, which is a characteristic of each isotope. The Larmor frequencies are in the range of the radio frequencies (2-50 MHz for different isotopes at a strength of 1 Tesla of the external magnetic field). The pulse of EM radiation that causes the magnetisation to progress to a direction orthogonal to the external magnetic field is called the 90 o RF excitation pulse (see figure 21). Since only one frequency will cause the precession of the nuclei of a certain isotope, it is a kind of resonance phenomenon. December 98 31

32 Figure 20: Rotation of a nucleus around the external magnetic field. Figure 21: Precession of magnetic moment due to EM radiation. After the termination of the RF pulse, the magnetisation will gradually precess back to the original equilibrium direction. From physics, we know that a changing magnetic field will induce a current in a coil (as happens in a dynamo). The size of the current is proportional to the strength of the magnetisation component along the axis of the coil. The frequency of the current will be equal to the precession frequency of the magnetisation. When a coil is placed perpendicular to the external magnetic field, the largest current will therefore be observed when the magnetisation also precesses perpendicular to the external field. December 98 32

33 Figure 22: Signal amplitude of the induced current as a function of time. When the magnetisation returns to its equilibrium position, the component along the axis of the coil will decrease gradually and so will the amplitude of the induced current. Since the initial value of the current in the coil is proportional to the strength of the magnetisation (when the magnetisation precesses perpendicular to the external field) and since the strength of the magnetisation is proportional to the number of nuclei, the initial value of the current is proportional to the number of resonating nuclei. The signal that is observed directly after the termination of the RF pulse is called the free induction decay (FID) signal (see figure 22). There is also a simulation of this principle The principle of nuclear magnetic resonance (NMR) imaging (MRI) can now be explained. The goal of the NMR imaging process is to provide an image of the tissue distribution in a plane through the body by measuring for example the hydrogen density in that plane. The objective is to obtain a two-dimensional image of a two-dimensional slice through the body, and the assumption is that the hydrogen density is indicative of the type of tissue. The density of hydrogen nuclei at each location of interest in the body can be obtained by giving the external magnetic field a strength that is different at each location of interest. The hydrogen atoms at each location thereby obtain their own specific Larmor resonance frequency, depending on the local strength of the external magnetic field. By irradiating the body with electromagnetic radiation of a certain frequency in a direction perpendicular to the external magnetic field, only those hydrogen nuclei having a Larmor frequency equal to the frequency of the RF excitation pulse will resonate. These nuclei are concentrated in a small volume if the strength of the external magnetic field varies strongly with position. The RF excitation pulse is given such a duration that, after the pulse, the magnetisation vector will precess perpendicular to the external magnetic field (a 90 o RF pulse). A current is now induced in a coil perpendicular to the external magnetic field, which has an amplitude proportional to the density of the resonating nuclei and a frequency equal to the Larmor frequency. The amplitude is proportional to the number of hydrogen atoms present in the sampled volume, whereas the frequency determines the position of the sampled volume. As was stated earlier, the Larmor frequency is determined by the strength of the external magnetic field. Since the strength of this field varies from place to place each location has its own specific December 98 33

34 resonance frequency. After the measurement, the frequency of the RF pulse is changed and the procedure repeated. Now the density of hydrogen nuclei at another location where the Larmor frequency is equal to the altered frequency of the RF pulse is measured. By repeating this procedure, the density of hydrogen nuclei can be determined at all locations of interest. Although such a procedure is possible, it would clearly take a lot of time to execute it. But it explains the principle. In an MRI system, the external magnetic field consists of two parts: a strong homogeneous field and smaller magnetic fields, the strength of which change linearly in a certain direction (also called a magnetic field gradient). The gradient is produced by currents that flow through gradient coils. In an MR imaging system there are three gradient coils: one for each of the three orthogonal directions. If the gradient is directed from head to toe every transverse (cross-sectional) slice through the patient resonates at a different Larmor frequency. Figure 23: Relations between external field gradient and Larmor frequency In addition RF coils are used to produce the RF signal and a receiving frequency coil is used to detect the resultant magnetisation. All nuclei in a slice orthogonal to the gradient direction will experience the same external magnetic field strength and therefore will have the same Larmor frequency (this gradient is therefore called the slice selection gradient (see figure 23)). The amplitude of the current in the receiving coil after the application of a 90 o RF pulses will be proportional to the total number of hydrogen nuclei in the particular slice. The real interest is in the distribution of the nuclei in this slice rather than in their total number. It is necessary to be able to determine the position more selectivity. If after the 90 o pulse, the slice selection gradient is switched off and another magnetic field gradient is applied orthogonal to the direction of the slice selection gradient (this additional gradient is called read-out or measurement gradient, because it is applied after the 90 o pulse and just before the induced current is measured in the receiving coil) then the frequency of the resonating nuclei will change: nuclei in the slice located along different rows orthogonal to the read-out gradient direction will experience a different external magnetic field December 98 34

35 strength (the sum of the homogeneous field and the read-out gradient field) and therefore will have different Larmor frequencies. This means that each row of hydrogen nuclei in the slice induces a current with a different frequency in the receiving coil. The amplitude of the component with a certain frequency is proportional to the number of hydrogen atoms along the corresponding row and the frequency determines the position of the corresponding row. Since the sum of signals with different frequencies is detected simultaneously, a Fourier transformation needs to be carried out to obtain the required frequency spectrum. With the help of the Fourier transform, the several frequencies that compose a signal can be determined. The procedure is reminiscent of the way a prism divides light into colours. Each frequency corresponds to a row in the slice (figure 24). Figure 24: Frequency spectrum. The height of each part of the spectrum corresponds to the number of nuclei in the corresponding row through the slice. The frequency spectrum therefore corresponds to a profile representing the number of hydrogen nuclei in each row as a function of the location along the direction of the read-out gradient. We now have a similar situation as in the case of CT. By measuring the profile for different directions of the read-out gradient and using a back-projection technique the distribution of the hydrogen nuclei over the slice can be obtained. The procedure can be speeded up by one step more. By applying a second field gradient orthogonal to and before the read-out gradient is applied, all positions along each row obtain a different phase: each point along the row precesses with a slightly different frequency due to the second field gradient (also called the phase encoding gradient). These phase differences alter the frequency profile (that is obtained after application of the read-out gradient). By applying phase encoding gradients with different strengths a number of profiles are obtained. Now the density distribution can be immediately obtained by performing a Fourier transformation twice. After termination of the 90 o RF pulse, the magnetisation gradually returns back to its equilibrium position, which is parallel to the external field. This phenomenon is called relaxation. The speed with which the magnetisation can realign itself with the external magnetic field depends on the easiness with which the nuclei can exchange energy with the surrounding nuclei (called the lattice). This so-called spin-lattice relaxation or longitudinal relaxation is characterised by the so-called longitudinal relaxation time T 1. T 1 is the time that is required for 63% of the nuclei December 98 35

36 to realign themselves with the external field (figure 25). The transverse component of the magnetisation, the component perpendicular to the external field, also decays as a function of time. The transverse component of the magnetisation will decrease because the magnetisation returns back to the direction of the external magnetic field. However, another mechanism - spin-spin interactions - makes the transverse component decrease faster than can be expected on the basis of the longitudinal relaxation. Figure 25: (Above) The strength of the magnetisation in the direction of the external field after a 90 o FR pulse. (Below) The strength of the component of the magnetisation perpendicular to the external field. The precessing spins create changing magnetic fields causing fluctuations in the precessional frequencies of neighbouring nuclei. Directly after the termination of the RF pulse all nuclei precess in phase, creating the transverse component of the magnetisation. Due to spin-spin interactions the nuclei de-phase because they all obtain slightly different frequencies, causing the transverse component of the magnetisation to decrease more quickly than could be expected from the longitudinal relaxation time (due to the differences in frequency, some of the magnetic fields of the nuclei have a transverse component that is cancelled by the transverse component of the magnetic fields of other nuclei close to them). This relaxation process is independent of the spin-lattice relaxation and is faster. The simulation can also show this action. Spin-spin relaxation is characterised by the transverse relaxation time T 2, measuring the time after which the transverse component has decreased by 63%. T 2 cannot be measured directly, because also in-homogeneities in the external magnetic field contribute to the de-phasing process. Therefore an effective T * 2 is measured, which is smaller than T 2 (see figure 26). The external field in-homogeneities exert a constant influence on the phase of the nuclei. By applying an 180º RF pulse some time after the 90º RF pulse, the December 98 36

37 individual spins are flipped and therefore come back into phase again, thus rebuilding the FID (see figure 27). This second signal is called an echo. The amplitude of the echo is smaller than that of the original FID signal, since the dephasing influence of the spin-spin interactions is irreversible and therefore cannot be compensated for by flipping the spins. The time between excitation and the maximum echo signal can be varied. The strength of the signal will be proportional to the hydrogen density and will depend on the transverse relaxation time (see figure 28). In this case, the contrast of the image is determined both by the hydrogen density and the value of T 2. Other techniques also enable us to visualise the influence of the T 2 values. Figure 26: Difference between the actually measured T 2 and the real T 2 December 98 37

38 Figure 27: Compensating for inhomogeneities in the external field by flipping the spins. Other factors that can influence the image are flow phenomena in blood or the cerebral spinal fluid. Flows can produce both signal enhancement (flow enhancement) or signal reduction (flow void). Magnetic resonance angiography (MRA) is based on signal enhancement. With MRA, arteries and veins may be selectively imaged without the need for contrast media. Figure 28: Measuring the real T 2 It is possible to use contrast agents - para-magnetic compounds such as gadolinium, manganese or iron, to enhance the contrast and to produce better visualisation of lesions in the brain, spine, breast and urinary tract. MRI can also be used for placing fiducial markers. This process is called December 98 38

39 tagging and is used in myocardial imaging. In this case, RF pulses are produced (triggered by the rising edge of the R-wave) in such a way that parallel planes perpendicular to the image plane are magnetised. If, directly afterwards, slices are imaged in the normal way, dark lines appear in the image where the tag planes intersect the image plane. The black lines are formed by the hydrogen nuclei that were already magnetised by the time that the second RF pulse was generated. The second RF pulse hardly effects these nuclei and therefore these nuclei are invisible in the image. By imaging cross-sectional slices through the heart at different time points through the systolic interval, any deformation of the heart wall can be studied by following the position of the dark lines. 8. NUCLEAR MEDICINE In nuclear medicine investigations, radioactively labelled material (a radiopharmaceutical) is injected into a patient. The choice of the material depends on the type of investigation to be performed. For example, when investigating thyroid disorders, an iodine compound can be used that is preferentially taken up by the thyroid, whereas erythrocytes labelled with technetium are applied in vessel investigations. The radioactive isotope is chemically bound to a carrier that is chosen because it is taken up by the organ of interest. The radioactive isotope is used for the creation of the image. The radiation emitted by the isotope is partly absorbed inside the patient, the remaining part can be detected by a detector, called a gamma-camera. Usually a computer system is connected to the gamma camera. The data coming from the gamma camera are used by the computer to produce and enhance images. Radioactive isotopes disintegrate to stable isotopes. For medical investigations, isotopes with a relatively short half-life (the time in which the activity reduces by a factor of two) should be used so that the radiation burden for the patient is kept to a minimum. Generators contain radioactive material with a relatively long half-life. The radioactive material produces the short-lived isotope by disintegration. Because the generator material has a relatively long half-life it can be used over several days for producing the required wanted short-lived isotope. Each day the shortlived isotope is removed from the generator and used for medical investigations. The technetium generator, for example, contains the radioactive isotope molybdenum-99 that has a half-life of 67 hours. This isotope disintegrates to technetium-99m, having a half-life of 6 hours. This isotope is used for imaging purposes. December 98 39

40 9. THE GAMMA-CAMERA This section include: 9.1. Histogram mode 9.2. List mode 9.3. Synchronised recordings 9.4. Processing of scintigrams The radiation that passes through the body is detected by a so-called gamma camera. Radioactive isotopes can emit several types of radiation. Nuclear medicine mainly utilises isotopes that emit gamma radiation of a suitable energy. Gamma radiation is similar to X-rays. The difference is the origin of the radiation. Gamma radiation is produced by the nuclei whereas X-rays are produced by decelerated electrons. What is suitable energy is determined by two contradictory requirements: the energy must be high enough to avoid the total absorption of all the radiation by the body of the patient. The extent of attenuation decreases with increasing energy the energy of the radiation also determines the sensitivity of a detector. The higher the energy, the lower the sensitivity of a detector In early applications in which radioactive iodine-131 was used to determine its accumulation in the thyroid, a Geiger counter was used. This detector had a low sensitivity- only 1% of the radiation was detected. By making use of scintillation detectors the sensitivity could be increased. In these detectors the incident gamma radiation is transformed into scintillations (light pulses). A sodium-iodide (NaI) crystal activated with thallium was used as a scintillation detector. The scintillations are transformed into electric pulses with the help of a photo-multiplier tube. In order to be able to measure the distribution of the activity inside an organ, scanners were developed. In order to measure the activity in small volumes, a collimator is placed between the detector and the patient. This collimator is a lead block in which cone-shaped channels are drilled in such a way that all channels appear to originate from one focal point, that is positioned at a certain depth below the collimator. Only radiation that passes through these channels can reach the detector. The detector therefore is sensitive to the radiation coming from a small volume in the organ, located at the focal point. When the detector is moved along the organ, the activity distribution over the whole organ is measured. Using this method was very time-consuming: scanning of the organ took some time in the order of half an hour. A new development was reported by Anger in He described a scintillation camera with a field of view of 10 inches. With this camera, a whole organ could be December 98 40

41 visualised directly. The camera consisted of a NaI crystal with a thickness of half an inch and a diameter of 10 inches. The thickness of the crystal was a factor of four thinner than usual for the detectors used in scanners, in order to be able to accurately determine the location where a scintillation in the crystal was generated. Because of its small thickness, the sensitivity of the camera was much lower than that of the scanner when using radioactive isotopes available at that time. Moreover the resolution produced by the camera was also worse than that of the scanner. When technetium-99m, emitting gamma radiation with an energy of 140 kev, became available, the Anger gamma-camera was reconsidered. The sensitivity of the crystal for this energy was much higher. A collimator was used in front of the crystal to permit imaging. In the collimator, a large number of parallel channels are present. Only gamma radiation passing through these channels will reach the crystal. Each parallel channel only lets through those gamma rays that have a direction parallel to the channel walls. This means that radiation coming from a 'tube' of tissue in the body that fits the channel of the collimator will produce scintillations in one position of the detector. Therefore we have a similar situation as with X-rays. We again obtain a shadow image, so also here tomography can be considered worthwhile. A number of photo-multiplier tubes are positioned on the crystal. Each of these photo-multipliers delivers a pulse when a scintillation is produced within the crystal. The photo-multipliers closest to the scintillation produce the largest signals. By suitably connecting the photo-multiplier outputs, two signals are obtained. One signal is proportional to the X-coordinate of the location where the pulse was generated and the other is proportional to the Y-coordinate. A third signal that is used is equal to the sum of the signals produced by all photo-multipliers together. This signal is called the Z signal (figure 29). December 98 41

42 Figure 29: Lay-out of the gamma camera As can be seen from the figure, the output signals of the photo-multiplier tube (PMT) array are amplified and then summed using a matrix of resistors. In modern camera's the signals coming from each photo-multiplier tube are digitised (with the help of an analog-to-digital converter). The digital information coming from all photo-multipliers is now used by software-based algorithms to determine the position of the scintillation in the detector (figure 30). December 98 42

43 Figure 30: Lay-out of a digital gamma camera According to modern physics, radiation can be regarded as a stream of quanta (wave packets) and these quanta are either totally absorbed or not absorbed at all. When a quantum is absorbed in the NaI crystal, a photon (light quantum) is released having the same energy as the gamma-quantum. The magnitude of the Z signal is proportional to the energy of the absorbed quantum of radiation. With the help of this signal, gamma radiation coming from the injected radioisotope can be discriminated from unwanted background contributions. When the Z signal has the correct magnitude, the X and Y signals are digitised and stored in computer memory. These signals can be stored in different ways. The most common ways are the so-called histogram mode and the list mode. December 98 43

44 9.1. HISTOGRAM MODE In the histogram mode, the field of view of the gamma-camera is divided into a matrix of pixels. Matrices of 256 x 256 pixels are common. In the computer memory, as many memory locations (bytes or words) are reserved as there are pixels. Before data acquisition, the contents of the memory locations are set to zero. After the detection of a gamma quantum at a certain location in the field of view, the contents of the corresponding memory location is incremented by one. This procedure is repeated until a pre-defined time period has elapsed or a predefined number of counts (quanta) have been detected. The contents of each memory location are now equal to the number of quanta detected in the corresponding location within the field of view of the gamma camera. The count distribution representing the activity distribution over an organ can be displayed as a colour distribution on a video screen. This colour distribution (or grey scale distribution) is useful for diagnostic purposes. It is used in so-called static studies, which investigate whether the observed activity distribution is normal for the organ under consideration. With the histogram mode, dynamic studies can also be performed. In such dynamic studies, the computer stores the data as several images. After a predefined time interval, the counts are stored in the next image. This procedure is repeated each time the pre-defined time interval has elapsed. In this way it is possible to study dynamic processes: the activity distribution as a function of time is presented as a number of images. How much time is to be allocated to the acquisition of each image depends on the type of study. For example, in renal studies, the initial images will each be acquired over one second. For the subsequent images the acquisition time can be longer. This is because of the fact that a radioactively labelled material usually arrives quickly in the kidneys, whereas the washout period during which the agent leaves the kidney will take more time LIST MODE If the appropriate time interval for each image in a dynamic study is not known in advance, the list mode will be used. In this case the X and Y co-ordinates of the detected quanta are stored sequentially. This means that much more storage is needed than in the histogram mode. In the histogram mode as many memory locations are necessary as there are pixels. In the list mode, one needs as many memory locations as there are counts. After acquisition of the basic data, images of different formats and acquisition times can be constructed. In addition to information about the X and Y co-ordinates, time intervals are also stored (say, every ten milliseconds). By using the time intervals in conjunction with the coordinates many different image sequences can be generated. The list mode is usually used for new investigations where it is not yet clear how many images with what acquisition time intervals are needed to produce the best results. December 98 44

45 9.3. SYNCHRONISED RECORDINGS Many physiological phenomena are so fast that it is impossible to produce satisfactory images using small enough acquisition time intervals. If we have (quasi-) periodic phenomena, a synchronised recording is possible. In heart studies the R-wave from the EKG is used for synchronisation. By dividing the RRinterval in a sufficient number of sub-intervals it becomes possible to obtain scintigraphic images of the beating heart. The images, replayed as a video, show the movement of the heart muscle during one heart cycle. Only acquiring data during one heart cycle will not provide enough counts to obtain interpretable images. The quasi-periodic character of the heart movement can be used to increase the number of counts in each image. Each time a new R wave is detected, the acquisition sequence is repeated. In this way each heart cycle adds data to each image. The procedure is continued until each frame (image) contains enough counts to generate a clear image. In heart studies, one is usually interested in the contraction of the heart muscle as a function of time and location. To that end in the order of forty frames are accumulated (each covering an interval in the order of 20 to 30 milliseconds) PROCESSING OF SCINTIGRAMS Computer systems are not only used for the accumulation of images but also for the processing of captured images. With the help of the computer: the contrast can be increased images can be corrected for background activity and non-uniformities of the crystal images can be filtered to erase features in the images that are irrelevant for the study the images can be analysed (by determining the number of counts in different areas) In heart studies, the interest may be in the ejection fraction, indicating how efficiently the pump works. The contour of the left ventricle is automatically detected, both in the end-systolic and end-diastolic phase and on the basis of the activity remaining inside the heart at systole, after correction for background activity, the ejection fraction can be calculated. Similarly determining the radioactivity in the left ventricle by sector provides the ejection fraction for each of the sectors. 10. TWO DIMENSIONAL RECONSTRUCTIONS The images obtained with a gamma camera are two-dimensional projections of three-dimensional activity distributions. By rotating the gamma camera, patient December 98 45

46 activity profiles can be obtained from a large number of angles. In a similar way as in CT scanning, the activity distribution of various slices through the organ of interest can be obtained from the activity profiles. More Anger detectors are used (two to four) to speed up this procedure in modern gamma cameras (figure 31). The procedure is called SPECT (single photon emission computed tomography). SPECT is for example used to make three-dimensional reconstructions of the heart, to perform brain studies, or generate skeletal scintigraphy and similar studies (figure 32). Figure 31: Example of a two-head gamma camera The activity distribution in slices through the organ can be more accurately obtained with another technique, called PET (positron-emission tomography). In PET studies, radio-pharmaceuticals are labelled with positron-emitting isotopes. A positron (a positively charged electron) combines rather quickly with an electron. As a result two gamma quanta (each with an energy of 511 kev) are emitted in December 98 46

47 almost opposite directions. In PET studies, the patient is surrounded by a circle of detectors. When, simultaneously, a gamma quant is detected in opposing detectors it is clear that a pair of quanta have been detected and that these quanta were created somewhere along the path between the detectors. Reconstruction of the image takes place using the same principles as discussed in CT. The advantage of PET scanners is that no collimating is necessary. This enhances the sensitivity of the detection. With the help of PET, biochemical processes can be studied. Biological systems consist for a large part of hydrogen, carbon, nitrogen and oxygen. Of the latter three elements, short-lived isotopes that emit positrons can be produced with the help of a cyclotron - 15 O, 13 N and 11 C with a half-life of respectively 2, 10 and 13 minutes. Using these isotopes various metabolic processes can be studied. Figure 32: Example of a brain scan 11. REFERENCES FOR FURTHER READING Basic Principles of MR Imaging, Phillips Medical Systems, Eindhoven, The Netherlands Bildgebende Systeme fur die Medizinische Diagnostik (H. Morneburg, Editor) Publicis MCD Verlag, Erlangen, (1995) ISBN December 98 47