Gamma Camera, SPECT, MRI PDF

Magnetic resonance imaging (MRI), gamma camera, SPECT The text under the slides was written by Peter Nagy 2022 This educational material was prepared for the biophysics lectures held by the...

Magnetic resonance imaging (MRI), gamma camera, SPECT The text under the slides was written by Peter Nagy 2022 This educational material was prepared for the biophysics lectures held by the Department of Biophysics and Cell Biology Faculty of Medicine University of Debrecen Hungary https://biophys.med.unideb.hu 2 The purpose of the lecture is to review three medical diagnostic procedures applying electromagnetic radiation. The gamma camera and its three-dimensional counterpart, SPECT (single photon emission computed tomography), are based on detection of gamma photons emitted by gamma-decaying isotopes. While understanding these techniques is simple, MRI (magnetic resonance imaging), based on the principle of NMR, is difficult to comprehend, since several methodological steps are in between the basic physical phenomenon (NMR) and image generation. Therefore, it is strongly advised to master the principles of NMR before studying this topic. All three imaging approaches demonstrate how basic physical principles (NMR, radioactivity) can be used in medical practice. 3 Before taking an image with a gamma camera a radiopharmaceutical labeled with a gamma- decaying radioactive isotope is injected into the patient, and gamma photons emitted by this isotope will be detected by the gamma camera. As opposed to conventional X-ray or X-ray absorption CT, in which absorption of radiation from an external source is measured, the distribution of a radioactive isotope in the body is detected by a gamma camera. Image formation is based on converting gamma photons, emitted by the radiopharmaceutical, to visible photons (scintillation) followed by converting the visible light flashes to electric impulses (PMT, photomultiplier tube). A radioactive isotope selectively taken up by on organ (e.g. heart, “A”) emits gamma photons in all directions, which will first go across the collimator. A collimator is thick lead slab (for its image consult the next figure), in which there are holes perpendicular to the surface. Since gamma photons not parallel to the holes are absorbed, flashes will only be generated in those parts of the scintillation crystal that are above the source (A). Visible photons generated in the scintillation crystal are converted to electrical impulses by photomultiplier tubes. Consequently, only the pixels corresponding to areas above the  radiation source will be bright in the computer-generated image (B). A collimator is an indispensable part of a gamma camera, essential for image formation. Its role is similar to that of an objective in light microscopy. A pitfall associated with collimators is the fact that they absorb most of the radiation impinging on them. This circumstance is to blame for fact that only a tiny fraction of emitted gamma photons is detected during gamma camera 4 imaging. The role of the collimator is easiest to understand if we imagine what would happen if we removed it from the device (C). In this case the collimator would not only be exposed to gamma photons above the place where the isotope is concentrated. Consequently, the source could not be located, no image would be formed. 5  A collimator is a metal slab containing many holes (A). The scintillation crystal is found above the collimator in a gamma camera. Photomultiplier tubes are located on the other side of the scintillation crystal. A typical gamma camera contains approximately 50-100 photomultiplier tubes (B). The collimator determines the resolving power of the gamma camera, similar in principle to a microscope objective. The gamma camera or the scintillation crystal is unable to form a point-like image of a point-like source, just as the light microscope. Instead, image spots, whose size is determined by the geometrical properties of the collimator (e.g., length and diameter of holes), are generated. The red and green ovals in figures C and D represent image spots corresponding to two point-like gamma radiation sources. If the diameter of the collimator holes is suitably small, images of the two sources are distinguishable (red and green ovals do not overlap, C). If the diameter of the collimator holes is larger, the two images overlap, i.e. the sources are not distinguishable (D). The figure demonstrates that the collimator functions similar to a microscope objective in many respects. 6 The lecture about the detection of ionizing radiation has already discussed the basic phenomenon of scintillation. This figure describes the process in more detail. The gamma photon entering the scintillation crystal (1) liberates electrons by photoeffect or Compton effect (2). These electrons induce secondary electrons while interacting with the scintillation crystal (3). The primary electrons, generated by the photoeffect or the Compton effect, or the secondary electrons excite the material of the scintillation crystal (4). Relaxation of these excited particles leads to the emission of visible photons (5), which will be converted to electrical impulses by photoelectron multipliers not shown in the figure. 7 Images produced by the gamma camera are not rich in anatomical-morphological detail. However, the metabolism, the blood flow and many of the biochemical processes, i.e. the function of a certain organ or tissue, are shown precisely. (A) The thyroid gland actively accumulates iodine for producing iodine-containing hormones. Therefore, radioactive iodine (123I) or pertechnetate (99mTcO4–), behaving similarly to iodine, is injected to patients intravenously, and the accumulation of radioactivity in the thyroid gland is measured. This parameter changes when the thyroid gland is underactive or overactive. (B) 99mTc-pyrophosphate injected into a patient will accumulate at places of bone rebuilding. Since bone metastases of malignant tumors induce significant bone rebuilding, i.e. the biochemical processes in the bone are changed, these pathological features can be detected by gamma cameras. The figure also demonstrates that gamma camera images show the two-dimensional projection of the distribution of radioactive isotopes. 8 Since gamma camera images do not provide three-dimensional information about the spatial distribution of isotopes, the tomographic, three-dimensional version of the gamma camera, named SPECT (single photon emission computed tomography) was developed. The word “emission” in the name emphasizes that emission of photons is detected, as opposed to X- ray absorption CT, in which absorption of radiation is measured. In X-ray absorption CT a three-dimensional image is produced from two-dimensional projection images by recording such two-dimensional images from many directions. The same principle is applied in SPECT as well. A gamma camera image is produced at a certain position followed by the camera rotating around the body and recording another image from another direction. In order to accelerate the process SPECT devices equipped with more than one gamma camera are also available. The three dimensional image is produced from the 2D projection images generated by the rotating gamma cameras (or cameras) according to the principle discusses in the lecture about X-ray absorption CT. 9 Contemporary SPECT devices typically contain more than one gamma camera in order to accelerate investigations. In addition, combined devices are also available in which CT and SPECT images are produced at the same time (SPECT-CT). Thus, functional (SPECT) and morphological (CT) information can be obtained from the patient in a single investigation. 10 Since SPECT is a functional imaging modality, SPECT images shed light on the physiological and pathological processes in a certain organ. SPECT is often used to assess blood perfusion, which is proportional to the activity of a certain brain region. The cortical areas display the highest activity in a healthy individual (A). Characteristic hypoperfusion (lower blood perfusion, and consequent lower activity) can be seen in Alzheimer’s disease (white circle, B). The functional information provided by SPECT is supplemented by the morphological background provided by CT, a principle demonstrated by the bone metastasis of breast cancer (C). The anatomical location of the metastasis can more accurately be determined by overlaying the SPECT image on the CT picture. 11 MRI (magnetic resonance imaging) is based on the principle of NMR discussed previously. Nuclei are placed in a strong external magnetic field (A), in which the energy of the  and  spin states splits. Spins are then excited by radiofrequency photons, whose frequency corresponds to the resonance condition. As a result, the macroscopic magnetization is rotated (B). The extent of the rotation is determined by the duration of the excitation; in most cases a rotation of 90 degrees is induced. After turning off excitation, the macroscopic magnetization rotates back to its original position, which is parallel to the external magnetic field (relaxation). MRI not only involves these steps, but it is also aimed at locating the signal. In other words, we would like to know how rapid the relaxation processes are in different parts of the body. To this aim, the strong external magnetic field is combined with magnetic field gradients. 12 Locating the signal is based on the principle that the resonance frequency is always determined by the local magnetic field (for a certain type of nucleus). Consequently, if the local magnetic field strength is increased, so is the resonance frequency. This principle can be utilized for locating the signal with linear magnetic field gradients. If the magnetic field strength changes in space, the resonance frequency of a certain type of nucleus (typically 1H) is also altered: different location different Blocal different resonance frequency. Reversing the chain of thought we can state that if we know the resonance frequency (i.e. we know that a photon of a certain frequency induces excitation), we can figure out the source of the signal if the parameters of the gradient field are known. The local magnetic field sensed by the spins during their excitation is determined by two factors in MRI: the homogeneous magnetic field (B0) onto which a linear magnetic field gradient is added. When interpreting chemical shift atoms surrounding a certain 1H nucleus were also considered enabling us to explain why the resonance frequency of 1H is different in e.g. CH3 and OH. Although these processes obviously take place during MRI as well, the change in the local magnetic field induced by the chemical environment (e.g. the chemical shift) is negligible compared to the magnetic field gradient. Consequently, chemical shift is ignored. 13 The figure demonstrates how field gradients can be used to localize the signal. The resonance frequency of 1H at three different places (red circles on the head) are examined. The equation required to understand the figure: N f  Blocal 2 where f is the resonance frequency, Blocal is the local magnetic field, and N is the gyromagnetic ratio. If the patient is placed into a homogeneous magnetic field (A), protons at each of the three locations experience the same local magnetic field, and their resonance frequency will be identical. If a nose-nape gradient is superimposed on the homogenous magnetic field (B), then the magnetic field will be stronger at the two points farther from the forehead. Consequently, the resonance frequency of these protons will be higher than that of protons closer to the forehead. 14 (A) Let us imagine that a patient is placed into a field, in which both a homogeneous magnetic field (B0) and a gradient field are present. Therefore, the local magnetic field (Blocal) is determined by the two fields. For slice selection the strength of the gradient field changes along the z axis, i.e. along the foot-head direction, although the gradient can be oriented at any arbitrary angle making “slicing” of the patient in any orientation possible. (B) Since the resonance frequency is proportional to the local magnetic field, the resonance frequency of 1H nuclei will change in a similar fashion along the foot-head axis. (C) In other words, splitting between the energies of the  and  spin states changes along the z axis as a result of the inhomogeneous magnetic field. The resonance condition (E-E=E=hf) is only met in a certain slice of the body when using a photon of a certain frequency and E=hf energy. Consequently, spins will only be excited in this slice. If the slope of the gradient is changed, photons of the same energy will select another slice, i.e. spins will be excited in another part of the body. 15 It was explained in the previous slide how slice selection takes place with the Z magnetic gradient. This animated figure demonstrates what happens in the selected slice. Since the resonance condition is only fulfilled for protons in the selected slice (only these protons absorb the radiofrequency photons), the macroscopic magnetization will only rotate by 90 degrees in this slice (red vectors). Since the NMR signal is always provided by the macroscopic magnetization rotated to the horizontal plane, the detected signal originates only from the selected slice, because the macroscopic magnetizations in other parts of the body are located parallel to the external magnetic field (vertical in the figure). The animation in the lecture slide can be viewed at the following link: https://youtu.be/yQdOYfEsrLU 16 Although the source of the signal was restricted to a single “slice” of the patient with the methods described in the previous figures, the detected signal cannot be located within the selected slice along the X and Y axes without further steps. After slice selection (i) two more gradient fields perpendicular to the direction of the slice selection gradient are applied enabling us to locate the signal in three dimensions. First, a phase-encoding gradient (i) is turned on. After turning off this phase-encoding gradient another, frequency-encoding gradient, perpendicular to both previous gradients, is turned on (iii) and the signal is detected with the frequency-encoding gradient still on (iv). Steps ii-iv are repeated many times in order to locate the signal in the X-Y plane. (The effect of the gradients is described in more detail in the next figure.) The animation in the lecture material can be viewed at the following link: https://youtu.be/Q9VAQ57ZF1M 17 The figure shows the processes taking place in the slice selected by the slice-selection gradient during the application of the phase- and frequency-encoding gradients. The slice is viewed from the top, and the red lines represent the macroscopic magnetization rotated by 90 degrees to the horizontal plane. These vectors precess (rotate) by the Larmor frequency determined by the external magnetic field. The Larmor frequency is equal to the frequency of those radiofrequency photons, which are capable of exciting the spins. The figure neglects T1 and T2 relaxation processes, and only displays the aforementioned precession. 1. The magnetic field is homogeneous in the selected slice before turning on the phase- and frequency-encoding gradients. Precession frequency is determined by the external magnetic  field ( f  B ). Since the magnetic field is identical at every location within the slice, all 2 macroscopic magnetization vectors precess with the same frequency. 2. As a result of the phase-encoding gradient applied in the horizontal direction the magnetic field becomes stronger from left to right. Due to the direct proportionality between the magnetic field strength and the Larmor frequency spins on the right precess more rapidly. 3. After turning off the phase-encoding gradient the magnetic field is again homogeneous in the selected slice. Therefore, all spins again precess with the same frequency at every location within the plane. However, the spins on the right precessed more rapidly, they have run forward relative to the spins on the left. Although the precession frequencies are now identical 18 for all the spins, there exists a phase difference between spins on the left, in the middle and on the right. 4. Finally, the last gradient, the frequency-encoding gradient is turned on, which is perpendicular to the slice selection and the phase-encoding gradients. As a result, spins on the top will precess faster due to the stronger magnetic field than spins at the bottom. The NMR signal is detected while the frequency-encoding gradient is on. During detection spins differ in their phase along the horizontal axis (because they “remember” the effect of the phase-encoding gradient turned on in point 2), while they have different precession frequencies along the horizontal axis. Since all the nine spins (more precisely the macroscopic magnetizations in all the nine voxels) differ from each other either in their phase or frequency, we have made localization of the signal in the selected plane possible. For reasons not discussed here the signal must be recorded after several phase shifts so that the signal can be unequivocally be localized, i.e. the following procedure is repeated: 1. turn on the phase encoding gradient; 2. turn off the phase encoding gradient; 3. detect the signal in the presence of the frequency-encoding gradient; 4. restart the sequence from step 1 using a phase-encoding gradient with a different strength. The four animations in the slide can be viewed at the following links: 1. https://youtu.be/bjAN2YddLGk 2. https://youtu.be/32Me0NM8ffA 3. https://youtu.be/otmBUVT8eQ4 4. https://youtu.be/A78TStZXUdI 19 This animated figure demonstrates what kind of principles help us locate the signal in the selected slice along the X and Y axes using a specific, calculated example. The animation merely demonstrates how the combined application of a phase- and a frequency-encoding gradient allows determining the source of the signal, since the applied mathematical approach is completely different from and much simpler than the one used in clinical MRI devices. There are four pixels in the selected slice, designated by A-D. The length of the vectors is also displayed (A=1; B=0.9; C=0.8; D=0.7). Similar to the previous figure, the vectors represent the macroscopic magnetization rotated to the horizontal plane. Only the precession of these vectors is simulated, and their relaxation is neglected. During the animation the plot above “NMR signal” displays the oscillating NMR signal, i.e. the projection of the sum of macroscopic magnetization vectors on one of the horizontal axes (X or Y). Alternatively, the NMR signal can also be interpreted as the current induced in a coil in the X-Y plane. The plots above this part display the Fourier transforms of the NMR signal. Fourier transformation is a mathematical procedure that determines the contribution of different frequency components to an oscillating signal. The eight Fourier transforms calculated in the eight steps of the simulation are displayed in fields FT1-FT8. It is worth noticing that if the NMR signal is a single-component sine wave, then the Fourier transform only contains a single peak. If, on the other hand, the NMR signal is composed of two different frequency sine waves, the Fourier transform also contains two peaks. 20  First, spins precess with identical frequency in a homogeneous magnetic field, and consequently only a single peak appears in the Fourier transform (1). Then, the phase- encoding gradient is turned on, and the spins in pixels on the left and right exhibit different precession frequencies. Consequently, there are two peaks in the Fourier transform (2). After turning off the phase-encoding gradient, the frequencies are again identical for every pixel, but a phase difference persists (3). The phase difference generated between pixels on the left and right is determined and displayed (1). The Fourier transform again contains only one peak in accordance with identical precession frequencies for all pixels. Afterwards, the frequency-encoding gradient is turned on along the vertical direction, and spins precess with two different frequencies, and there are two peaks in the Fourier transform (4). The two peaks in the Fourier transform are measured and displayed (P1, P2). Peaks P1 and P2 are determined by variables A-B and C-D, respectively, and by the phase difference according to the equations displayed in the lower right corner. The phase-encoding – frequency-encoding pair must be repeated once more (5-8) so that four independent equations can be written for the four unknowns (A-D). At the end of the animation the program calculates A-D by solving the equation set consisting of four equations. The solutions for A-D are identical to the assumed values of the variables. A web-version of this simulation with executable animations is available at https://peternagy.webs.com/mri. The whole animation can be viewed at the following link: https://youtu.be/AT648Qiup8I 21 Spatial resolution is a characteristic property of every imaging technique, which can be described by the size of voxels. Voxel size is determined by the thickness of a slice and the number pf pixels in a slice. Spatial resolution can be improved by decreasing the size of voxels, but this improvement comes at the price of reduced NMR signal due to fewer protons being present in smaller voxels. Reduced signal undermines the reliability of the measurement. A compromise must be reached between good spatial resolution and high enough signal, leading to a practical resolution of clinically applied MRI devices in the range of a couple of millimeters. 22 The magnetic field in clinical MRI devices is extremely strong, enhancing the strength of Earth’s magnetic field by five orders of magnitude. Therefore, it is prohibited to wear ferromagnetic materials for the patient and the staff since such materials are accelerated by the magnetic field. 23 In an MRI examination the relaxation of the macroscopic magnetization removed from its equilibrium position is measured. Due to intrinsic contrast generating mechanism, briefly discussed later, three different types of information can be visualized in MRI images, which are the time constants characterizing the rate of the two relaxation processes (T1 and T2) and density of spins, typically 1H nuclei, providing the signal. Extrinsic contrast agents can also be used in MRI, which alter the relaxation processes locally. According to the values displayed in the table T1 relaxation in biological tissues is significantly slower than T2 relaxation, and soft tissues can be differentiated from each other based on their relaxation times. The source of the MRI signal is also part of the interpretation of MRI images. In most cases water hydrogen nuclei provide the signal, although in some cases 1H nuclei in fatty acids also contribute. In practice, no source other than water and fatty acids contributes to the signal in MRI (if 1H nuclei are excited). 24 Timing of measurement of the NMR signal forms the basis for specific measurement of the relaxation times (T1 and T2) and the spin density. Parts “A” and “B” of the figure provide help for understanding these principles. Remember that the NMR signal is provided by the precession of the transverse magnetization vector (MXY) produced by the 90-degree pulse. The size of MXY and so the amplitude of the signal depends on the size of the longitudinal magnetization (MZ) rotated to the xy plane by the 90-degree pulse. After switching off the 90- degree pulse the longitudinal magnetization increases in size due to the spin-lattice relaxation process, as shown in Figure A. The rate of recovery is characterized by the T1 relaxation time. In NMR applications, including MR imaging, several excitation and detection cycles are repeated after each other. In order to get the maximal signal (i.e. full recovery of the equilibrium magnetization), at least a time interval of 5 T1 should separate subsequent excitations (90-degree pulses). If the waiting time or repetition time (TR) is shorter, then a smaller magnetization is rotated to the xy plane, so the signal will also be smaller. This way we can distinguish tissues with different T1 relaxation times even if they have similar spin densities: the next excitation (i.e. the next 90-degree pulse) is to be done relatively early, at the time point shown by the dotted line in figure A. If the measurement is performed according to this timing (with the aim of emphasizing differences in T1 relaxation), the signal will stronger in the pixel with rapid T1 relaxation (T1,B). If the aim is to suppress differences in T1 relaxation times (e.g. to display T2 relaxation times or spin densities), then the waiting time (TR) should be relatively long (continuous line in figure A), when the signal from pixels 25 exhibiting different T1 relaxation times will be nearly identical, i.e. independent of the rate of T1 relaxation. The other process influencing the magnitude of the signal is spin-spin relaxation causing the decay of transverse magnetization after the 90-degree pulse (Figure B). This process is characterized by the T2 relaxation time. The rate of spin-spin relaxation should be taken into account when the signal is detected: the later the signal is detected, the lower its amplitude is. If the spin-spin relaxation process is over, the signal is lost. This principle is demonstrated for the measurement of T2 relaxation by figure B. If we would like the signal to depend on T2 relaxation, the measurement is to be carried out relatively late, at the time shown by the dotted line. With such a long detection time (TD), the signal coming from pixels characterized by slow relaxation (T2,A) will be larger than the signal collected from rapidly relaxing pixels (T2,B). If differences in the pace of T2 relaxation is to be suppressed (e.g. in order to generate T1-weighted or spin density-weighted images), the signal is to be measured at an early time point (short TD), shown by the continuous line, when the signal from pixels with short and long T2-s is comparable. Let us investigate in a model, shown in part C, how T1-, T2- and spin density-weighted images can be made building upon the principles described in figures A and B. Spin densities are identical in pixels A and B, but both T1 and T2 relaxations are slower in pixels labeled by A. A T1-weigthed image can be generated by adjusting both the detection time and the waiting (repetition) time to be short. In this case the signal will depend on T1 according to figure A, and differences according to T2 will be suppressed. The signal in pixels exhibiting fast T1 relaxation is going to be larger, therefore these pixels will be bright. According to the same logic the waiting (repetition) time must be adjusted to be long in order to obtain a T2-weighted image, since differences according to T1 will be suppressed (see figure A). In order to emphasize differences in T2 the detection time must also be long (according to figure B). Therefore, if a pixel exhibits rapid T2 relaxation, the signal will be small in a T2-weighted image, and these pixels (pixels B) will be dark. If the effect of both relaxation processes is to be eliminated from an image, the waiting (repetition) time is to be adjusted to be long, while the detection time must be short. In this case the signal will be independent of the relaxation times, but will instead be influenced by the density of spins. Figure D displays, using brain MRI images, how weighting on spin density, T1 and T2 relaxation can be achieved by appropriate timing of measurement of the signal. According to the figure shown on page 24 the T1 and T2 relaxation times of white matter are shorter than those of gray matter. Fast T1 relaxation leads to rapid recovery of the signal, i.e. large signal at the time of the measurement (see figure A). Therefore, white matter is brighter in T1- weighted images. In contrast, rapid T2 relaxation means rapid decline of the signal in white matter resulting in a small signal at the time of measurement. Therefore, the signal of white matter will be smaller at the time of measurement (see figure B), and white matter appears darker in T2-weighted images. 26 Besides intrinsic contrast introduced in the previous figure extrinsic contrast agents injected into patients can also generate or enhance contrast of MR images. MRI contrast agents are paramagnetic or ferromagnetic materials, and due to their toxicity they are typically applied in complexed form. The most commonly applied MRI contrast agent is gadolinium. MRI contrast agents locally change, typically accelerate, both T1 and T2 relaxation processes. Using the very same weighting methods mentioned in the previous slide, regions accumulating the contrast reagent can be “highlighted” on the image. At the bottom of the figure application of gadolinium for revealing the damaged blood-brain barrier is presented as an example. In the T1-weighted image of the brain white matter is brighter than gray matter, and the signal is somewhat weaker (due to slower T1 relaxation) in a faintly visible area on the right. This alteration becomes much more obvious after applying the gadolinium-containing contrast material. The contrast agent cannot go across the blood-brain barrier in healthy parts of the brain, and therefore the relaxation time is not influenced there. On the other hand, it can enter the damaged brain across the damaged blood-brain barrier, and by shortening T1 relaxation time the signal will be stronger in the T1-weighted image. 27 MRI maps the relaxation and density of 1H nuclei in the body, and it is, therefore, suitable for imaging hydrogen-rich soft tissues. It is exquisitely well suited for investigating the brain due to its high water content. While the strong X-ray absorption potential of the skull casts a shadow on the brain in CT, such a detrimental effect does not take place in MRI. Another significant advantage of MRI is that no ionizing radiation is used, but ferromagnetic materials must not be brought close to the device. The price of the examination also significantly limits the more widespread application of MRI. Although not discussed in the biophysics course MRI can also be used for functional imaging (e.g. for measuring perfusion), and not only for anatomical imaging. Medical applications of MRI are exemplified by a T1- and T2-weighted image of the healthy brain and by T1- and T2-weighted images of a brain tumor. In tumors both relaxation times are elongated, therefore in T1-weighted images the signal will be lower, whereas in T2-weighted images it will be higher compared to the surrounding areas. 28 29