X-rays are produced by accelerating electrons across a high voltage in a vacuum tube. The x-rays are directed in a beam through the part of the body to be visualized. Some of the x-rays are absorbed in interactions with electrons in the patient's atoms. Others are scattered more or less uniformly in all directions by electrons. Those that are not attenuated (removed from the beam) by the patient are absorbed by a sheet of fluorescent material called a radiographic screen that converts the x-ray energy into light. The light in turn exposes a photographic film in a type of shadowgram. That is, the absorbing properties of all of the objects in the path of the x-ray beam are represented in the image.
The amount of absorption and scattering of the x-rays depends on the energy of the radiation, which can be adjusted by the technologist. Higher energy radiations penetrate the patient much more readily, but also penetrate the radiographic screen. The amount of absorption also depends on the number of electrons in the absorber (i.e., the atomic number). The number of x-rays absorbed per centimeter of a material is proportional to the cube of the material's atomic number. Hence high atomic number materials like barium and iodine make good contrast agents and very high atomic number materials like lead provide maximum protection. Computed radiography makes use of a screen called a photostimulable phosphor plate that stores the energy released by the x-rays until it is scanned by a laser. The laser releases the stored energy as light that is electronically detected to produce a digital image. Computer display of the image allows adjustment of the contrast and brightness by the viewer, which enhances visibility.
Live television images of x-ray absorption are obtained using fluoroscopy. In this case the fluorescent screen is part of a device called an image intensifier which amplifies the light from the screen by converting it to electrons and accelerating the electrons toward a smaller screen. The combination of reducing the image size and giving energy to the electrons increases the light intensity by several thousand times. The light output of the intensifier is viewed by a TV camera. Using a mirror the light can be redirected to a still camera or motion picture camera for recording images. The TV signal can also be digitized for a digitally stored image. This allows advanced techniques such as digital subtraction of images with and without contrast agents for vascular imaging.
For CT (or CAT) scans the x-ray source rotates around the patient. A very narrow beam of x-rays is aimed at a string of small electronic detectors. Strings of information called profiles are collected at many different angles as the tube goes around the patient. These profiles are mathematically projected back across an image (an array in a computer) in such a way that an image of a slice of the patient appears. By appropriately selecting the x-ray energy, CT scans are formed by removal of the radiation from the beam by means of scattering rather than absorption. Because the amount of scattering does not depend on the atomic number as absorption does, the resultant image is a map of tissue density. The computer generated images can be manipulated to enhance visibility of different tissues. Large numbers of consecutive images make up a three dimensional picture of the body which can be sectioned in different orientations or displayed in several 3D methods including virtual "fly-throughs" of vessels, internal airways or the gastrointestinal track.
Images in nuclear medicine the patient is injected with a small amount of a gamma ray emitting radioactive material. This "radionuclide" is distributed in the body based on the chemical composition of the molecules to which the radioactive atoms are attached. The intensity of the radiation source in nuclear medicine is much less than that in x-ray imaging. Information is visualized using a gamma camera that contains a large crystal of sodium iodide, usually about 15 inches in diameter and 3/8" thick. This converts the gamma rays into light in the same way that the screen does in x-ray imaging. Because the radiation intensity is so small the crystal has to be very thick. The individual light flashes from each gamma ray are detected by an array of very sensitive photomultiplier tubes mounted behind the crystal. The tube closest to where the gamma ray created the light flash in the crystal will detect the most light. The exact position of the source of light will be determined from the relative amount of light reaching different tubes. Because the radioactivity is distributed in the patient a large piece of lead with many small holes (a collimator) is used so that only radiation traveling to the detector along specific lines is detected.
Nuclear medicine also has its CT scanners. Those that use the more common radionuclides, such as Technetium-99m, which emit only one gamma ray (or photon) at a time, are referred to as single photon emission computed tomography or SPECT. Some radionuclides, such as Fluorine-18, emit positrons which annihilate themselves when they come in contact with an electron. These positron emitters produce two photons that travel in opposite directions. These are detected by a ring of detectors similar to those in x-ray CT in a process called positron emission tomography or PET. Positron emitters have very short half lives and have to be used shortly after they are produced. Thus PET scanners have to be relatively close to a high energy linear accelerator used to produce the radionuclides
Magnetic resonance imaging (MRI):
MRI involves placing the patient in a large magnetic field (field strengths about one tesla). The hydrogen atoms in the patient tend to line up with the magnetic field like compass needles due to their own magnetic properties. (A rotating proton creates a magnetic field called its nuclear magnetic moment.) Thermal energy (Brownian motion) keeps the atoms from aligning. The total magnetization from all the hydrogen atoms turns out to be only about six millionths of what it would be if all the atoms could be aligned. The alignment process increases following a logarithmic curve where the time it takes to get 63% of the maximum possible alignment (called T1) describes the shape of the curve. This time is different for different tissues and is one source of information in MRI. Maximum alignment is achieved in a couple seconds.
In order to see anything the hydrogen atoms (actually the effective magnetization) have to be precessing (i.e., wobbling like a top or gyroscope that has been bumped). In order to get the maximum amount of precession (ideally flipping the atoms over 90 degrees) a rotating force has to be applied at the frequency at which the atoms are spinning. This so-called Larmore frequency is directly proportional to the magnetic field the atoms experience. At one tesla the Larmore frequency is 43 megahertz (i.e., a radio frequency). So radio frequency energy is applied to "tip" the hydrogen atoms over. Once the atoms are tipped over enough, the radio transmitter is turned off. The precessing atoms emit a radio signal at their Larmore frequency. The signal dies away in a few tenths of a second. This happens because the atoms all experience slightly different field strengths due to the magnetic moments of their neighbors. The time it takes for the signal to die down to 37% of its original intensity is called T2. The difference in T2 among different tissues is also a source of MRI signal information. Some difference in signal also comes from the differences in the density of hydrogen atoms in different tissues, but these differences are small compared to differences in T1 and T2. By adjusting the timing of the sequence of transmitting and receiving the radio signals ( so called echo times and repetition times) the relative effects of T1, T2 and hydrogen density on the image can be adjusted.
The radio transmitter sends its signal throughout the patient. If it were not for some tricky manipulation the received signal would also be picked up from all over the patient. To localize the source of the MR signal, gradients in the magnetic field are used. To begin with when the radio transmitter is on, the external magnetic field is changed in the direction perpendicular to the plane that is to be imaged. Only the hydrogen atoms in the plane that has a field, and therefore a Larmore frequency, corresponding to the transmitted frequency will be tipped. Similarly at different times during the imaging process gradients in the other two directions are used to locate the source of the MRI signals within the excited plane. The timing of signal transmission, reception and gradients and the mathematics used to reconstruct an image are quite complex. In comparison with the simple shadowgrams obtained with x-rays, MR images are not only far more complex, but offer a wide variety of methods that can be used to optimize images of different materials. Contrast agents such as Gadolinium or iron can be introduced. These shorten the T1 because they have different magnetic properties. Thus these agents can be used to study physiologic changes in the same way as materials tagged with radionuclides.
Unlike all the other imaging modalities discussed here, ultrasound does not use electromagnetic radiation to obtain an image. It uses mechanical vibrations at ultrasonic frequencies that are transmitted into the patient and reflected off boundaries between materials that have different acoustic impedance (product of density and speed of sound in the material). The ultrasound scanning device (the "scan head") contains either a row of small acoustic transducers (piezoelectric crystals that convert electrical pulses to sound and vice versa) or a single rotating transducer. The sound waves are sent as pulses. The distance to a boundary is determined by the time it takes the sound pulses to be detected by the transducer after transmission. Fluids (e.g., in cysts) do not produce reflected signals or echoes. They're "anechoic". Tissues like striated muscle have many surfaces and are "echogenic".
The ultrasound pulses are attenuated by tissue. So the brightness of the image has to be adjusted based on how far (how long) the signal has traveled. Ultrasound attenuation depends on the resilience and viscosity of a material. Material like bone is such a good attenuator that very little sound energy passes through it. Hence it produces an "acoustic shadow" beyond it. Air also is a good attenuator. More importantly air has a very different acoustic impedance than either the transducer or soft tissue. Thus the ultrasound transducer has to be in acoustic contact with the skin of the patient via a special acoustic gel. Presence of air in the lungs and bone in many locations limits the use of ultrasound. Some scan heads are small enough to fit in body cavities allowing the radiologist to avoid going through bone or air. Attenuation is also dependent on the frequency of the sound waves used (lower frequencies transmit better). Different scan heads operating at different frequencies are used based on the desired depth of interest. The Doppler effect, caused by changes in frequency of pulses reflected from moving blood cells, is used to analyze movement of blood in the vessels. A two dimensional map of the Doppler signals can be displayed as color on top of a black and white anatomical image in "color Doppler" images.
Sharpness is the opposite of blur. It is the ability to distinguish a sharp edge of an object. A term related to this is spatial resolution, the ability to separate two small objects. We often describe spatial resolution in terms of how close together two lines can be before they appear as one. We have to be able to see a line and a space between it. We call this a "line pair". So sharpness can be described as the number of line pairs per millimeter (lp/mm) that can be seen in an image. The spatial resolution of medical images varies quite a bit. Most radiographic images can display 5 to 7 lp/mm. Mammograms can demonstrate 20. Photographic film (for comparison can show a few hundred lp/mm. Fluoroscopy has poorer resolution than radiography, about 1 or 2 lp/mm. Most CT scans can only show about 1 lp/mm. Surprised? The great image quality of CT doesn't come from spatial resolution but from contrast resolution (more on this later). MRI spatial resolution is even worse than CT. Then comes ultrasound and the worst is nuclear medicine where we are lucky if we get a couple line pairs per centimeter. Obviously sharpness isn't everything.
Sharpness is an important issue in mammography. In order to get sharp mammograms the radiographic screens that expose the film have to be thinner than usual so that the light does not have room to spread out before it gets to the film. Even a small speck of dust between the film and screen in mammography can cause enough spreading of the light to cause significant image degradation. The sharpness of the digital imaging modalities is limited by the size of the displayed image matrix, which is related to the space needed in computer memory. For example a typical CT image of the body will be displayed as an array of 512 by 512 rows and columns of picture elements. If this corresponds to a typical field of view used for imaging the head of 256 mm, you can see that there would be two picture elements for each mm (i.e., 1 lp/mm).
Contrast is the difference between the appearance of light and dark objects. One of the best demonstrations of good and bad contrast is to place a radiograph on a bank of view boxes with all the room lights out and only the view box light behind the film on. Take a good look at it, noting what you can just barely see. Then turn on all the lights in the boxes and all the room lights. The glare from the lights makes it much harder to see the information on the film. This type of contrast we call as system contrast. When you photograph a white object against a black background it has a lot of contrast. A slightly lighter than neutral object against a neutral gray background has low contrast. We call this type of contrast subject contrast. The sources of both subject and system contrast are different in the technologies mentioned above.
Subject contrast in x-ray imaging comes from the difference in the attenuation of the x-rays by the material as discussed above. Subject contrast in CT is far better than in radiography or fluoroscopy because the very narrow beam in CT eliminates scattered radiation from being detected. The contrast resolution in CT is so good that you could see differences in tissue density of 0.1% were it not for the noise (more on this later). Subject contrast is greatly improved in radiography when the technologist properly collimates the x-ray beam (or "cones down") to the body part of interest. A major source of poor system contrast in radiography is over and under-exposure of films. Film contrast depends on the optical density. Contrast is reduced in the very light and dark regions of the film. Over- and underexposure can be caused by improper technique settings or improper processing. Daily quality control of film processing is one of the most important aspects of getting good films. System contrast in CT (or any digital imaging format) can be adjusted by "windowing" the display scale. This allows a different range of densities to be displayed as shades of gray. All the digital systems described above have essentially no system contrast losses other than poor adjustment of the contrast display window.
Subject contrast in nuclear medicine comes from the ability of the agent to be selectively taken up by the organ of interest and is therefore related to the purity of the radiopharmaceutical and by the timing of the imaging process to coincide with the maximum uptake of the radiopharmaceutical.
MRI subject contrast comes from differences in the T1, T2 and hydrogen densities of materials as discussed above.
Contrast in ultrasound comes from the relative density of surfaces in an area that can cause reflections.
We usually refer to noise as the random variations in intensity that obscure small changes in contrast. Its that snowy appearance you get when the TV signal is weak or the grainy appearance of an old photograph. These two kinds of noise are due to the imaging system. There is another kind of noise due to the random nature of the source of the information. In x-ray imaging we call this quantum mottle. It comes from the fact that x-rays are individual packets (quanta) of energy caused by separate atomic events. Radiography is like making a picture with raindrops. If you have only a light sprinkle the shadows of objects aren't well described. Its worse if the objects are partly transparent as they are to x-rays. Quantum mottle obscures low contrast objects. The only cure for it is more quanta. In x-ray imaging or nuclear medicine this translates to more radiation dose for the patient. One of the reasons there is not more sharpness (smaller pixels) in CT is that there would be fewer quanta in each pixel and therefore more noise. This would just mess up the great contrast resolution that is the strong point of CT.
The signal in MRI also comes from individual events (proton interactions) and so gives rise to a type of quantum mottle also. In order to get low noise images in MRI it is necessary to repeat the measurements a number of times. This leads to long exam times. Great efforts are being made to figure out how to get short MRI exposures with acceptable noise levels.
Ionizing radiation:
X-rays, as well as gamma rays from radionuclides, have enough energy to ionize atoms. Ionization (removing electrons from the atom) leads to free radical formation and cellular changes that may lead to cancer. It is thought that this process has no threshold. That is, one photon might cause one chemical change in a particular strand of DNA that controls cell growth. While this may be true, the probability is very small and the possibility of other outcomes (DNA repair, cellular repair or cell death) are sufficient that probable causation can not be determined. While there is probably more epidemiological data about radiation and cancer than any other carcinogen, the uncertainty is great. The nuclear bombing of Hiroshima and Nagasaki gave rise to fewer than 40 more breast cancers than would be expected based on the size of the surviving population. As carcinogens go, radiation seems to be a pretty weak one.
Nevertheless, the possibility of a low dose causing a lethal alteration has lead to a philosophy of maintaining exposures "ALARA" - as low as reasonably achievable - both in the workplace and in medical practice. Further regulations set upper limits on workplace exposures. An average whole body dose (referred to technically as the "effective dose") is limited to 50 millisieverts (mSv) per year due to employment. For comparison, the average effective dose from natural sources of radiation in the environment, referred to as background radiation (cosmic rays and radioactive minerals in soil and food) is about three mSv per year in the U.S. This background varies by about one mSv per year from one geographical location to another. Higher dose limits are allowed to individual organs based on the possibility of direct injury such as burns, hair loss, cataracts and sterility. These so called deterministic effects do not appear to occur in organs that have doses below a threshold of about two sieverts (67 times background) and the effects are severe only at doses several times higher than this threshold.
Deterministic effects haven't been seen in radiologists since the early days of the profession when they would stick their hands in the beam to check if the machine was on. However, in the past few years there have been a number of incidents of patients experiencing alopecia and erythema from high skin doses associated with lengthy and complicated interventional fluoroscopic procedures in both radiology and cardiology.
Children in utero are at a few times higher risk for radiation induced cancer as well as at risk for birth defects. Hence women of child bearing age should be screened for the possibility of pregnancy. Pregnant patients should not have an exam that can be postponed if it will directly expose the fetus, especially in the first trimester after implantation of the embryo in the uterine wall. Nevertheless, doses of more than 50 mSv to the fetus are almost unheard of for diagnostic procedures. Below this dose therapeutic abortion on the grounds of congenital abnormality is not advised and cancer risk is low (recall it's the workplace limit for adults).
One of the worst misconceptions about radiation bioeffects is related to genetic mutations. Fear of mutation based on early radiobiological studies gave rise to the myths of spider man, the incredible hulk and other comic book characters. Recent reviews of the survivors of the bombings in world war two failed to demonstrate a statistically significant increase in genetic mutations in their offspring compared to a matched control group.
Radiation doses to patients vary by several orders of magnitude from one procedure to another. To better understand this a brief foray into the jungle of radiation dosimetry is in order. The radiation absorbed dose refers to the energy deposited per unit mass of tissue. When someone quotes a dose it could be dose to a speck of skin, dose averaged over the body (as in the effective dose), or anything in between. For example the skin dose from a typical mammogram might be around 12 mSv. The average dose to the glandular tissue of the exposed breast is closer to 2 mSv. The effective dose due to that one film is about 0.1 mSv (5% of the gland dose). Now the breast makes up less than 5% of the body mass, but it makes up 5% of the risk of cancer. That's how effective dose is determined. Typically a mammographic study involves two films of each breast. The average glandular dose is therefore 4 mSv, not eight, because you have twice as much tissue in two breasts as one.
To further confuse the neophyte, radiation experts change the names of the units from time to time. (How else could we stay experts?) The following is here just in case you see these terms in the literature and want to translate them. The old unit for the sievert, called the rem was a hundred times smaller. The gray is a unit that is the same as a sievert as long as we limit our conversation to x- and gamma rays. Likewise the rad is a unit equivalent to the rem with the same restrictions. Technically rads and grays refer to dose and rems and sieverts refer to something called equivalent dose or to some people dose equivalent. In some circles the effective dose is called the effective dose equivalent. The roentgen is a unit used to describe the intensity of the radiation beam (technically the exposure). An exposure of one roentgen gives rise to a soft tissue dose of a little less than one rad (or one rem, 10 mSv or 10 milligray) if you put some skin at that location. One roentgen would cause a dose to bone of about 2 rad at the same location however. Confused? Well I have to make a living somehow.
Radiation safety can pretty much be reduced to three things. Shorten the time you are exposed, increase your distance from the source, and put some shielding material between you and the source. If you ever find yourself operating a fluoroscope you can be very effective in reducing exposure time by just taking your foot off the switch when you aren't looking for movement. This reduces dose to the patient, to yourself and to anyone else working in the area. To make use of shielding and distance you need to know where the radiation source is. The x-ray tube is very heavily shielded so unless you're dumb enough to stick your hands in the beam, most of the radiation you will be exposed to is scattered from the part of the patient being examined. The more tissue in the beam, the more the scatter will be. Stepping back from this secondary source of radiation can dramatically reduce your exposure. The exposure goes down as one over the square of the distance from the source (like the size of your shadow as you walk away from a bright light source). Lead is the preferred material for shielding. The typical lead apron will reduce dose to tissue behind it to about one to ten percent. Remember high atomic number materials all make good shields. A couple layers of sheet rock in walls are as good as a typical apron. Even a nice thick medical textbook will absorb about half the radiation striking it. Of course, you have to be selective about what you want to cover with the book.
Nuclear Medicine:
Radiation safety in nuclear medicine is a bigger problem. When an x-ray machine is turned off, the radiation is absorbed and converted to heat or chemical changes at the speed of light. Radioactive materials are the radiation source used in nuclear medicine studies. They decay naturally with a half life varying from a few minutes (Fluorine-18 used in PET) to a week (Iodine -131 used in thyroid therapy). Patients carry the source material around in them. Patients may bleed or regurgitate this material. Technologists can drop the material or dribble it out of a syringe by accident. Constant routine monitoring is needed anywhere significant amounts of radioactive materials are used. The details of these procedures are beyond this discussion. Suffice it to say, the U.S. Nuclear Regulatory Commission (NRC) has very strict and detailed regulations regarding radiopharmaceutical use and it enforces them like the IRS enforces tax laws, except that every licensee gets audited by the NRC.
There are two primary potential safety hazards in MRI, fixed magnetic fields and radio frequency (RF) radiation. The powerful magnet field used in MRI (typically 1.5 tesla), is beyond most peoples normal experience. The magnetic field can make projectiles out of ferromagnetic materials. Screwdrivers, scissors, clamps, IV poles, wheelchairs and even computers can, and have become airborne in and around MR facilities. The field can stop a pacemaker, and tear aneurysm clips or other prosthetic devices from their location in a patient. Tiny metal scraps in some ones eye can be a serious problem. All patients have to be screened to avoid the possibility of damage from the strong static magnetic field. All staff entering the MR exam room have to remove any loose ferromagnetic objects before entering. Patients requiring emergency assistance must be removed from the room for treatment. The field is strong enough to pick up a crash cart. Anything that becomes mobile due to the field goes toward the center of the magnet. Patients have been impaled and seriously injured in MRI suites.
The RF fields are in the same energy range as those in a microwave oven. While they are non-ionizing and therefore do not have the same effects as x-rays, RF energy can heat things up. FDA regulations and the computer program running the MR machine will keep the RF heating to an acceptable level for normal soft tissue. But a metal wire (not necessarily ferromagnetic) will act as an antenna if it are formed in a loop and will get hot. Patients have been burned by leads for electronic devices laid carelessly close to their skin. A sufficiently large internal loop of wire could have the same effect.
Other contraindications to an MRI include claustrophobia and obesity (the tunnel is quite small).
The energy used in ultrasound is essentially of a mechanical nature. Vibration of tissue at ultrasonic energies has been shown to cause chromosome aberrations and at sufficiently high energies, heating and cavitation. The energy levels used in diagnostic imaging does not cause significant heating or cavitation, although that used in continuous fetal heart monitoring approaches significant levels. No one knows what the effects of ultrasound induced chromosome aberrations are. Perhaps they are the same as x-ray induced chromosome aberrations, but epidemiological studies to date have been insufficient to show a significant effect.
Further reading
W. R. Hendee and E. R. Ritenour, Medical Imaging Physics
, 4th ed., John Wiley, 2002