| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Special Reviews |
1 From the Department of Radiation Medicine, Georgetown University Medical School, Washington, DC (A.B.W.); and Medical College of Wisconsin, 8701 Watertown Plank Rd, Milwaukee, WI 53226 (W.R.H.). Received September 20, 2004; revision requested November 15; revision received December 9; accepted January 14, 2005; updated July 14; final version accepted August 5. Supported in part by grants from the National Institutes of Health (RO1 CA80490, P01 CA 87634) and GE Medical Systems. Address correspondence to W.R.H. (e-mail: whendee{at}mcw.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONADVANCES IN CURRENT TECHNOLOGIESDEVELOPING AND EXPERIMENTAL...EVOLVING ROLE OF COMPUTERSCONCLUSION: IMAGING WITH A...ESSENTIALSReferences |
|---|
© RSNA, 2006
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONADVANCES IN CURRENT TECHNOLOGIESDEVELOPING AND EXPERIMENTAL...EVOLVING ROLE OF COMPUTERSCONCLUSION: IMAGING WITH A...ESSENTIALSReferences |
|---|
As a result, there are a number of widely accepted standard film-based and digital technologies to choose from in posing and addressing a particular clinical question. Part of the task of the clinician is to be adept at selecting the one with the greatest likelihood of providing the correct diagnostic answer safely and at acceptable cost.
There are also several nonstandard imaging modalities, some new and others that have been around for awhile. These are being explored and may (or may not) eventually find routine clinical application. A few of them, such as electroencephalography, magnetocardiography, and thermography, create images from extremely faint signals produced by the body itself. Tissue impedance tomography and diaphanography operate, like x-rays, on the basis of observations of how probes passing through the body interact with it.
Meanwhile, biomedical research is becoming ever more interdisciplinary, as it depends increasingly on cross fertilization between the physical sciences and engineering on the one hand and biology and clinical medicine on the other. Examples of the fruits of such collaboration abound—such as recent developments in molecular imaging, biomedical informatics, nanobiotechnology, computer-aided detection and diagnosis, and the merging of diagnosis and therapy as in image-guided intervention.
This review will provide a brief overview of the current state of image-based diagnostic medicine and offer comments on the demonstrated or potential clinical efficacy of some of the newer imaging modalities. Additional viewpoints may be found in the reviews and other articles listed among the references.
|
ADVANCES IN CURRENT TECHNOLOGIES |
|---|
|
TOPABSTRACTINTRODUCTIONADVANCES IN CURRENT TECHNOLOGIESDEVELOPING AND EXPERIMENTAL...EVOLVING ROLE OF COMPUTERSCONCLUSION: IMAGING WITH A...ESSENTIALSReferences |
|---|
|
In planar x-ray imaging, new electronic image receptors are taking over the field (2). The standard combinations of fluorescent screen plus x-ray film and of image intensifier plus television monitor or charge-coupled device optical camera have served radiology well for a good part of a century, but their evolution has largely plateaued. Meanwhile, digital image receptors keep on getting better, and fully digital systems, built around picture archiving and communication systems (PACS), may soon be dominating x-ray projection imaging. A number of radiology departments already are filmless, and the direction of the trend is clear (3). Even in breast imaging, where resolution is essential for the assessment of microcalcifications, digital systems are becoming increasingly accepted (4,5).
Photostimulable phosphor plates composed of BaFBr and BaFI have been used in computed radiography cassettes since the 1970s. More recently, the real-time, active-matrix, flat-panel imager of digital radiology and digital fluoroscopy has begun displacing its analog predecessors. A flat-panel imager is an array of hundreds, thousands, or millions of tiny independent semiconductor detectors that are themselves sensitive either to high-energy photons (direct-detection array) or to light from an adjacent thin layer of fluorescent material (indirect-detection array) (6,7). It is possible that recently developed ink-jet printing (rather than photographic) lithographic technology will allow the fabrication of flat-panel imagers on flexible plastic substrates of virtually any size.
There are advocates of both computed and digital radiography, but the two are complementary modalities, and there is room for both (8). Computed radiography cassettes are somewhat more flexible—coming in a range of sizes and shapes—and more cost-effective: To install, one need only exchange a film cassette assembly with a computed radiography cassette; also, a damaged computed radiography cassette can be replaced at reasonable cost. Digital radiography allows faster throughput and is more dose effective, but it is also more expensive to implement. Both are likely to be around for a while, providing quality images of teeth, the chest, bone, the gastrointestinal tract, the extremities, and the breast (9,10) and contributing to a growing variety of interventional procedures.
Among the great advantages of both technologies (and of digital imaging in general) are (a) the linearity of response of digital radiation receptors over exposure ranges that are orders of magnitude greater than the latitude of any film and the overall improvement in image quality that results from that property; (b) the separation of image acquisition from image processing (windowing, edge enhancement, noise reduction, etc) and from display, so that these three steps can be optimized independently; and (c) the overall benefits of digital signal storage, communication, and analysis.
Through the digital subtraction of images created with different x-ray tube kilovoltage settings, for example, dual- (and higher-) energy methods (11) can be used to remove unwanted visual background noise, such as bonelike or soft-tissue structures, that obscures what the clinician is looking for. In a related vein, dual-energy x-ray absorptiometry is the standard method for measurement of bone density, the major predictor of fracture due to osteoporosis. However, dual-energy x-ray absorptiometry is now being challenged by US and other techniques (12,13).
Digital tomosynthesis takes a different approach to achieving the same end (14,15). Like its film-based predecessor, digital tomosynthesis generates multiple planar images at arbitrary depths within a patient. A series of discrete projection radiographs are acquired with digital detectors while the x-ray tube and beam are pivoted through a small angle. Through manipulation of the projection images to remove the over- and underlying planes, high-quality images of each selected tomographic plane can be obtained with section thicknesses ranging from 1 mm to several centimeters (Fig 1). Tomosynthesis is a type of limited-angle tomography and has relatively poor resolution in the depth direction. The level of detail in the imaging plane, however, can be superb. Digital tomosynthesis has been applied to chest, breast, angiographic, orthopedic, dental, and other kinds of examinations.
|
|
|
In axial-mode nonhelical CT systems, the tube is rotated once about an immobile table; the table is then advanced by a small amount in the z-axis direction, and the tube is rotated once in the opposite direction. This procedure is repeated over and over, with each rotation producing data for a flat plane of tissue. With helical CT, also known as spiral CT, the table and x-ray tube both move continuously throughout data acquisition. The technology is either so-called third generation, with a rotating array of detectors, or fourth generation, in which a thousand or so very small detectors surround the patient and are fixed in space. Slip rings are needed to deliver electric power to the tube and, for third-generation machines, to retrieve the signals from the detectors (Fig 3). Helical CT allows much faster scanning and provides images with higher resolution in the transverse plane.
|
|
Multi–detector row CT scanners use classic third-generation scan geometry but with a cone- rather than a fan-shaped x-ray beam. For 16 or fewer sections, it is possible to employ well-established multi–parallel-section approximate reconstruction algorithms. More sophisticated approximate and exact cone-beam algorithms have been created and yield better images, but they are computationally more complex and time consuming (19–22).
CT scanner manufacturers continue to strive for greater scanning speed and more longitudinal anatomic coverage with each gantry rotation. Because the centrifugal forces would become overwhelming, the speed of gantry rotation will probably not increase much beyond its present value of three cycles per second. This is good enough to examine the lung (23,24) or capture a heart at systole or diastole with little motion blurring, but shorter scan times are clinically desirable and might be achievable with multiple tubes. So, too, is z-axis coverage, to allow functional imaging of the heart, lung, brain, and other structures. These, and the development of better detector systems, will further improve resolution in the transverse dimension, so that CT is becoming a truly isotropic modality, similar to MR imaging. A difficulty with this, of course, is the increase in Compton scatter radiation reaching the image receptors. Indeed, one of the touted benefits of single-section scanning, in the early days, was the virtual absence of scatter—and the best approach to dealing with it, now that it has returned with multisection imaging, is not obvious (25,26).
Researchers at the University of Aachen, Germany, have reported on a device in which the multiple rings of detectors have been replaced with a flat-panel imaging array (26,27). Several manufacturers have already produced flat-panel prototypes, and quasi-CT C-arm devices have been marketed. However, there are daunting problems associated with the very high rates (gigabytes per second) of data transfer.
With dual- or multienergy techniques, moreover, CT could provide information on tissue parameters beyond simple pixel-average relative attenuation coefficients (ie, Hounsfield units).
Revolutionary when it was introduced in the mid-1980s, electron-beam CT can produce a transverse section in as short a time as 50 msec, fast enough to freeze the beating of the heart in any phase of the cardiac cycle, and at a relatively low dose—about 1 mSv for angiography, compared with 10 times that with 64-section CT—and even less for coronary calcium scoring (Fig 4). Electron-beam CT images tend to be noisy, however, and the far more versatile 64-section CT has better resolution in all three directions (about 0.3 vs 1.5 mm for electron-beam CT). Fast MR imaging can also support cardiac screening or image-guided intervention and can do so without ionizing radiation.
|
|
While biomedical informatics is perhaps best known for its role in untangling the vast quantities of data coming from work on the genome, another area of considerable interest is the need to display the information content of hundreds, perhaps even thousands, of CT sections per patient (with numerous such patients each day) in diagnostically useful ways. A radiologist clearly cannot examine them all one by one (despite the possibility that a critical clinical sign may appear faintly in one of them—with serious legal implications). Conversely, the use of a helical 64-section machine just to create a smoother surface rendering of some organ would rarely be justified—although in some areas that require very high resolution, such as CT mammography or CT virtual colonoscopy, it might be justified (29). So, it is essential to develop computer-based bioinformatics programs that can locate, and perhaps interpret, the few clinically essential visual items contained within such an overabundance of imaging data (30).
Finally, there is strong pressure to reverse the upward "radiation dose creep" for all x-ray technologies, especially for CT in children and for mass screening applications. Fortunately, most operators of the machines appear to be increasingly aware of the need to tailor the technique factors to the dimensions of the patient, which can reduce the dose dramatically (31,32).
Gamma Ray Imaging
In standard nuclear medicine, images are produced by using radiopharmaceuticals and a gamma camera (33). A radiopharmaceutical consists of two parts: (a) an agent designed to concentrate, through chemical or other physiologic means, in a specific organ or compartment of the body and (b) a radioisotope (most commonly, technetium 99m), which emits a medium-energy gamma photon. While a nuclear imaging study cannot compete with radiography or CT in revealing precise anatomic detail, it may more than compensate by providing potentially invaluable information about the physiologic status of an organ or other tissue.
Radiopharmacology continues to evolve rapidly, and it is to be expected that scintillation-detector materials for gamma image receptors will continue to be improved and that, at some point, either the photomultiplier tubes or the entire scintillation crystal–photomultiplier tube assembly will be replaced by sensitive, low-noise, solid-state detectors, perhaps of the flat-panel imager variety.
SPECT, the gamma ray counterpart to CT, has become nearly indispensable for the assessment of myocardial perfusion and other cardiac functions. More than half of SPECT studies are for coronary artery disease, another quarter are for bone, and the rest are for brain, prostate, thyroid, and other organs (34,35).
Likewise, PET has long enjoyed a prominent role in neurologic research. More recently, PET has been used as a primary clinical tool for the detection, localization, staging, and monitoring of malignancies (34,36–38). PET makes use of radionuclei that emit positrons, the positively charged antiparticles of electrons. Positron emitters include oxygen 15, nitrogen 13 (13N), carbon 11, rubidium 82, and, the most extensively used one fluorine 18 (18F), which is normally employed in the form of the glucose analog fluorodeoxyglucose. A positron travels a few millimeters in tissue and collides with an atomic electron. The particles annihilate one another and create a pair of 511-keV "annihilation" photons that fly off in almost exactly opposite directions. Only events that trigger two of the PET device's detectors on opposite sides of the patient and in coincidence (nearly simultaneously) contribute to the creation of the PET image. The operation of opposed detectors in coincidence improves both the spatial resolution and the signal-to-noise ratio of the images.
The number of clinical PET and fusion PET/CT procedures in the United States has grown from about 250 000 in 2001 to roughly 900 000 in 2004. Ninety percent of the studies are performed in search of tumors; the remainder are split between cardiac and neurologic studies. It is possible that, because of its better resolution, physiologically more interesting agents, and decreasing costs, PET will come to displace SPECT and conventional nuclear medicine for many of the latter modalities' current roles.
The fusion of SPECT or PET images with CT or MR images (39,40) allows the projection of visual physiologic information on a detailed anatomic map background so that tissue function can be correlated directly with tissue structure (Fig 5). With fusion, moreover, the CT attenuation data also make possible substantial correction of the PET or SPECT image for attenuation of gamma rays within the body. It has even proved useful in radiation treatment planning to merge CT and MR images (Fig 6).
|
|
|
|
|
Meanwhile, advances in molecular biology are yielding receptor-specific radiopharmaceuticals and other pharmaceuticals that are opening up a new realm of opportunities for in vivo imaging at the molecular and cellular scale. Research in radiopharmacology and in the technologies for producing radionuclides are areas of importance but of uncertain near-term funding.
MR Imaging
A hydrogen nucleus (proton) in a water or lipid molecule in a cell acts rather like a spinning positively charged ball. As with any other moving charged body, it produces its own local magnetic field, somewhat like that of a compass needle.
When placed in a strong (eg, 1.5- or 3-T) external magnetic field, a proton tends to settle down comfortably into its ground state, also like a compass needle, with its spin axis and local field aligned along that of the external field. But a proton follows the dictates of quantum mechanics, not of common sense, and it can also reside temporarily in its metastable, higher-energy spin state, pointing in the "wrong" direction. Indeed, a group of protons can be intentionally flipped or twisted upside down and into their higher-energy spin states in a process known as nuclear magnetic resonance through the absorption of a pulse of radiofrequency energy of the correct (ie, Larmor) frequency and duration.
The spins subsequently experience naturally occurring Larmor frequency magnetic noise, which tickles them back down to their lower-energy configuration. This relaxation occurs rapidly if the water molecules are tumbling and interacting with one another and with their environments in such a manner that, at least intermittently, they are rotating at the Larmor frequency. In this condition, each proton experiences the magnetic field from its partner as varying, at least briefly, at the Larmor frequency, which causes a downward spin transition. The average spin-relaxation time for this transition is called T1. Free water tumbles much faster than the proton Larmor frequencies for the external fields employed in MR imaging, so T1 is long (4000 msec or so); but for water bound to intermediate-sized biologic molecules (which themselves may happen to be tumbling at the Larmor frequency), T1 can be an order of magnitude shorter.
MR imaging involves manipulation of pulses of radiofrequency energy and magnetic field gradients to assess, in effect, T1 and the related parameter T2 at each point in the part of the body being examined (41,42). MR imaging is of diagnostic value because it can be used to generate two- or three-dimensional maps that reflect the spatial distribution of the values of T1 and T2. But the proton spin-relaxation times depend largely on the degree of binding of the water molecules (in which the proton spin flips are occurring) to the nearby biologic molecules, and the water–biologic molecule interactions in a tissue are sensitive to its histologic characteristics and also to its physiologic status. So a T1- or a T2-weighted MR image can reveal information on both the anatomy and the state of health of tissues. Similar considerations apply for hydrogen atoms in lipids.
T1 and T2 can be altered in clinically useful ways with MR contrast agents, many of which are built around paramagnetic gadolinium ions caged within molecules of a chelating agent (43). Dynamic contrast enhancement and related techniques, where the proton spin-relaxation times in a region are monitored during the administration of a bolus of contrast agent, can reveal much about the microvascular environment.
MR imaging will continue to find new areas in which to expand (44). It has already spun off MR angiography (a competitor with CT angiography) (45–48) and MR microscopy (49). Also, single-wire or other radiofrequency antennae that can be positioned within vessels, body cavities, and elsewhere may provide gateways to new types of studies of the heart and other accessible organs. Diffusion MR imaging, which is becoming more widely used in the assessment of acute stroke and malignancies, is able to reveal information (the diffusion tensor) about the actual directions of blood diffusion (50,51).
MR spectroscopy and the closely related modality of chemical shift MR imaging can elucidate chemical processes ongoing in small volumes of tissue (52). As with fusion PET/CT, MR spectroscopy is being combined with high-resolution MR imaging to enhance sensitivity and specificity in the detection and identification of cancers and other abnormalities (Fig 7). And MR spectroscopy of carbon 13, fluorine 19, and phosphorus 31 (rather than hydrogen 1) is also providing valuable insights on cellular metabolism and biochemistry (53).
|
|
|
|
|
Functional MR imaging, in particular blood oxygen level–dependent imaging, makes use of fast pulses to determine where the flow of oxygenated blood is unusually high and, by inference, where neurons are unusually busy. This method provides information on neural activity that is complementary to contributions from MR spectroscopy and PET (59,60). In a study of potentially staggering implications, functional MR imaging allowed researchers to determine which of several visual patterns that subjects were looking at—in effect, it allowed researchers to read the subjects' mind (61–63).
The nuclear MR signals that underlie MR imaging can be influenced by the magnetic coupling of nuclear spins that are far apart—even millimeters apart. The possibility of such an interaction, previously ignored because of its extreme weakness, has now given birth to a new branch of MR known as zero-quantum imaging (64). Still in the early developmental stage, zero-quantum imaging may be unusually sensitive for the detection of malignant tumors (Fig 9).
|
|
|
US Imaging
US produces images from high-frequency sound echoes that are created at boundaries between tissues with different elastic properties or densities (which, in turn, determine the speed of sound and the acoustic impedance for a medium). With its ability to do so in everyone from fetuses to the elderly and with no exposure to ionizing radiation, US is indeed a "womb-to-tomb" clinical imaging modality.
Over recent years, the evolution of diagnostic US has been quietly spectacular (67–69). There have been major strides in the design of miniaturized piezoelectric and capacitive transducers; contrast agents (which, at present, consist of suspensions in a fluid of erythrocyte-sized gas bubbles that may be coated with a tissue-targeting compound) (70,71); the exploitation of harmonic information, which is associated with higher frequency components created from nonlinear propagation of sound through tissues or nonlinear oscillations of contrast agent gas bubbles (72); the use of coded-excitation pulse trains and matched filters; image-processing software; techniques based on the absorption or speed of ultrasound radiation, rather than just on reflections; and small size and portability. These advances allow, with two-dimensional arrays of piezoelectric elements, the creation of real-time images that appear truly three-dimensional and that display high spatial, contrast, and temporal resolutions over a large field of view (73,74).
Clinical US systems normally operate in the 2–15-MHz range. Despite the technical difficulties, devices are now being designed with higher frequencies, with the shorter wavelengths allowing the imaging of correspondingly smaller objects. Micromachined US transducers, consisting at present of tens of separate piezoelectric or capacitive elements a few millimeters across—fine enough to fit within a narrow-gauge catheter and then into a coronary artery—are being built by way of microelectromechanical systems semiconductor manufacturing technologies (75). Vascular imaging is currently being performed in the 10–40-MHz range; at even higher frequencies (up to several 100 MHz), there are potential applications in ophthalmology, dermatology, and perhaps cellular-level imaging.
There are also early but promising results for true four-dimensional tomographic US imaging (Fig 10). US tomography may require hundreds of transducer elements covering a large-angle area of the body and perhaps firing and gathering echoes separately—although it is always possible that a much simpler solution to four-dimensional image acquisition may be found. Transmission US tomography has been demonstrated in relatively homogeneous tissues such as the female breast, where (as in refraction seismography, used by geophysicists in the search for oil and gas) variations in the speed and/or absorption of sound are measured and processed to create images. Reflection US tomography, with its multiplicity of omnidirectional echoes, continues to pose major technical problems that remain intractable at this time.
|
A new area of physics, known as time-reversed acoustics, provides the means to detect a sound wave from a submarine, for example, and theoretically time-reverse the sound wave and send it back to its origin. This ability to locate the source may lead to novel ideas for both diagnostic and therapeutic applications of ultrasound (78).
Finally, there is active research on biologic effects (79), especially as US applications move to much higher frequencies. This is ongoing not only to ensure patient safety in diagnosis but also to support novel applications in therapy (80).
|
DEVELOPING AND EXPERIMENTAL TECHNOLOGIES |
|---|
|
TOPABSTRACTINTRODUCTIONADVANCES IN CURRENT TECHNOLOGIESDEVELOPING AND EXPERIMENTAL...EVOLVING ROLE OF COMPUTERSCONCLUSION: IMAGING WITH A...ESSENTIALSReferences |
|---|
Gamma rays and x-rays reside at the top of the useable electromagnetic spectrum, and those fields are well plowed, as already indicated. Heading downward in energy, after meandering through the ultraviolet region (where currently there are few applications), one comes to visible and near-infrared light.
Optical and Near-Infrared Imaging
Diaphanography, also termed transillumination, is the medical counterpart to shining a bright flashlight through the hand to display the bones. As with x-rays, beams of visible and near-infrared light are differentially attenuated as they transit different types and thicknesses of tissues, and their shadows can be recorded as images on film or with an electronic camera. Transillumination of the breast, for example, can sometimes help distinguish benign from malignant masses—but this modality produces more false-positive and false-negative results than do conventional mammography and US, and it has not been accepted as a standard clinical tool.
The clinical utility of optical methods depends largely on the ability to extract image information about objects embedded in turbid media, so there have been serious efforts to reduce the level of scattered radiation reaching the image receptor. Improvements in time-resolved spectroscopy techniques and in the theory of photon transport in tissues have both contributed to recent advances in the field. In the former category, one family of methods employs gated detectors that are fast enough that, following an extremely short laser pulse, only unscattered photons (which travel directly from source to detector and arrive first) are accepted. This approach can display tumors and other abnormalities by recording light transmission as a laser is stepped through a region of interest—a much more refined version of diaphanography.
Laser optical tomography or, for one particular application, tomographic laser mammography, yields cross-sectional images of tissue obtained by projecting laser beams inward from many directions. These images are useful in studies of blood perfusion, tissue oxygenation, and neovascularity in the brain, breast, and extremities. In one interesting variation on this idea, photographic-acoustic tomography makes use of differences in the tendencies of tissues to absorb brief radiofrequency pulses of laser light and to heat up and thermally expand extremely rapidly, thereby producing ultrasonic waves that can be detected with sensitive piezoelectric receptors.
With a somewhat different approach, confocal scanning laser tomography can be used to noninvasively acquire three-dimensional images of the posterior segment of the eye, creating a quantitative description of the optic nerve head and the surrounding retinal surface (Fig 11). A laser beam is focused to some depth within the eye and scans a two-dimensional plane. Only light from that focal plane is allowed to reach the detector. A sequence of such two-dimensional optical planar views is acquired for increasing depths of the focal plane, and the result can be displayed as a three-dimensional topographic image of the optic nerve head and peripapillary retinal nerve fiber layer. The approach is useful not only for detecting abnormalities directly but also for tracking subtle changes over time that would be difficult to detect clinically. Authors of a recent study (81) suggest that, with expert interpretation, the power of the technique as an aid in the diagnosis and management of glaucoma is comparable to that of the accepted standard of stereofundus photography.
|
Diffuse optical tomography provides measurements of hemodynamics and neural activation at depths of several centimeters in tissue. Nonlinear microscopy employs nonlinear optical methods, such as multiphoton molecular excitation, optical harmonic generation, and depletion of stimulated emission. Nonlinear microscopy can be used to image subcellular morphology and trace molecular dynamics at subnanometer resolution at depths of up to a fraction of a millimeter in living tissue.
Tissue is nearly opaque throughout most of the infrared, visible, and ultraviolet parts of the electromagnetic spectrum. One way to deal with this challenge is to use a very intense laser beam, as in diaphanography. Another approach is to exploit a narrow window of transparency in the near-infrared region, with wavelengths in the 700–900-nm range. An important set of applications makes use of the fact that oxygenated and deoxygenated hemoglobin both absorb these wavelengths well but have spectra that differ enough to allow distinction between the two species. This distinction can be important in studies of blood flow and oxygen consumption.
Each year, 1 million women undergo core-needle breast biopsy in the United States. While considerably easier on the patient than excisional biopsy, the core-needle technique suffers from a false-negative rate of up to 7%, despite needle guidance with x-ray fluoroscopy or US. A more sophisticated type of probe currently under development may reduce the false-negative rate substantially (83,84). After the probe has been guided to the tumor site within the breast, it transmits near-infrared laser light into the breast tissue and then senses the light coming back from the tissue. The light returning from a tumor may differ from that returning from normal tissue because of irregularities in the degree of oxygenation, fluorescence characteristics, or other properties. In cases where no difference is detected, the tip of the probe can be repositioned in further search for disease. On finding tissue that yields a difference in returning light, the probe is used to acquire a core biopsy specimen. Unlike mammograms, the laser probe works well in dense tissue, such as in the breasts of young women, and may find applications in other organ systems, as well.
Terahertz Imaging
The terahertz portion of the electromagnetic spectrum lies between 300 and 100 µm in wavelength. So-called terahertz rays, or T rays, do not penetrate water or tissue well, so they would be of little use in the examination of deep-seated tissues. A substantial fraction of cancers lie in the epithelium, however, and while many of these are readily apparent to the trained eye, some that are small and flat can be overlooked or are simply not visible. Standard modalities are not adept at depicting or characterizing epithelial tumors—but terahertz imaging, because of the ability to recognize the spectral fingerprints of surface proteins that are markers for certain cancers, seems capable of demonstrating them at an early stage when they can still be treated effectively (Fig 12). Terahertz rays are nonionizing and offer imaging resolution of less than a millimeter; the equipment is safe and portable. There are major difficulties with terahertz sensing and imaging systems: They are inefficient and extremely costly to produce, but recently created photonic band-gap materials may soon radically change that situation (85–87).
|
|
Electron-spin resonance imaging, also known as electron paramagnetic resonance imaging, is analogous to MR but involves spin transitions of unpaired electrons rather than of (unpaired) protons (60,89). Since an electron is three orders of magnitude lighter than a proton, it has a considerably greater magnetic moment and a correspondingly higher Larmor frequency. At 1 T, an electron's magnetic resonance occurs at 28 GHz, rather than at 42 MHz for a proton (1 GHz = 1000 MHz). Microwave energy tends to be strongly absorbed in the process of excitation of rotational states in water—indeed, that is how microwave heating works—and at gigahertz frequencies, microwaves are absorbed in a millimeter of skin. At considerably lower resonance frequencies and field strengths (eg, 0.01–0.04 T), however, the penetration permits imaging of free radicals in vivo in small animals. A variation on the theme, known as proton-electron double-resonance imaging, or PEDRI, involves the simultaneous performance of electron-spin resonance and nuclear MR imaging (Fig 13) (90).
|
|
The warmer something is, the more heat it emits, at a rate that is approximately proportional to the fourth power of the absolute temperature (ie, approximately T4). Like some night-vision techniques, emission thermography has long been used to sense infrared radiation (which has a lower frequency than visible light) emitted through the skin as a result of heat brought to the surface by blood flowing from deeper regions. Irregularities that affect blood flow in the outermost few millimeters of the skin or that influence its temperature directly (eg, angiogenesis and neovascularity that accompany breast disease) can be detected by means of an infrared camera.
While there is not a great deal of evidence supporting the relative value of thermography in helping search for breast tumors, it has proved to be effective in monitoring of inflammatory conditions such as rheumatic disease, injured muscles (Fig 14), diminished enervation of muscles, burns, and frostbite (91).
|
With the traditional approach, a few electrodes attached to the skin surface record the faint voltages produced when the heart "fires" or an area of the brain is stimulated. At present, it is the shapes of the voltage pulses as they change over time that are clinically useful. Arrays of many electrodes spread over an area of the patient may eventually succeed in generating useful information in image form, as well. One major problem is the rapid attenuation of electric signals in the body, so that those that do reach the electrodes are very weak and difficult to extract from background electric noise.
Magnetoencephalography and Magnetocardiography
Like any other electric currents, those in and between neurons generate weak magnetic fields. Magnetoencephalography is similar to electroencephalography, except that the former is used to detect and monitor the weak varying magnetic (rather than electric) fields produced in this fashion (92–95).
Magnetoencephalography measures such magnetic fields with detectors positioned around the head and back-calculates the locations and strengths of the neural currents that gave rise to the fields. These fields are about 1 billionth as strong as that of the Earth, so near-perfect magnetic shielding and highly sensitive detectors are required. Fortunately, the exquisitely sensitive superconducting quantum interference device, or SQUID, fills the bill nicely: It is responsive enough to detect the movements of a small magnet a quarter mile away. With millimeter and millisecond spatial and temporal resolutions, magnetoencephalography demonstrates neural activity directly, as it takes place; functional MR imaging and PET, in contrast, represent indirect assessments of brain function, because they show relatively slow variations in metabolic activity. Magnetoencephalography has already proved to be clinically useful in noninvasive studies of epilepsy, migraine headache, and diabetic coma; and like PET and functional MR imaging, it can follow the brain's response to stimuli. It can be used to demonstrate subtle abnormalities in a region of brain that appears to be structurally and physiologically normal when viewed with the aid of other modalities (Fig 15). Magnetocardiography works the same way as magnetoencephalography, but it tracks the much stronger magnetic field changes that accompany the beating of the heart.
|
Also being explored, as an alternative to electroshock therapy, is the converse process of magnetic nerve stimulation, in which a pulse of electric current flow is generated in a well-localized region of the brain by rapidly changing the field from a powerful external magnet (96).
Tissue Electric Impedance Tomography
Electrons and ions in the fluids in and between cells can move about, to a greater or lesser extent, in response to an applied electric field. The standard measure of the resistance of tissue to this movement of electrons is referred to as the tissue's electric impedance. To assess the impedance (Z) of tissues in a region, numerous electrodes are attached to the skin, low voltages (V) are applied between pairs of electrodes, and the electric currents (I) produced are assessed; Ohm's law (Z = V/I) then yields the impedance. After many such measurements, electric impedance tomography uses CT-like reconstruction calculations to generate three-dimensional maps of tissue impedance throughout the region (85,97,98).
A closely related alternative for data acquisition, known as applied potential tomography, involves application of voltages at two or more points in the tissue. The resultant voltages are then measured elsewhere. Yet another approach is to place the region of interest in a strong magnetic field, after which passage of a high-frequency alternating current between electrodes generates a detectable ultrasound signal.
Like MR imaging, tissue impedance imaging is sensitive to changes in the water content of tissues. Indeed, tissue impedance imaging is being developed by the military to monitor blood loss from wounded soldiers. Other applications include studies of gastric emptying, pulmonary ventilation, cardiac output, blood flow, and intracranial hemorrhaging in infants. Since malignant tumor tissue may be an order of magnitude or so more conductive than surrounding normal tissue, instruments have been built to facilitate the assessment of ambiguous breast masses.
Complications in impedance imaging are that only a finite number of electrodes can be placed on a given surface area and that electric currents (unlike x-rays) do not follow straight lines. On the other hand, electric impedance tomography is safe, can be used at the bedside, allows for long-term data acquisition, costs little, and images a new physiologic function that may turn out to be uniquely revealing for some disorders.
Cellular and Molecular Imaging
Molecular imaging refers to the minimally invasive in vivo sensing, depiction, and characterization of spatially localized biologic processes at the cellular and molecular levels (99). Having evolved out of conventional nuclear medicine and still largely in an active preclinical stage, molecular imaging now represents "the convergence of multiple image-capture techniques, basic cell/molecular biology, chemistry, medicine, pharmacology, medical physics, biomathematics, and bioinformatics into a new imaging paradigm" (100).
The development of molecular imaging is attributable, in part, to the synthesis of novel molecular agents that attach with high specificity to genes, proteins, or other biomolecular targets and that can then be detected with various imaging modalities. Among these are agents that are activated only in response to particular changes in the local biochemical environment, such as specific gene expression or enzyme activity (43,101).
Molecular imaging is being driven also by the design of micro-CT, micro-MR, micro-PET, optical bioluminescence, fluorescence molecular tomography, and other methods for the study of small (approximately 70-g) laboratory animals (Fig 16): The corresponding clinical machines designed for 70-kg humans are scaled down—often with improved image quality and almost always at substantially reduced cost. The image receptor for micro-CT, for example, can be a commercially available "off-the-shelf" charge-coupled device (102,103).
|
|
Fluorescent lifetime imaging microscopy, for example, exploits the dependence of the fluorescence decay time on tissue type to help identify abnormalities (105). Another approach makes use of quantum dots, or "qdots," which are single fluorescent semiconductor nanocrystals only nanometers in diameter (106–108). At such sizes, a nanosystem begins to act somewhat like a large molecule or a set of interacting large molecules, with quantum mechanical properties quite different from those of a bulk piece of the same crystalline or amorphous material (eg, as in a transistor). Because of quantum confinement effects, the narrow laser-stimulated emission spectra of quantum dots, in particular, depend on the size of the dots, which can be precisely controlled during synthesis. When quantum dots emitting various colors are conjugated to antibodies, antibiotics, or other recognition molecules and are then injected intravenously, they can help detect and map out a number of specific cellular markers of diseased tissues simultaneously—a form of multiplexed in vivo optical biopsy. Also, MR, PET and CT contrast agents can be grafted to their surface so that they, too, can provide molecular imaging diagnostics. It might even be possible to attach therapeutic agents to quantum dots, which could be activated once in proper position.
Considerable numbers and varieties of molecular imaging systems are being investigated and, as with any new field, are undergoing rapid growth. Few are similar to one another in design. A plethora of recent publications describe these systems, including a recent issue of Medical Physics, which contains abstracts from a major symposium on the subject (109). Topics include, in addition to the ones discussed above, diffuse optical tomography (110), fluorescence resonance energy transfer, Fourier transform infrared spectroscopy, autofluorescence imaging, hyperspectral imaging, video capillaroscopy, and others.
Mini- and Nanotechnology
Small-scale technology offers exciting possibilities for medical imaging. There already exists a disposable video camera–transmitter–light source combination the size of a pill that sends color images to a receiver-recorder (strapped to the waist) as it wends its way through the bowel. Compared with endoscopy, limitations of the system are related to the inability to control the position of the camera and its direction of view and battery lifetime. These challenges, however, should be surmountable.
At the finest level, nanoparticles (eg, quantum dots) that display specific physical or chemical characteristics (such as paramagnetism or stimulated fluorescence) or objects coated with nanoparticles may be detected with appropriate imaging technologies (111). Ceramic, metallic, and organic nanotubes, nanowires, and other structures have been the subject of intense investigation throughout the decade, as have been (more recently) nanostructures such as minute sensors, motors, or other electromechanical devices (which might contain biologic components) that could enter a blood vessel and be directed to an area of interest, where they could acquire and transmit images; implement enzyme-, antibody-, or genome-based biologic tests; perform biopsies; build tissuelike materials and organs; and even carry out small-scale surgeries. While some might find this all to be a bit too "Star Trek–ish", others would argue that it is already happening (112,113). The National Cancer Institute, in particular, has recently announced a 5-year, $144 million program to develop nanotechnology in cancer detection and treatment.
|
EVOLVING ROLE OF COMPUTERS |
|---|
|
TOPABSTRACTINTRODUCTIONADVANCES IN CURRENT TECHNOLOGIESDEVELOPING AND EXPERIMENTAL...EVOLVING ROLE OF COMPUTERSCONCLUSION: IMAGING WITH A...ESSENTIALSReferences |
|---|
The amount of visual and other medical data being generated at the levels of molecules and cells, tissues and organs, and individuals and communities is populating vast databases at an ever increasing rate. The relatively young discipline of biomedical informatics is beginning to provide tools to search for, identify, retrieve, integrate, analyze, model, display, store, communicate, and manipulate all this information (117–119). The roles of informatics in management of the data from large numbers of CT sections and in searches for diagnostically meaningful patterns in images have already been mentioned. More generally, researchers in the field will continue to develop ways to integrate and search effectively through the databases, which are growing in number, size, and complexity, that are needed for clinical care and for research. There are related applications, such as when there is a crisis on a spaceship halfway to Mars or after detonation of a nuclear weapon in a city, where rapid access to essential medical information might prove to be invaluable.
PACS, DICOM, Computer-aided Detection, and More
Increasing numbers of medical centers are adding information management and communication systems, in which images of concern are stored and manipulated in the PACS. Ideally, a PACS is integrated with the radiology information system and the hospital information system as part of an integrated health care environment. That way, the electronic health record or integrated clinical record and much other relevant clinical information about a patient can be rapidly available wherever the network infrastructure allows. PACS got off to a shaky start years ago, largely because early systems were overhyped and turned out to be user unfriendly. With simpler user-PACS interfaces, better and more standardized PACS-machine connectivity such as that brought about by the Digital Imaging and Communications in Medicine (DICOM) 3.0 protocol (120), more effective training provided by vendors, better security, lower purchase and operational costs, and the availability of mini-PACS packages that can be upgraded as physicians choose to make fuller use of them, PACS may soon be considered indispensable for the provision of high-quality cost-effective health care (121–123).
A PACS generally has teleradiology capability that allows medical personnel to respond to images on any compatible workstation, even at home; to communicate and share images effortlessly with other specialists; to obtain input from around the world on difficult cases; and to provide diagnostic expertise in remote areas. The next step, telesurgery, in which the practitioner can be thousands of miles from the robot holding the scalpel and needle, is actively being taken. The Internet is already providing an invaluable conduit for all such telemedicine services, and further applications will evolve as the more powerful and versatile next-generation Internet and Internet2 become increasingly accessible. (The next-generation Internet protocol, or IPng, which is a new version of the current Internet protocol, or IP, is designed to deal with the addressing and routing problems arising from the massive size and continuing expansion of the system. Internet2 is a not-for-profit consortium of universities and companies that are collaborating on these, and related, issues.) The medical value of next-generation Internet and Internet2 depends, of course, on there being in place agreements on a medical information (or more general) protocol to define interface uniformity, data formats, semantics, and all the other details.
The advantages to placing all patient-related information into Web-based interoperable data sets that link physicians, hospital medical records, and laboratories may seem self-evident. For most clinics and medical centers, however, this idea remains a remote dream. The health care industry invests only 2% of its gross revenues in information technology, compared with other information-intensive industries (eg, finance, utilities, communications), which typically spend a five-times greater fraction (124,125). Furthermore, clinical information technology systems that do exist often cannot readily connect with one another.
The impact of misinformation on patient safety is a closely related, huge issue. The Institute of Medicine has estimated that 44 000–98 000 people die annually because of medical errors in the United States, making this the eighth leading cause of death, ahead of acquired immunodeficiency syndrome and breast cancer.
An important and expanding role for small high-powered computers is in pattern recognition and computer-aided detection, or CAD. Computers excel at highly analytic jobs that involve logic and memory, such as playing chess or calculating just about anything. But even the most "clever" computer has great difficulty with the elementary tasks of everyday life that require intuition, judgment, or common sense, such as recognizing a face with a new beard or distinguishing a liver with an abscess from one that is normal but just looks odd.
Still, there have been great strides in machine learning. For example, neural networks and other species of computer programs can improve their success at performing certain tasks by "learning" from past experiences. Machine learning, pattern recognition, and statistical inference programs began with the detection and characterization of a point, a straight line, or a circle on a blank background. Programs that search for irregularities in mammograms, chest CT sections, US scans of the liver, and other images are much more complex and sophisticated, but the basic concepts are much the same. Some successes in computer-aided detection have already found clinical application and are tantalizingly suggestive of even more wonderful things to come (126–128).
Now that Medicare reimbursement is available, mammographic computer-aided detection is being used with both digitized and digital images as a sharp-eyed and untiring second reader to pick out subtle cancers, thereby reducing the false-negative rate. It is not clear how quickly computer-aided detection will progress to full computer-aided detection and computer-aided diagnosis, or CADx, in which the system offers a statistically informed second opinion—but a number of people involved in the development of computer-aided diagnosis believe that the writing is on the wireless flat-panel monitor on the wall.
An important aspect of this effort will be the continued development of meaningful values of merit for the "quality" and clinical utility (which are not necessarily the same thing) of various types of images. Resolution may be quantified as, say, the minimum separation between two tiny dots or thin lines that can just be distinguished from one another visually or by computer and more fully in terms of the modulation transfer function. Similarly, image contrast and noise have their own quantitative measures, such as the detective quantum efficiency. But while resolution or contrast or noise alone may be critical in particular diagnostic tasks, they all interact to affect the quality and utility of most types of medical images. Viewing conditions and the training, experience, eyesight, and frame of mind of the interpreting physician, as well as relevant clinical information accompanying the image, can also individually and collectively affect the diagnostic process.
There are ongoing attempts to find ways to combine these factors into mathematic functions that gauge diagnostic efficacy, however it is defined, for various medical situations (129–131). With such measures of overall clinical effectiveness, it would be possible to modify the design parameters of an imaging or pattern-recognition system until this function achieved a local or global maximum value in the appropriate measure of diagnostic usefulness. Receiver operating characteristic curves, for example, have long been used in the judgment of overall performance of a technology (131–133), and it is to be expected that investigators will continue searching for ways to deal with their well-known limitations in the clinical setting. Receiver operating characteristic curves are largely empirical in nature, rather than derived from first principles, and are restricted in what they can offer in basic development work.
Diagnostic efficacy is a fascinating and highly promising topic of research in which physics and engineering meet psychology and clinical science and evaluation on an equal footing. But it is likely that much more research remains to be done before anything very substantial and practical emerges in this area.
Computer Technology
The rapid evolution of basic computer technologies will continue to have a tremendous effect on imaging. One ever-growing area of influence will undoubtedly be in systems for image display. For reasons of image quality, weight, size, required voltage, ruggedness, energy efficiency, and cost-effectiveness, flat-panel display technology is rapidly displacing the traditional cathode-ray tube, not only in medical displays but also in computer monitors and high-definition television sets (134–136). At present, liquid crystal panels with several million pixels account for something like 99% of all flat-panel displays produced, but other approaches, such as plasma, field emission, and electroluminescent panels, are also under development. Flexible, rewritable, plastic-based electronic "paper" is already available—it remains to be seen what applications arise for it in medical imaging.
True three-dimensional viewing, moreover, is now a reality. The reader may have seen stereoscopic drawings and movies years ago in which two slightly different perspectives of an object were presented in blue and red; when one wore special glasses with one red and one blue lens, the spears and arrows really did seem to come right out of the screen. Now there are electro-optical glasses in which the lenses themselves can display slightly different electronic images for the two eyes—the images are perceived to be fully three-dimensional and can be of value in virtual reality applications such as surgical planning or virtual endoscopy. Holographic displays, in which the viewer does not even need glasses, are in the early stages of development.
The boundary between computation and communication is blurring. As of early 2005, a half billion cell phones are filling the airwaves with chatter. Already, 10% of them are smart cell phone–computer hybrids that cruise the Web, support complex games, and send and receive e-mail, text messages, and images. This wireless technology is already nosing its way into the clinic in the form of systems for computerized physician order entry and real-time verbal updating of patient charts by way of tablet personal computers that incorporate dictation and transcription and error-prevention software.
Computers themselves will keep on getting faster, smarter, and less costly. Today, a fast new desktop personal computer may be able to manage a billion operations in a second. The most advanced supercomputers operate more than 1000 times faster, and there are plans for future machines that will perform a quadrillion (a thousand trillion) operations per second and draw on a quintillion (a million trillion) bytes of mass storage in doing so.
The most advanced computers play increasingly central roles in circuit design, and Darwinian (ie, "genetic" programming) methods are being developed to encourage their rapid, and sometimes nonobvious, paths of evolution (137). New machines may come into being that function in desirable ways but that will conceivably be too complicated for humans to understand. And these machines will, in turn, design future generations of computers, and so on, and so on—an interesting prospect, and not the stuff of science fiction.
Beyond producing ever bigger and faster computers, some researchers at the cutting edge are taking the first tentative steps toward creating computing machines that are different in kind, not just in degree. Rather than sending currents of electrons along routes determined by the settings of silicon-based switches, a DNA computer, for example, would operate by snipping and joining segments of DNA in solution (138). Such a computer would consume a billion times less energy than its solid-state siblings, require a trillionth of the space, and perform some kinds of computations that are otherwise impossible now.
At the same time, rudimentary quantum computers are being explored that operate according to the counterintuitive quirky rules of quantum mechanics rather than the linear, "if-this-then-that" Boolean logic of everyday life (139). While digital computers operate, ultimately, with electronic switches that can be in either the "on" or the "off" state, quantum devices would manipulate states that are combinations of the two. Like DNA devices, they would provide calculational powers currently undreamed of but—also like DNA computers—probably not for a decade or more.
DNA and quantum computers will act somewhat like massively parallel conventional computers—in effect, performing gargantuan numbers of relatively simple computations with blinding speed. They will be able to crack certain mathematic problems in minutes that would give today's computers centuries-long nightmares, such as separating a thousand-digit number into its two prime-number factors—a task that is of critical importance in making and breaking secret codes. But they may also be relatively inflexible and, as with supercomputers, challenging to provide software for and to program; still, all it might take is one good idea to change all that.
A neuron computer, on the other hand, would be as slow as molasses, but it could open up a limitless universe of possibilities (140). Perhaps grown as clusters of human nerve cells and self-assembling according to building plans contained within its genetically modified DNA set, a neuron computer might interface with the outside world by way of a solid state substrate—at the confluence of semiconductor, information, and nano- and biotechnologies. Or perhaps it could talk directly with the neurons of a human brain through electrodes or other means. Conceivably, a neuron computer could be made to mimic the structure or function of some of our own neural complexes. And that might be just the start.
|
CONCLUSION: IMAGING WITH A CRYSTAL BALL |
|---|
|
TOPABSTRACTINTRODUCTIONADVANCES IN CURRENT TECHNOLOGIESDEVELOPING AND EXPERIMENTAL...EVOLVING ROLE OF COMPUTERSCONCLUSION: IMAGING WITH A...ESSENTIALSReferences |
|---|
The long tradition of importing ideas, devices, and software from other disciplines into medical imaging will certainly expand in the future. The National Aeronautics and Space Administration, the Defense Advanced Research Projects Agency, and the Central Intelligence Agency, for example, have developed highly sophisticated imaging technologies to assist astronomers in searching the heavens back to the beginning of time, to sharpen the capabilities of spy satellites, to keep cruise missiles on track, to allow surveillance of individuals suspected of terrorism or criminal behavior, and to practice medicine in space (142). These agencies are constantly looking for civilian applications of their technologies and have already transferred numerous ideas and equipment designs to medicine. Similarly, much of the technology that makes modern clinical imaging so vibrant and flexible was introduced first in the entertainment and communications industries for applications such as video games and special effects for movies and television. These technologies will evolve, and their spin-offs into medicine will continue to improve the quality and efficiency of patient care.
What more is there to be said about computers? First appearing in the clinic in the 1970s, computers have enabled novel computation-based imaging (CT, MR imaging, PET); provided image processing and analysis tools ranging from windowing and noise reduction to computer-aided detection for all modalities; and allowed the instantaneous storage, retrieval, display, and transmission of images and other information with PACS, electronic notepads, et cetera. In so doing, they have altogether changed how medicine is practiced. The deployment of digital imaging systems; high-bandwidth networks to transport images between acquisition, interpretation, consultation, and storage sites; and algorithms for computer recognition of image features for computer-aided detection and diagnosis are ushering in a whole new approach to medical imaging. The only certainty is that what we see today is just the beginning. Tomorrow—perhaps the one-stop digital diagnostician. And after that?
This article closes by posing the ultimate medical-imaging-of-the-near-future question: How likely is it that a totally new modality, one not imagined at present, may burst onto the scene over the next few years? Is another CT or MR imaging revolution in the cards? Some, if not all, of the existing technologies will doubtless evolve and change tremendously, but will we see a completely unanticipated, fundamentally different imaging approach come along?
Medical physicists, biomedical engineers, and research radiologists have long been exploring all known physical processes in search of useful new probes for imaging, and they do find them from time to time. But they have not lately stumbled onto much that looks overwhelmingly dramatic. This leads some in the field to think that most of the technology that we will see in the imaging clinic of 2010, or even of 2025, will be pretty much what we find today, only much, much better. Functional MR imaging, for example, may be fully integrated with PET and magnetoencephalography, yielding splendid high-contrast, high-resolution, and concurrent images of both brain function and neural activity—but these technologies will still be MR, PET, and magnetoencephalography at heart.
Yet several of the new fields opening up today could alter the picture dramatically. Molecular imaging and the use of image-based surrogate markers and biomarkers and of biosensors (143) are in their infancy, as are computer-aided detection and diagnosis, biomedical informatics, imaging nanotechnology, and the convergence of diagnosis and therapy—and the potential of each of these seems boundless (144). Also, once in a while big surprises pop up even in well-plowed fields; it was recently shown, for example, that stimulated emission depletion fluorescence microscopy can provide optical images with resolution down to tens of nanometers—50 times finer than the diffraction barrier that classical optics normally allows (145).
So maybe the clinic of 2025 will be as astonishing to us as MR imaging would be to Roentgen. You just never know!
Everything that can be invented has been invented.Charles H. Duell, Commissioner, U.S. Patent Office, 1899
Note to readers.—The references in this paper comprise a mix of books, recent reviews in journals, and interesting, representative research articles. In-depth discussion of the standard and newer imaging technologies and more extensive bibliographies may be found in references 146 and 147. Nontechnical introductory information on imaging for patients and general education are presented in reference 148 and at http://www.nlm.nih.gov/medlineplus/diagnosticimaging.html.
|
ESSENTIALS |
|---|
|
TOPABSTRACTINTRODUCTIONADVANCES IN CURRENT TECHNOLOGIESDEVELOPING AND EXPERIMENTAL...EVOLVING ROLE OF COMPUTERSCONCLUSION: IMAGING WITH A...ESSENTIALSReferences |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Abbreviations: PACS = picture archiving and communication system
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONADVANCES IN CURRENT TECHNOLOGIESDEVELOPING AND EXPERIMENTAL...EVOLVING ROLE OF COMPUTERSCONCLUSION: IMAGING WITH A...ESSENTIALSReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  T. G. ODLE Breast Ultrasound Radiol. Technol., January 1, 2007; 78(3): 222M - 242M. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. B. Wolbarst and W. R. Hendee The National Institute of Biomedical Imaging and Bioengineering and NIH Grant Process: An Overview Radiology, January 1, 2007; 242(1): 32 - 55. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Martino and T. Odle New Skills Acquisition. Radiol. Technol., November 1, 2006; 78(2): 97 - 100. [Full Text] [PDF]
|
|
|
|
 |
|
 G W L Boland Government reform of the National Health Service: implications for radiologists and diagnostic services Br. J. Radiol., November 1, 2006; 79(947): 861 - 865. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
) in a 10-year-old patient with Landau-Kleffner syndrome. (Image courtesy of M. Funke, MD, PhD, Center for Advanced Medical Technologies, University of Utah, Salt Lake City, Utah.)
