High-frequency sound waves are transmitted through the body by a transducer. The transducer converts electric pulses into sound waves using a piezoelectric material. Transducer frequencies range from 2 to 16 MHz. The time of flight of the sound wave to be received back to the transducer is used to determine the depth of an object. The speed of sound in tissue is more than five times that in air; ultrasound functions only when there is no air between the probe and the target. Sound waves are reflected back in varying degrees, depending on the reflectivity of the tissues at interfaces. The amount reflected is based on both the angle of incidence and the difference between the acoustic impedances of the tissues on either side of the interface. The received echo signals are detected and translated into luminance to form an image with depth perception.

Spatial resolution in ultrasonography has two distinct components: an axial component and a lateral component. The axial resolution is defined as the ability to discern two objects that lie on top of one another. It is dependent on the length of the transmitted ultrasonographic signal. The lateral resolution, the ability to discriminate between two adjacent objects, is determined by the transducer beam width.

Tissue harmonic imaging refers to use of the second and greater fundamental frequencies to generate an image, thus reducing artefact and scatter at least in the near field. This technique has been shown to improve visualization of many structures, especially in the “difficult to scan” or obese patient.

CT uses conventional x-rays to generate two- and three-dimensional images with the help of a computer. The technology has continued to improve as scanners become faster each year. Early CT scanners used an incremental technology in which images were produced at a given table position. Each slice of the patient’s anatomy was obtained by moving the x-ray tube in a circle around the patient, with the table remaining stationary. Image reconstruction started during the scan and was completed shortly after the scan was done. The table was
then moved and the process completed for each slice until the predetermined field of view (FOV) was covered.

All newer generation CT scanners are spiral, in which the tube (x-ray source) continuously rotates around the patient who is continuously moved (translated) at a certain speed defined by the pitch. There is continuous data acquisition, which allows for superior temporal resolution. The computer then generates true axial slice data from the oblique spiral data. Since 2002, CT scanners are now multidetector (4, 8, 16, or 64), meaning that instead of using a single detector, they use a row of detectors called a detector array. The detector array enables simultaneous acquisition of 4 or 16 slices during one gantry rotation. Subsecond gantry rotation speeds have enabled chest abdomen and pelvis to be fully scanned in a single breath-hold.

Because data are acquired as a volume, this volume can be reconstructed even retrospectively to obtain slices of different thickness from the raw data (as long as it has not been deleted). Thin reconstructed image slices of ≤1 mm can then be stacked together to generate a volume for three-dimensional reconstruction for angiography and “virtual” endoscopy. These newer generation multidetector (16- and 64-detector or “slice”) CT scanners can generate isotropic volume data sets that can generate clear images in any plane similar to MRI.

CT images are considered high-resolution images (compared to MRI and ultrasonography) at a matrix of 512 × 512 pixels. The FOV divided by the number of pixels determines the resolution or pixel size (or voxel size when discussing three-dimensional volume). For example, if a 51.2-cm FOV is chosen, then each pixel is 1 mm; if a 25.6-cm FOV is chosen, then each pixel is 0.5 mm. Slice thickness plays a role in the voxel size; for example, if the slice thickness is 5 mm, then the voxel is 1 mm × 1 mm × 5 mm. The thinner the slice, the smaller is the volume of the imaging voxel, and this decreases volume averaging.

MRI is a technique that uses strong magnetic fields and multiple radiofrequency pulses to generate an image with outstanding spatial resolution and tissue contrast. Certain nuclei have a magnetic moment because they are composed of an odd number of protons and neutrons. When placed in a strong magnetic field, these nuclei attempt to align with the field and rotate or spin (precess) about the axis of the magnetic field. The
frequency of precession (the Larmor frequency) depends on the specific nuclei and the field strength of the magnet.

The most commonly imaged nucleus is that of hydrogen 1 because of its abundance in the body and its high gyromagnetic ratio compared with other nuclei, such as those of other commonly found nuclei in the body (e.g., phosphorus 31, sodium 23, carbon 13). When hydrogen nuclei in a strong magnetic field are excited by the addition of a radiofrequency pulse, these nuclei gain energy. When the radiofrequency pulse is turned off, the nuclei return to their resting state and emit the previously absorbed energy at the same frequency. The magnitude of the emitted signal and the time it takes for the nuclei to return to the resting state depend on certain intrinsic properties of the nuclei, which include nuclear spin density (or proton density), longitudinal relaxation, transverse relaxation, and flow.

Spin density (proton density) refers to the density of nuclei potentially available to be imaged. The net signal measured in a given volume is the sum of the hydrogen protons. Therefore, tissues that have few or no hydrogen protons have no MRI signal, such as the air in the lungs or bowel and cortical bone (composed of calcium). The longitudinal (or T1) relaxation time is a measure of the time it takes the precessing nuclei to return to their baseline state (i.e., oriented parallel to the magnetic field) after the radiofrequency pulse has been turned off. The energy released during this process is absorbed by the surrounding molecular environment or lattice; hence, longitudinal or T1 relaxation is also known as spin-lattice relaxation. T1 is the time for the MRI signal to rise to 63% of its maximum equilibrium value. T1 varies between 200 and 800 milliseconds for most tissues. Tissues with a short T1 appear bright on T1-weighted images, and tissues with long T1 (those containing water) are dark.

The transverse (or T2) relaxation time is a measure of the loss of signal in the plane orthogonal to the long axis of the magnetic field due to the loss of phase coherence between the protons. The loss of phase coherence is due to the minute changes in the magnetic field caused by the spinning protons. For this reason, transverse or T2 relaxation is also known as spin-spin relaxation. T2 is the time it takes for the MRI signal to decrease to 63% of its original value. T2 varies between 50 and 200 milliseconds for most tissues. Tissues with a long T2 (those containing water) appear bright on a T2-weighted image.

A number of extrinsic parameters or variables may be manipulated to change the tissue contrast and thus the appearance of an image. The spin echo pulse sequence begins with a 90-degree radiofrequency pulse followed by a 180-degree refocusing radiofrequency pulse. A coherent signal is emitted from the nuclei at the echo time (TE). The complete image data set is produced by repeating this spin echo pulse sequence many times. The time between each successive 90-degree radiofrequency pulse is the repetition time (TR). By varying the TR and TE, one can change the image contrast to produce an image that is either T1 or T2 weighted. Images with a short TE and short TR tend to be T1 weighted, whereas images with a long TE and long TR tend to be T2 weighted. Image matrix size (number of pixels), FOV, and slice thickness may be varied according to the area being imaged, the required detail, and the time available to acquire the image.

PET generates images by detecting energy given off by decaying radioactive isotopes. As they decay, radioactive isotopes emit positrons that collide with electrons and produce gamma rays that shoot off in opposite directions. PET systems use the information from the paths of these two gamma rays to determine the original collision point. The scanners have a circular array of gamma ray detectors similar to the arrangement in a CT scanner. The signals are then processed and converted into two- and three-dimensional images.

PET (superseded now by combination PET-CT) is a relatively new technique for use in the staging of GI malignancies (6). The most commonly used compound in clinical practice is FDG-PET scanning, whereby a glucose derivative is tagged with a radioactive label (fluorine 18). This tagged molecule is transported across the cell membrane, and because it is not suitable for complete glycolysis, it accumulates in the metabolically hyperactive cell. Because accumulation occurs at a greater rate than in most normal tissues, a measure of uptake, called the standard uptake value (SUV), may be calculated. Malignant processes tend to have a higher metabolic rate than benign processes, with the notable exception of inflammation or infection. Combination scanners now fuse the functional (PET) images onto anatomical images (CT), which has increased the accuracy of PET (7). Combination PET-CT is also up to 30% faster due to synergies using CT data for registration. Spatial resolution of PET images is currently limited to 5 mm.

Barium studies remain an excellent way of diagnosing GI malignancies, particularly tumors involving lumens. They are obtained using conventional x-rays with the addition of a fluoroscopy unit. Fluoroscopy allows real time x-ray visualization of the patient with the help of variable amounts of barium and air (hence, “double” contrast). X-rays are continuously transmitted onto a fluorescent screen, and with the help of an image intensifier, are projected onto a television monitor for the radiologist. Other contrast agents such as nonionic intravenous contrast can be given in certain situations. Double-contrast examinations show mucosal detail to a greater extent than single-contrast (barium only) examinations and more readily reveal early lesions (8,9,10). Barium examinations are an efficient way to evaluate the GI tract from the esophagus to the rectum and give useful dynamic information that cannot be achieved with cross-sectional imaging. Discussion of the relative roles of barium examinations and endoscopy is beyond the scope of this chapter. Numerous studies have addressed this issue (11,12,13,14,15,16,17). Barium examination and endoscopy each have advantages and disadvantages. The principal advantage of endoscopy is biopsy capability. In patients who are known to have a malignant polyp in the rectosigmoid region, the remainder of the colon should be examined endoscopically (15).

Primary radiographic evaluation of the esophagus usually entails a barium study (Fig. 9.4), although the esophagogram may be prompted by the appearance of abnormalities on images obtained using other modalities. Double-contrast esophagography has 95% sensitivity for detecting esophageal or gastroesophageal junction carcinoma (21). Several methods are available for staging esophageal carcinoma once diagnosed (22,23,24). However, no single imaging modality is ideal, and imaging assessment of esophageal cancer both preoperatively and post resection remains a difficult challenge. The current evidence shows that PET is more accurate than CT in the detection of occult metastatic disease, prediction of the response to therapy, and detection of recurrent disease (25,26). A combination of CT and EUS has been used until recently for locoregional staging (Fig. 9.5). PET has been shown to be more accurate than CT for detecting metastatic esophageal cancer, and so where available, it should be used to avoid unnecessary major surgery (27). PET-CT is now part of the routine workup of esophageal cancer (Fig. 9.6). Recent studies have shown hybrid PET-CT fused images have incremental value over PET and CT in staging (28). Work has also shown that changes in SUV uptake after treatment with combined chemoradiation correlates with progressionfree survival (29). EUS is the superior modality for
T staging (30). CT has limitations in detecting local invasion of the aorta and other adjacent structures, but virtual CT airway bronchoscopy may obviate fiber-optic bronchoscopy to detect invasion of the bronchial system (31). All modalities may either understage or overstage esophageal carcinoma, and the imaging strategy needs to be tailored for each individual patient.

Gastric carcinoma may present in several ways. Malignancies may appear as ulcers or linitis plastica, or as metastatic disease (32). Linitis plastica has a typical appearance in which the stomach is rigid, narrow, and tubular. Malignant gastric masses may also manifest as ulcers. These ulcerated masses may have characteristic appearances on barium studies (Fig. 9.7). Differentiating between benign and malignant ulcers is not always possible. Certain classic findings, however, are associated with both benign and malignant ulcers and are summarized in Table 9.1. Location in the stomach does not differentiate between benign and malignant ulcers. Although benign ulcers are more common on the lesser curvature and in the gastric antrum, malignant ulcers may appear there as well. The appearance of the gastric rugae near the ulcer crater may aid in differentiation. The folds of a benign ulcer extend to the edge of the ulcer crater, whereas the folds of a malignant ulcer end at a distance from the margin of the ulcer. The latter occurs because a mass is usually associated with a malignant ulceration. On a high-quality double-contrast barium study, classic benign ulcers with the features discussed previously may be followed to resolution. A large percentage of ulcers are indeterminate and require biopsy.

Once the diagnosis of a malignant lesion is made, the patient generally undergoes a CT examination for appropriate staging (33). CT is useful both for assessing local extent of gastric cancer and for identifying metastatic disease (Fig. 9.8). Occasionally, patients with gastric carcinoma present with extragastric symptoms such as jaundice, the cause of which may be seen on CT. The wall thickness of a well-distended stomach is variable but should always be <1 cm; additional enhancement patterns may aid in the diagnosis of gastric cancer (34). Water is used as a low attenuation oral contrast agent for several reasons. First, adequate distension of the stomach is essential
for depiction of lesions, and enhancement of the stomach wall by intravenous contrast would be obscured by high-density oral contrast agents. Second, the use of traditional oral contrast agents would complicate two- and three-dimensional visualization of gastric vessels with CT angiography (35,36). The presence of a focal mass or focal thickening in a well-distended stomach should alert the physician to the possibility of a gastric malignancy. CT is a good modality for the identification of regional and distant spread; however, it is relatively insensitive for determining resectability. However, most prior studies used slice thicknesses of 5 to 10 mm, which have partial volume averaging with adjacent structures. The problem of partial volume averaging is greatly reduced by using section thicknesses of 1.25 mm with superior multiplanar reconstruction images due to the high speed of multidetector scanners, which also overcome breathing artefacts (37). Criteria used to assess resectability, such as irregularity of the tumor or loss of the fat plane between the mass and the adjacent organs, are not sensitive for invasion (38). When these findings are identified, they should alert one to the possibility of unresectability. It has been shown that PET can demonstrate response to treatment of GISTs by a reduction in SUU (39). This is particularly powerful in the setting of little or no change in tumor size.