The use of magnetic resonance imaging (MRI) has been rapidly expanding since it was introduced as a new clinical imaging modality in 1981. MRI relies upon the measurement of magnetization in tissues and it can distinguish tissue structures with far better contrast resolution than other imaging techniques. In contrast to computerized tomography (CT), MRI can provide images in arbitrary planes, including sagittal, coronal, and even oblique directions. CT displays only X-ray beam attenuation coefficients, while MRI uses four biophysiochemical parameters to produce images: resonant frequency, proton density, relaxation times (T&sub1; and T&sub2;), and motion. Scanning methods in MRI are called ‘pulse sequences’. Images are often classified as ‘T&sub1; weighted’, ‘T&sub2; weighted’ or ‘proton density weighted’, terms which refer to the source of the signal difference or relative brightness as displayed on the image.

There are few hazards or contraindications to MRI. The strong magnetic fields used in MRI may cause suppression of cardiac pacemakers due to induced current, or motion of intracranial aneurysmal clips. Clips on coronary arteries and other organs rarely migrate, however, because the torque in the field strength is 1.5 tesla or less. Modern types of heart valves and stainless steel orthopaedic implants which do not exhibit ferromagnetism are also unaffected by MRI.

A proton (nucleus of hydrogen) can be compared with a tiny elementary magnet. In a natural environment, the magnetic moments m of individual protons in the body are pointed in random directions. It follows that their sum, or macroscopic magnetization, M, is zero. If placed in a constant magnetic field B&subo;&sub'; many of the tiny elementary magnets spontaneously undergo a transient orientation in the direction of the field. Therefore, a small magnetization vector M&subz; appears parallel to B&subo;. The nuclear magnetic resonance phenomenon can be observed when protons absorb radiowaves at a specific frequency (the Larmor frequency). On a macroscopic scale, the magnetization vector M will be rotated away from its M&subz; equilibrium position parallel to B&subo;. The angle between M and B&subo; will become greater as the applied radiowave has a longer duration or greater power. In MRI, a transmitted radiowave that shifts vector M perpendicular to B&subo; (transverse or XY plane) is called a 90° pulse. As vector M is then parallel to the XY plane (M+M&subx;&suby;), the value of longitudinal magnetization M&subz; is zero. Resonance occurs when the magnetization rotating in the transverse plane emits radiowaves that can be detected by the imaging system.

When the radiowave excitation pulse ceases, total magnetization M slowly returns to its equilibrium state. This phenomenon is called ‘relaxation’. Once the 90° pulse has been terminated, M&subx;&suby; decays and M&subz; grows. This return of longitudinal magnetization (M&subz;) is characterized by the relaxation time T&sub1;. T&sub1; values vary among tissues according to the mobility of the hydrogen (proton)- containing molecules. Water molecules (such as cerebrospinal fluid) have a very long T&sub1;, while lipid molecules (such as fat tissues) have a very short T&sub1;.

The 90° pulse that shifts equilibrium magnetization from M&subz; to M&subx;&suby; also places the nuclear magnetic moment ‘in phase’, which means that they are moving in exact alignment. However, transverse magnetization is progressively dephased (protons lose alignment) soon after the pulse ceases. This phenomenon, known as transverse relaxation, is characterized by the relaxation time T&sub2; (Fig. 1) 131.

Nuclear magnetic resonance in biological tissues is measured by detecting transverse magnetization M&subx;&suby; before it relaxes (dephases). An electric current is induced in an antenna by the variation of the magnetic field caused by in-phase rotation of protons in the transverse plane (M&subx;&suby;). This electrical signal (the MR signal) represents the magnetic characteristics of each tissue. The amplitude and timing of various signals is mathematically converted to represent the number of protons present in a tissue (proton density) or the relaxation time (T&sub1; and T&sub2;).

Numerous studies in this field have been documented since MRI was first introduced to the clinical environment. MRI promises to be a unique tool, providing high quality images of the brain with few motion artefacts.

Cerebral infarction refers to a sudden and focal neurological deficit with an associated ischaemic abnormality in a localized region of the brain due to thrombosis. The hypoxia that occurs with ischaemia initially leads to an increase in the intracellular water content; tissue necrosis results if the ischaemia is persistent.

Intracranial haemorrhage can appear as various pathophysiological states, including intraparenchymal haemorrhage, intratumoural haemorrhage, haemorrhagic infarction, subarachnoid haemorrhage, epidural or subdural haemorrhage, and vascular malformations.

Precisely defining the extent of the tumour and grading the malignancy are the most important goals of any imaging study because of the implications for patient management and evaluation of treatment. So many classifications of brain tumours have been provided that a knowledge of general neuropathological features is essential for analysing magnetic resonance images.

The clinical course of multiple sclerosis is usually described as a relentless stepwise progression of neurological dysfunction, manifested by exacerbations and remissions. The plaques are pathologically considered to be gliosis, with oedematous change in the acute phase.

Ischaemic infarct and/or oedema is often visualized within 6 to 12 h of onset as an area of hyperintensity (white signal area) on T&sub2; weighted images. A well-circumscribed hyperintensity on T&sub2;-weighted images and hypointensity on T&sub1;-weighted images suggests chronic ischaemic infarction.

Pathophysiology and consecutive changes of erythrocytes and haemoglobin in haematomas make MRI findings complex. Within the first few hours after an intraparenchymal haemorrhage the signal intensity of the haematoma is similar to that of normal tissue. MRI cannot depict acute subarachnoid haemorrhage because the abundant oxygen contained in cerebrospinal fluid prevents the deoxygenation of haemoglobin. In the subacute stage (up to a week), haematomas begin to show central hypointensity due to the presence of deoxyhaemoglobin and hyperintensity of the periphery corresponding to oedema on T&sub2;-weighted images. Over a period of 1 week, haematomas appear as areas of hyperintensity, due to methaemoglobin, ringed with hypointensity due to haemosiderin.

Because the tumour may contain cysts, calcification, haemorrhage, and necrosis, a single signal pattern is not found. Peritumoural oedema increases relaxation times and makes it difficult to detect the tumour margin. Tumours often extend beyond the rims apparent on CT scans, and similar difficulties may reasonably be expected when MRI is employed.

The characteristic MRI findings consist of multiple, usually small, lesions with prolonged relaxation times, most commonly located in the periventricular region. The T&sub2;-weighted spin echo technique is currently the best method for detecting multiple sclerosis: this is positive in between 80 and 100 per cent of patients with definite multiple sclerosis by established clinical criteria.

The advantages of MRI over CT are apparent when delineating small infarcts in the brain-stem and cerebellum and lacunar infarcts in the basal ganglia. MRI can detect hyperacute ischaemic changes in the brain by 6 h after onset, while CT is usually useful in detecting them after 24 h.

MRI is less useful for detecting acute (within 1 day) cerebral haemorrhage. However, it gives a dynamic window through which a variety of haemorrhagic conditions may be observed.

Multiplanar capabilities and a reduction in artefacts from the skull can demonstrate relationships between brain tumours and the skull better than CT, particularly in patients with infratentorial tumours. However, MRI itself occasionally fails to delineate tumour margins due to peritumoural oedema. A solution would be provided by the use of the intravenous contrast agent, gadolinium-DTPA.

The role of traditional imaging studies in multiple sclerosis has been indirect, to rule out a space-occupying lesion. MRI can be used to determine regression or progression of lesions and can reduce the number of paraclinical studies. Additionally, tiny plaques in the brain-stem, cerebellum, and spinal cord, usually missed with CT, are often detected on T&sub2;-weighted images.

Normal parathyroid glands cannot be routinely imaged by CT, sonography, or MRI because of their small size and poor contrast against the thyroid. However, they can be visualized by these imaging modalities once neoplastic or hyperplastic enlargement occurs.

Parathyroid adenomas and hyperplastic glands usually show high signal intensity on T&sub2;-weighted images. Approximately 75 per cent of adenomas can be detected by MRI, but these are often indistinguishable from posterior thyroid adenomas.

MRI may be particularly valuable in patients with recurrent or persistent hyperparathyroidism following surgical exploration performed without preoperative localization of the tumours, since it is free of artefacts due to surgical clips or postoperative fibrous tissue. However, sonography is preferred as the first modality as it is a sensitive and inexpensive method, capable of detecting 80 per cent of parathyroid tumours.

Enlarged mediastinal lymph nodes are the most common mass lesions, accounting for 25 per cent of all mediastinal diseases. Thymomas, teratomas, neurogenic tumours, and bronchogenic and pericardial cysts are also common.

Mediastinal lymph nodes are easily visible on T&sub1;-weighted images, standing out clearly against hyperintense mediastinal fat. Mediastinal tumours such as thymoma, germ cell tumours, and neurogenic tumours usually have similar signal intensities to those of lymph nodes. Cystic lesions containing serous fluid can be distinguished from complicated cysts and solid lesions by their long relaxation times.

MRI may be more efficient than CT for the detection of hilar adenopathy: the condition can be distinguished from pulmonary vessels and bronchi without the need for intravenous contrast because of the superb contrast afforded by the lack of a signal from flowing blood. However, accurate discrimination between cancer and inflammation is still impossible by MRI: size criteria analogous to those used in CT (>1 cm) are used to identify abnormal nodes. MRI cannot provide tissue-specific diagnosis of mediastinal tumours on the basis of signal intensity, but can indicate spatial relationships between tumours and adjacent structures by its multiplanar capabilities.

Among the wide spectrum of hepatic diseases, detection, characterization, and staging (if malignant) of liver tumours currently command the greatest interest in clinical imaging. MRI has been remarkably advanced and employed in the diagnosis of focal hepatic diseases over recent years, and its efficacy has been described by many investigators.

Hepatic metastases often show well-defined margins and homogeneous internal morphology, with hypoechoic T&sub1;-weighted images and slightly hyperechoic T&sub2;-weighted images compared to normal liver parenchyma. Primary and metastatic liver tumours of different histology can be distinguished from each other: MRI commonly demonstrates specific morphological features that suggest the correct tissue diagnosis (Fig. 2) 132.

Hepatomas usually show signal intensities similar to those of metastases. However, about 10 per cent of hepatomas display a bright signal on T&sub1;-weighted intensities because of steatosis; this is very rare in metastases. T&sub2;-weighted images are more reliable for the detection of hepatomas than are T&sub1;-weighted images: using the latter method, 40 to 50 per cent appear as a low intensity capsule, considered to be fibrous tissue.

Cavernous haemangiomas essentially comprise a blood lake. They typically produce well-defined low signal lesions on T&sub1;-weighted images and extremely high intensity signals with T&sub2; weighting (Fig. 3) 133. The specificity of MRI in detecting haemangiomas more than 2 cm in diameter is over 90 per cent.

Focal nodular hypoplasia usually shows similar signal intensity to normal liver parenchyma on T&sub1;-weighted images and slightly higher intensity on T&sub2;-weighted images, with compression of hepatic vascular structures. MRI may demonstrate the characteristic stellate scar of focal nodular hyperplasia, which is not generally encountered in hepatic adenomas, but its differentiation from other hepatic tumours is often difficult.

Cirrhosis usually produces non-specific signal intensities from the liver and can be diagnosed by its morphological features, such as enlargement of the left and caudate lobe, irregularity of liver surface, enlarged portal veins, or prominent collateral veins and splenomegaly, which can also be seen on CT. Regenerating nodules in cirrhotic liver sometimes show high signal intensity in T&sub1;-weighted images due to steatosis, which is indistinguishable from hepatomas with fatty metamorphosis.

Focal fatty liver infiltration often shows regional high intensity on T&sub1;-weighted images, mimicking hepatomas with fatty metamorphosis, but may be clinically diagnosed by its isointensity on T&sub2;-weighted images.

The diagnostic performance of MRI varies substantially, depending on the instruments used, use of pulse sequences and other techniques, lesion size, and tumour histology. In general, T&sub2;-weighted images can show significant differences in the internal structures of hepatic tumours and provide unambiguous tissue contrast betwen lesions and normal hepatic parenchyma, while T&sub1;-weighted images show the anomaly in a similar way to CT. Many CT criteria can be directly applied to T&sub1;-weighted images. Failure to obtain good T&sub2;-weighted images is the most common reason for failure to reach a diagnosis. At present, MRI is not always superior to other modalities, but it is an important adjuvant.

One study showed MRI to have a sensitivity of 64 per cent in the detection of individual metastatic liver tumours, compared with 51 per cent for CT (p<0.0001). The sensitivity for identifying individual patients with liver metastases was 82 per cent, compared with 80 per cent for CT (non-significant). The specificity of MRI in patients without hepatic metastases was 99 per cent, versus 94 per cent for CT (p<0.05). In planning resection of liver metastases, MRI has the advantage over CT because of its superior specificity.

MRI can detect 85 per cent of tumours smaller than 2 cm and 100 per cent of those larger than 2 cm (an average of 97 per cent, compared with 98 per cent for CT). MRI can be almost equivalent to CT in the detection of primary tumours, but often demonstrates secondary nodules missed on CT. MRI is sometimes valuable, therefore, in staging hepatomas when planning surgical approaches for resection, avoiding exploration in patients with unresectable disease. However, total image diagnosis including CT, sonography, and selective angiography is still imperative to assess daughter nodules accurately. Tumour thrombi up to the second branches of the portal system can be usually visually by MRI, but it is difficult to demonstrate portal invasion.

MRI (especially T&sub2;-weighted imaging) is superior to other imaging modalities in diagnosing and delineating cavernous haemangionas from other malignant tumours. Small (< 1 cm) haemangiomas can be demonstrated reliably on T&sub2;-weighted images when sufficient anatomical resolution is achieved. CT usually requires both precontrast and contrast scans to diagnose haemangiomas and sometimes fails to demonstrate small lesions. MRI can prevent patients from requiring a second examination such as sonography or selective angiography after CT.

Cholelithiasis often accompanies inflammatory processes and malignancies in the biliary system. Diagnostic imaging allows the cause of obstructive jaundice to be determined and enables malignancies to be staged.

T&sub2; weighted MRI shows as a signal void (dark signal) surrounded by high intensity bile. Thickening of the gallbladder wall and a dilated biliary system are routinely demonstrated.

MRI does not seem to have any advantages over CT or sonography. Stone disease accounts for a considerable proportion of biliary pathology and represents a physical limitation for MRI as crystalline structures produce no signal. In addition, the poor spatial resolution of MRI is a disadvantage for the study of small biliary structures such as the common bile duct. The applications of MRI are, therefore, limited; they include detection of hepatic metastases from gallbladder cancers and cholangiocarcinomas, but not diagnosis of primary lesions.

The pancreas is an important target for surgeons because of the high mortality rate associated with pancreatic neoplasms. Detecting pancreatic cancer, determining the stage, and differentiating it from pancreatitis are undoubtably the major goals in imaging of the pancreas.

Pancreatic carcinomas larger than 2 cm produce focal mass lesions that are detectable by MRI using morphological criteria analogous to CT. These are associated with the appearance of the tumour and with indirect signs of biliary obstruction and invasion to the neighbouring structures. The contrast between tumour and pancreas derived from longer T&sub1; and T&sub2; of the tumour may clearly delineate the boundary of the tumour, but the invasion to adjacent tissues is often ambiguous because of the relatively poor spatial resolution of MRI.

The characteristics of pancreatitis on MRI are similar to those described with CT. Observed signs include alterations in the pancreatic volume, abnormal colour, dilatation of the pancreatic duct, fluid collections or pseudocysts, and peripancreatic infiltration. However, small pancreatic calcifications are often undetectable on MRI because they do not generate a signal, while the high sensitivity of CT in detecting small calcifications is well known.

CT is apparently superior to MRI for detecting and staging pancreatic tumours. Motion-induced ghost artefacts from surrounding fat and gastrointestinal tract are major problems that produce poor quality magnetic resonance images. MRI does not allow a definitive diagnosis between cancer and pancreatitis. The major advantage of MRI in staging pancreatic carcinomas is its sensitivity for detection of hepatic metastases.

Neoplasms of the gastrointestinal tracts are common, and while advances in barium studies and endoscopic procedures allow accurate diagnosis, assessment of invasion beyond the gut wall or of metastasis to other organs is essential to enable therapy to be planned.

MRI has not been used in the diagnosis of gastrointestinal tract tumours due to its poor spatial resolution, motion artefacts, and poor contrasts between tumours and the bowel wall. However, extension of oesophageal and rectal carcinomas into adjacent structures can be detected with relative success due to fewer motion artefacts and the high contrast of surrounding fat tissue. In other parts of the gastrointestinal tract, MRI usually illustrates only large masses fixed to other abdominal structures.

MRI can be equal or superior to CT in staging of rectal cancers since high quality pelvic images can be obtained. The multiplanar capabilities of MRI allow studies to be performed in the sagittal plane to delineate rectal anatomy. With current resolution, MRI is unlikely to be able to distinguish mucosal tumours from those invading muscle. However, MRI can reliably detect invasive tumours beyond the bowel wall; this is useful in assisting the planning of surgical excision. Sagittal and coronal images can clearly delineate the relationship between the tumour and the rectal sphincter, allowing planning of a sphincter-saving operation. MRI may be equal or superior to CT in detecting pelvic lymphadenopathy and is more specific in identifying the normal patient, but the criteria for lymphadenopathy depends upon the size of nodes: the borderline for enlarged nodes has been reported as ranging from 12 mm to 20 mm.

The detection of renal masses has been dramatically improved by the advance of imaging modalities. At present, the major application of MRI to the kidneys include staging of kidney cell carcinomas and characterization of retroperitoneal masses.

The detection of renal cell carcinoma using MRI is sometimes difficult, particularly when lesions are smaller than 2 cm, because its signal intensities are similar to those of renal parenchyma. There is no morphologically specific finding in this disease (Fig. 4) 134.

The appearance of angiomyolipoma on MRI depends upon the fatty component of the tumour, as with CT. It becomes impossible to differentiate renal angiomyolipoma from renal cell carcinoma if the fat component is poorly contained.

Haemorrhagic cysts occasionally mimic solid tumours on CT. They produce a very high intensity signal on both T&sub1;- and T&sub2;-weighted images and can be delineated from tumours.

Renal stones are not usually detected, except for staghorn calculi, which show a signal void.

MRI has similar results to contrast-enhanced CT in the staging of renal cell carcinoma, with an overall accuracy of 91 per cent. Evaluation of vascular extension and invasion of adjacent organs are two areas in which CT is less reliable than multiplanar resonance. Studies in coronal and sagittal planes allow more accurate differentiation of polar renal masses, adrenal masses, retroperitoneal tumours, or tumours of liver origin, which are notoriously poorly delineated in transverse planes on CT. Multiplanar studies are also useful for evaluating the extent of intravenous tumour thrombus. There are no specific signal characteristics that allow one to differentiate between malignant and inflammatory retroperitoneal lymph nodes using MRI.

MRI has not been extensively applied to the evaluation of renal pelvic tumours (transitional cell carcinomas) since intravenous urography and antegrade/retrograde pyelography have high diagnostic accuracy.

Intravenous contrast agents such as gadolinium-DTPA can be used in MRI, as with CT. MRI has no advantages in this field because CT and sonography are apparently better able to detect renal mass lesions than MRI.

Non-invasive evaluation of adrenal gland morphology has been facilitated by the advance of CT, which has good spatial resolution. Adrenal tumours are divided into functioning and non-functioning; the former can be diagnosed by hormonal assays, but the latter have to be differentiated from metastases.

There are no tissue- or hormone-specific findings which allow differentiation between adenomas, carcinomas, and metastases. The diagnosis is usually established on the basis of morphological appearance, as with CT, rather than signal intensities.

The advantages of MRI in this field lie in its multiplanar capabilities that allow adrenal tumours to be differentiated from tumours of adjacent structures. There appears to be an overlap of approximately 25 per cent in the signal characteristics of malignant (metastatic) and benign adrenal tumours. Needle aspiration biopsy is still imperative to characterize tissue reliably.

Cervical carcinoma, endometrial carcinoma, and uterine myoma are common disorders requiring surgical treatment. While gynaecological examinations can provide some information about tumour staging, imaging is essential for the accurate diagnosis of uterine cancers.

T&sub2;-weighted images provide superior contrast resolution of the zonal anatomy of the uterine body and the boundary between the uterus and cervix, the isthmus and the vagina. Delineation of these structures allows the depth of myometrial invasion and vaginal invasion of uterine neoplasms to be determined: tumours appear as high intensity lesions in low intensity muscular structures. T&sub2;-weighting is also sensitive in demonstrating as low intensity masses uterine leiomyomas as small as 0.5 cm, and can determine the site of origin as submucosal, intramural, or subserosal.

The advantages of MRI over CT in this field are multiplanar capability and the superior contrast resolution obtained with T&sub2;-weighted images. MRI has been reported to be able to demonstrate Stage II lesions in cervical cancers and Stage I, II, and III lesions in endometrial cancers, but it cannot detect Stage 0 lesions of either. T&sub2;-weighted imaging in the sagittal plane may be useful in detecting rectal or vesical invasion; however CT is the most accurate method for staging advanced cervical cancers. MRI is as accurate as CT in demonstrating lymph node involvement but benign and malignant adenopathy cannot be separated.

Another potential of MRI is the capability of imaging obstetric patients, since it avoids any prenatal exposure to radiation.

Ovarian neoplasms are currently classified according to a WHO scheme based on histogenesis: epithelial origin, stromal origin, and germ cell origin. Up to 10 per cent of neoplasms are metastases from other sites. Preoperative imaging affects only the type of operative approach, since all patients with malignant ovarian tumours should undergo laparotomy. In patients with benign tumours, imaging may be used for follow-up, but laparotomy is eventually needed to rule out malignancy.

Benign cystic ovarian tumours typically appear as low intensity on T&sub1;-weighted images and high intensity on T&sub2; weighting, but often contain haemorrhage and protein-rich fluid, causing a high intensity T&sub1;-weighted image. Some kinds of ovarian tumours, for example, ovarian fibroma, show characteristic low intensity on T&sub2;-weighted images but most ovarian tumours show non-specific findings that do not allow differentiation of benign from malignant.

MRI has not been sufficiently evaluated in the diagnosis and staging of ovarian tumours. Its multiplanar capabilities may be of value in detecting the spread of malignant tumours to adjacent organs, but MRI has not been proved to be more useful than CT or sonography in differentiating between benign and malignant tumours. MRI and CT are equally effective in the detection of pelvic and retroperitoneal lymphadenopathy, and MRI can also be used to characterize some specific ovarian tumours, such as fibroma, polycystic ovary, and endometrial cyst. Dermoid cysts, the most common benign ovarian tumours, are better characterized on CT than by MRI.

Bladder tumours are classified histologically as transitional cell carcinoma (most common), adenocarcinoma (from urachal remnants), squamous cell carcinoma, sarcoma, and others. Since most superficial tumours are treated by endoscopic resection, imaging should be performed, particularly to stage bladder tumours which have extended beyond the muscular layer of the wall and to assess metastases.

T&sub1;-weighted images and PDWI can usually delineate tumours from urine and perivesical fat, while T&sub2; weighting allows differentiation of tumour from adjacent normal bladder wall. Disruption of the low intensity line of the bladder wall on T&sub2;-weighted images correlates with invasion beyond the deep muscular layer.

Reported accuracies of MRI in staging bladder tumours range from 73 to 85 per cent, which is similar to or slightly better than CT (59–88 per cent). MRI is unreliable in predicting the depth of bladder wall invasion, especially in delineating invasiveness of superficial tumours (less than Stage B2) from invasive tumours (more than Stage B2).

Prostatic cancer is the second most common cancer in men. Although digital rectal examination is still recommended as the best screening technique for the detection of potentially curable tumours, transrectal sonography and MRI have been recently introduced as complementary techniques.

MRI demonstrates the zonal anatomy of the prostate, allowing characterization of prostatic nodules and of prostatic cancers. T&sub2;-weighted imaging allows the prostate to be divided into a hyperintense peripheral zone and a hypointense central and transitional zone. Prostatic carcinoma develops predominantly (about 70 per cent of cases) in the peripheral zone, although benign prostatic hyperplasia predominantly (95 per cent of cases) affects the transitional zone. Prostatic nodules arising from the peripheral zone are, therefore, likely to be malignant.

MRI can provide excellent tissue contrast in the prostate, but the shape and signal intensity of prostatic carcinoma are non-specific and mimic benign prostatic hyperplasia unless extracapsular extension or metastases are obvious. The accuracy of MRI in differentiating stages A and B (confined to the prostate) from Stage C and D (extracapsular extension or metastases) has been reported to range from 52 to 89 per cent.

Disease of the spine can be divided into three categories: intramedullary (intraspinal cord) lesions, intradural extramedullary lesions, and extradural lesions, including intervertebral disc diseases.

Syringomyelia and hydromyelia appear as low intensity images with T&sub1; weighting and as high intensity lesions spreading longitudinally within the spinal cord with T&sub2; weighting (Fig. 5) 135. In patients with Arnold–Chiari malformation, MRI which includes the posterior fossa of the skull can show the cerebellar tonsils and vermis projecting into the upper cervical spinal canal. Intramedullary plaques of multiple sclerosis may appear as high intensity on T&sub2;-weighted images, but are not visible on T&sub1;-weighted images.

Intramedullary spinal tumours (ependynomas and astrocytomas) have no reliable characteristic magnetic resonance feature, but appear as lesions with prolonged T&sub1; and T&sub2;. Spinal arteriovenous malformation may be sometimes demonstrated as a high intensity lesion with a spotty signal void, consistent with abnormal large vessels detectable on T&sub2;-weighted images.

Paravertebral extension of the dumb-bell shape intradural extramedullary tumours (neurofibroma, neurinoma, etc.) are usually well demonstrated on T&sub2;-weighted images but poorly differentiated with T&sub1; weighting from contiguous muscle (Fig. 6) 136. Spinal meningiomas show similar signal intensities to neural tissues on both T&sub1;- and T&sub2;-weighted images.

Degenerative discs usually produce lower signal intensities than normal on T&sub2;-weighted images. Protruded or extruded fragmented discs are easily visible on T&sub1;-weighted and PD weighted images. Involvement of the spine by metastatic tumours is demonstrated as low intensity vertebrae on T&sub1;-weighted imaging.

For the investigation of almost all conditions of the spine, MRI is an alternative to contrast myelography and is complementary to radiographs and CT scans. Sagittal scans of the spine and oblique scans along disc spaces are particularly helpful.

Syringomyelia and hydromyelia can be demonstrated in their superior and inferior extents in most cases (about 90 per cent). CT-myelography is the most effective way in confirming syringomyelia when results of MRI are equivocal. MRI is the only method capable of demonstrating intramedullary plaques of mutiple sclerosis. A survey examination of the brain in patients with suspected spinal lesions of multiple sclerosis may demonstrate lesions compatible with this diagnosis.

Intradural extramedullary tumours can be demonstrated on MRI as well as on CT, but secondary erosive changes of the subarticular canal are less precisely visualized than with CT. Calcification is frequently difficult or impossible to detect on MRI. The intravenous contrast agent gadolinium-DTPA may be valuable, particularly in detection and estimation of extent of tumours.

Degenerative disc diseases (spondylosis, disc herniation, and canal stenosis) are widely applicable to investigation by MRI. There is 83 per cent agreement between the results of preoperative MRI and surgical findings in patients with degenerative disc diseases, a rate equivalent to that of CT–myelography. T&sub2;-weighted imaging in the sagittal plane is useful in the survey of degenerative discs, which show lower signal intensities than normal (Fig. 7) 137. Sagittal T&sub1;-weighted imaging allows simultaneous visualization of multiple metastatic tumours affecting the vertebrae; this is impossible with CT.

MRI findings complementary to plain radiographs in the assessment of vertebral disorders, such as atlantoaxial subluxation, spondylosis, spondylolisthesis, and tethered cord syndrome (myelomeningocele with dysgenesis of the spina and lipoma).

Injuries of the knee joint, such as meniscal tears, ligament injuries and fractures, arthritis, and neoplasms are routinely encountered. Plain films, CT, and arthrography are used for diagnosis. The use of MRI in the assessment of knee joint diseases is increasing because of its capability to be used in multiple image planes and fewer problems with motion artefacts.

Excellent visualization of menisci and ligaments can be achieved by coronal and sagittal MRI. For example, meniscal tears show disruption of low intensity meniscus in high intensity intra-articular fat or fluid on T&sub2;-weighted images.

MRI may be helpful in evaluating patients with minor knee injuries and may reduce the need for arthrography and arthroscopy. Meniscal tears that are difficult to detect on arthroscopy can be demonstrated, along with surrounding fluid or blood. Images along the longitudinal direction of the femur, tibia, and fibula are useful to show the extent of marrow involvement in patients with bone tumours.

The most common clinical indications for MRI are pain or restricted range of motion in the shoulder, patients referred with suspected rotator cuff tears or impingement, defects in the glenolabrum, infections, and neoplasms.

Sagittal T&sub2;-weighted imaging is well suited for evaluating rotator cuff disease. Tendinitis and fluid collections show as high intensity lesions, often associated with the cuff tear. Disruption of the labrum, the main cause of recurrent anterior dislocation, are seen as linear high intensity lesions on T&sub2;-weighted imaging.

Routine radiography, radionuclide studies, CT, and sonography have been important techniques for evaluating the glenohumeral joint, but arthrography has been essential when surgical treatment is considered. MRI may reduce the indications for arthrography by its excellent soft tissue contrast and multiplanar capability.

Although there are numerous causes of hip pain, MRI has progressed most rapidly in evaluating patients with suspected avascular necrosis of the femoral head because of its high sensitivity and specificity.

Early in the course of avascular necrosis, an inhomogeneous loss of signal intensity can be seen on T&sub1;-weighted images. A low intensity line of demarcation may also be evident at the margin of the necrotic zone during the early phase, when radiographs are typically normal.

Bernardino ME, Steiberg HV, Pearson TC. Comparison of MRI and angiography in the determination of shunt patency for portal hypertension. Radiology 1986; 158: 57–61.

Berquist TH, et al. Magnetic resonance imaging of the chest: a diagnostic comparison with computed tomography and hilar tomography. Magnetic Resonance Imaging 1984; 2: 315–27.

Berquist TH. Shoulder and arm. In: Berquist TH, ed. MRI of the Musculoskeletal System, 2nd edn. New York: Raven Press, 1990: 313–56.

Biondetti PR, Lee JKT, Ling D, Catalona WJ. Clinical stage B prostate carcinoma: staging with MR imaging. Radiology 1987; 162: 325–9.

Bradley WG Jr, Crooks LE, Newton TH. Physical principles of NMR. In: Newton TH, Potts DG, eds. Modern Neuroradiology, Vol. 2. San Francisco: Clavadel, 1983: 15–61.

Brant-Zawadzki M, et al. Primary intracranial tumour imaging: a comparison of magnetic resonance and CT. Radiology 1984; 150: 435–40.

Brant-Zawadzki M, et al. Basic principles of magnetic resonance imaging in cerebral ischaemia and initial clinical experience. Neuroradiology 1985; 27: 517–20.

Butch RJ, et al. Staging rectal carcinoma by MR and CT. Am J Roentgenol 1986; 146: 1155–60.

Butch RJ, Stark DD, Malt RA. Magnetic resonance imaging of focal hepatic nodular hyperplasia. J Comput Assist Tomogr 1986; 10: 874–7.

Buy J-N, et al. MR imaging of bladder carcinoma: correlation with pathologic findings. Radiology 1988; 169: 695–700.

Bydder GM, et al. Enhancement of cervical intraspinal tumors in MR imaging with Gd-DTPA. J Comput Assist Tomogr 1985; 9: 847–51.

Choyke PL, Kressel HY, Pollack HM, Arger PM, Axel L, Mamourian AC. Focal renal masses: magnetic resonance imaging. Radiology 1984; 152: 471–7.

Dooms GC, Hricak, H, Crooks LE, Higgins CB. Magnetic resonance imaging of the lymph nodes: comparison with CT. Radiology 1984; 153: 719–28.

Dooms GC, et al. Cholangiocarcinoma imaging by MR. Radiology 1986; 159: 89–94.

Ebara M, et al. Diagnosis of small hepatocellular carcinoma: correlation of MR imaging and tumor histologic studies. Radiology 1986; 159: 371–7.

Ferruci JT Jr. MR imaging of the liver. Am J Roentgenol 1986; 147: 1103–16.

Gallimore G, Harms SE. Knee injuries: high-resolution MR imaging. Radiology 1986; 160: 457–61.

Gamsu G, et al. Magnetic resonance imaging of benign mediastinal masses. Radiology 1984; 151: 709–13.

Glazer GM, et al. Evaluation of focal hepatic masses: a comparative study of MRI and CT. Gastrointest Radiol 1986; 11: 263–8.

Gomori JM, et al. Intracranial haematomas: imaging by high field MR. Radiology 1985: 87–92.

Grossman RI. Haemorrhage. In: Stark DD, Bradley WG Jr, eds. Syllabus: a Categorical Course in Diagnostic Imaging—MR Imaging. Radiological Society of North America 1988: 189–95.

Hricak H, et al. Uterine leiomyoma: correlation of MR, histopathologic findings and symptoms. Radiology 1986; 158: 385–91.

Hricak H, Thoeni RF, Carrol PR, Demas BE, Marotti M, Tanagho EA. Detection and staging of renal neoplasms: a reassessment of MR imaging. Radiology 1988; 166: 643–9.

Idy-Perenti I, Bittoun J. Physical basis. In: Vanel D, McNamara MT, eds. MRI of the Body. Paris: Springer-Verlag, 1989: 1–29.

Kerlin P, et al. Hepatic adenoma and focal nodular hyperplasia: clinical, pathologic and radiologic features. Gastroentrology 1983; 84: 994– 1002.

Lee BCP, Deck MDF. Sellar and juxtasellar lesion detection with MR. Radiology 1985; 151: 143–7.

Lee JKT, et al. Magnetic resonance imaging of abdominal and pelvic lymphadenopathy. Radiology 1984; 153: 181–8.

Lee BCP, et al. MR imaging of syringomyelia and hydromyelia. Am J Neuroradiol 1985; 6: 221–8.

Li KC, et al. Distinction of cavernous hemangioma from hepatic metastases with MRI. Radiology 1988; 169: 409–15.

Mitchell DG, Rao VM, Dalinka MK. Femoral head avascular necrosis: correlation of MR imaging, radiographic staging, radionuclide imaging and clinical findings. Radiology 1987; 162: 715.

Modic MT, et al. Imaging of degenerative disc disease. Radiology 1988; 168: 177–86.

Mueller NL, Gamsu G, Webb WR. Pulmonary nodules: detection using magnetic resonance and computed tomography. Radiology 1985; 155: 687–90.

Nelson RC, Chezmar JL, Sugarbaker PH, Bernardino ME. Hepatic tumors: comparison of CT during arterial portography, delayed CT, and MRI imaging for preoperative evaluation. Radiology 1989; 172: 27–34.

Peck WW, Dillon WP, Norman D, Newton TH, Wilson CB. High-resolution MR imaging of pituitary adenomas at 1.5 T: experience with Cushing disease. Am J Roentgenol 1989; 152: 145–51.

Renig JW, et al. Adrenal masses differentiated by MR. Radiology 1986; 158: 81–4.

Rummeny E, et al. MR imaging of liver neoplasms. Am J Roentgenol 1989; 152: 493–9.

Schiebler ML, et al. Prostatic carcinoma and benign prostatic hyperplasia: correlation of high-resolution MR and histopathologic findings. Radiology 1989; 172: 131–7.

Sipponen JT, Sepponen RE, Sivula A. Nuclear magnetic resonance (NMR) imaging of intracerebral haemorrhage in the acute and resolving phases. J Comput Assist Tomogr 1983; 7: 959.

Stark DD, Moss AA, Goldberg HI, Davis PL, Federle MP. Magnetic resonance and CT of the normal and diseased pancreas: a comparative study. Radiology 1984; 150: 153–62.

Stark DD, et al. Detection of hepatic metastases by magnetic resonance: analysis of pulse sequence performance. Radiology 1986; 159: 365–70.

Stark DD, Wittenberg J, Butch RJ, Ferrucci JT Jr. MR detection of hepatic metastases: a randomized controlled comparison with CT. Radiology 1987; 165: 339–406.

Stark DD, et al. Magnetic resonance imaging and spectroscopy of hepatic iron overload. Radiology 1985; 154: 137–42.

Stark DD, et al. Magnetic resonance imaging of the neck. II. Pathologic findings. Radiology 1983; 150: 455–61.

Stark DD, Clark OH, Moss AA. Magnetic resonance imaging of the thyroid, thymus and parathyroid glands. Surgery 1984; 96: 1083–90.

Stark DD, Moss AA, Goldberg HI. Nuclear magnetic resonance of the liver, spleen and pancreas. Cardiovasc Interventional Radiol 1986; 8: 329–341.

Stark DD, et al. Magnetic resonance imaging of the neck. I. Normal anatomy. Radiology 1983; 150: 447–54.

Stark DD, Goldberg HI, Moss AA, Bass NM. Chronic liver disease: evaluation by magnetic resonance. Radiology 1984; 150: 149–51.

Steiner E, et al. Imaging of pancreatic neoplasms: comparison of MR and CT. Am J Roentgenol 1989; 152: 487–91.

Sze G, et al. Gadolinium-DTPA in the evaluation of intradural extramedullary spinal disease. Am J Neuroradiol 1988; 9: 153–63.

Webb WR, et al. Evaluation of magnetic resonance imaging of the normal and abnormal pulmonary hila. Radiology 1984; 152: 89–94.

Wehrli FW. Principles of magnetic resonance. In: Stark DD, Bradley WG Jr, eds. Magnetic resonance Imaging. St Louis: Mosby, 1988: 3–23.

Winkler ML, Hricak H, Higgins CB. MR imaging of diffusely infiltrating gastric carcinoma. J Comput Assist Tomogr 1987; 11: 337–9.