Radiographically negative avascular necrosis: detection with MR imaging

To correlate the morphologic appearance on magnetic resonance (MR) images of radiographically negative avascular necrosis (AVN) of the femoral head with that on computed tomographic (CT) and radionuclide scans, the radiographic and clinical records of 24 patients were reviewed retrospectively. In 18 patients the MR signal intensity features were monitored by means of serial imaging. All MR studies included T1-weighted (short repetition time [TR], short echo delay time [TE] ) imaging and T2-weighted imaging (long TR, long TE). Thirty-one hips were determined with MR to be involved by AVN; 27 were staged on the basis of signal intensity characteristics within the low-intensity rim. Core decompression was performed on 18 hips. Afterward, progression of disease occurred in only one hip. Fourteen of the 16 asymptomatic patients (88%) had early-stage focal lesions. CT scans were obtained in 15 patients and radionuclide scans in 21. Ten hips at radionuclide imaging and five at CT appeared normal when MR results were distinctly abnormal. MR can depict early radiographically negative AVN in asymptomatic individuals. At this early stage, the lesions in this series appear to be nonprogressive after treatment.     

MR imaging of the female pelvis at 3 Tesla: evaluation of image homogeneity using different dielectric pads

PURPOSE: To evaluate improvements in image homogeneity in pelvic MR imaging at 3 Tesla (T) using two different dielectric pads. MATERIALS AND METHODS: A total of eight healthy females were scanned using a 3T MR scanner equipped with a body-array coil. Axial and sagittal fast spin-echo T2-weighted images (T2WI) (TR/TE = 3200 msec/94 msec), axial fast spin-echo T1-weighted images (T1WI) (TR/TE = 700 msec/11 msec), and sagittal half-Fourier acquisition single-shot turbo spin-echo (HASTE) images (TR/TE = 3000 msec/100 msec) were performed for pelvic imaging. Sequences were repeated with dielectric pads (consisting of either ultrasound [US] gel or water), and without pads. Three or four regions of interest (ROIs) were placed on fatty tissues and the ratio of minimum to maximum signal intensity (RSI) was calculated as a marker of image homogeneity. RESULTS: RSI was significantly higher on T2WI and T1WI when using dielectric pads than when no pad was used. A similar tendency was observed in RSI on HASTE. No significant difference was found between images with US gel pads and those with water pads. CONCLUSION: Dielectric pads consisting of either US gel or water are effective in improving image homogeneity of the pelvis on 3T MRI.     

Femoral head avascular necrosis: correlation of MR imaging, radiographic staging, radionuclide imaging, and clinical findings

To better correlate the appearance of avascular necrosis (AVN) of the femoral head on magnetic resonance (MR) images with the stage of disease, MR images of 56 proved AVN lesions were compared with staging from corresponding radiographs (n = 56), Tc-99m scans (n = 41), and grade of symptoms (n = 28). Fractures complicating AVN were seen in 28 (50%) of 56 radiographs (radiographic stages III-V). With long repetition (TR) and echo delay (TE) times, a characteristic "double line sign" consisting of high signal intensity inside a low-intensity peripheral rim was seen in 45 lesions (80%). The central region within the rim was isointense with marrow fat on both short and long TR and TE images in 20 (71%) of 28 lesions uncomplicated by fracture (stages I-II) but in only four (14%) of 28 stage III-V lesions (P less than .001). Symptoms were least severe in lesions isointense with fat and most severe in lesions with low-signal central regions at short and long TRs and TEs. The peripheral double line sign on long TR/TE images may add specificity to the diagnosis of AVN by MR imaging. A chronologic pattern of central MR signal features is presented which may allow staging of AVN by MR imaging.     

Imaging of pancreatic neoplasms: comparison of MR and CT

Thirty-two patients with pathologically proved pancreatic carcinomas or cystadenomas were evaluated with MR images obtained with T1-weighted spin echo (short TR/short TE), inversion recovery, and T2-weighted spin-echo (long TR/long TE) pulse sequences. CT was used as the reference standard to determine the ability of MR to delineate normal and abnormal pancreatic anatomy and thereby to exclude or detect pancreatic malignancy. Short TR/short TE spin-echo sequences were significantly better (p less than .05) than inversion recovery or T2-weighted spin-echo sequences in resolution of both normal and abnormal anatomy. Resolution of pancreatic anatomy correlated (r = .9) with the image signal-to-noise ratio. In seven (22%) of 32 cases, MR visualized pancreatic tumors better than CT did because it showed a signal intensity difference between the tumor and normal pancreatic tissue. Overall, the slight superiority of MR over CT for tumor visualization tended to occur in larger tumors and was not statistically significant. On T1-weighted images, 63% (20 of 32) of pancreatic tumors studied had lower signal intensities than normal pancreatic tissue, whereas on T2-weighted sequences (TE = 60, 120, and 180 msec) only 41% (13 of 32) of tumors had increased signal intensities. Currently available MR imaging techniques offer no significant advantages over CT for evaluating the pancreas for neoplasia.     

Radiation fibrosis: differentiation from recurrent tumor by MR imaging

Magnetic resonance (MR) images of 21 patients who had undergone radiation therapy were analyzed and compared with those of 15 patients who had untreated tumors. T2-weighted images (TR = 1,500 msec, TE = 90 msec) were most helpful in distinguishing recurrent tumor from radiation fibrosis. Radiation fibrosis, like muscle, usually remained low in signal intensity on T2-weighted images, while tumor demonstrated higher signal intensity. In no patient was the signal intensity of tumor the same or less than muscle on the T2-weighted images. However, relatively high signal intensity on T2-weighted images is not specific for tumor recurrence and may be seen in acute radiation pneumonitis, infection, hemorrhage, and even pulmonary radiation fibrosis.     

MR of Leigh's disease (subacute necrotizing encephalomyelopathy)

MR images of three patients with Leigh's disease (subacute necrotizing encephalomyelopathy) were compared with CT findings. In all patients typical lesions in the basal ganglia were identified with both MR and CT. In two patients MR permitted identification of additional lesions not detected with CT. In one patient progression of MR abnormalities over a 4-month period correlated well with clinical deterioration in neurologic status. T2-weighted images with a repetition time (TR) greater than 1950 msec and an echo time (TE) greater than or equal to 60 msec or inversion-recovery images with a 50-msec TE, 1213-msec inversion time, and 3000-msec TR were advantageous in identifying multiple necrotic lesions in the brainstem, deep gray matter, periventricular white matter, and cerebral cortex. In this series MR was more sensitive in detecting and localizing multifocal necrotic lesions of Leigh's disease than CT was, and thus may be a useful diagnostic tool for patients with the appropriate clinical and laboratory abnormalities.     

MR imaging of the lungs: value of short TE spin-echo pulse sequences.

OBJECTIVE. An experimental short echo delay (TE = 7 msec) T1-weighted spin-echo sequence was compared with a conventional (TE = 20 msec) T1-weighted spin-echo sequence in the assessment of normal and abnormal lung parenchyma. Comparison was also made with high-resolution CT of abnormal lung parenchyma. SUBJECTS AND METHOD. At 1.5 T, an experimental short echo delay T1-weighted multislice spin-echo sequence (TR = RR interval, TE = 7 msec) was compared with an optimal conventional T1-weighted spin-echo sequence (TR = RR interval, TE = 20 msec, spatial presaturation). Ten healthy volunteers were examined with both sequences. The mean signal intensity and signal-to-noise ratios were calculated in lung parenchyma for both sequences. Two radiologists compared the visualization of normal lung parenchymal structures with the two techniques. In 24 patients with diffuse lung disease, results with both MR sequences and with high-resolution CT were compared. RESULTS. The signal intensity was significantly greater (p < .001) with the TE of 7 msec than with the TE of 20 msec, resulting in a 3.5-fold improvement in the signal-to-noise ratio. The 7-msec TE improved visualization of lung parenchymal structures, including peripheral vessels, interlobular septa or veins, and centrilobular arteries. In the patients with diffuse lung disease, pulmonary parenchymal abnormalities were better visualized on the images with TEs of 7 msec than on images with TEs of 20 msec. When compared with high-resolution CT, the sequence with a TE of 7 msec provided comparable assessment of air-space opacification and dense consolidation, but it was inferior to high-resolution CT in the anatomic assessment of lung parenchyma. CONCLUSION. This experimental spin-echo sequence with a TE of 7 msec significantly improves the signal-to-noise ratio, allowing improved visualization of normal and abnormal pulmonary parenchyma when compared with conventional spin-echo images with a TE of 20 msec. Although anatomic detail remains inferior to that seen with high-resolution CT, the improved image quality with a TE of 7 msec suggests that assessment and follow-up of parenchymal lung disease might be possible with MR, thereby avoiding ionizing radiation.     

Acute hypertensive intracranial hemorrhage: MR imaging at 1.5T

Twelve patients with acute hypertensive intracranial hemorrhage underwent magnetic resonance (MR) imaging within 7 days after the ictus. T1-weighted (TR = 400 msec; TE = 20 msec) and T2-weighted (TR = 2000 msec; TE = 80 msec) images were obtained on a 1.5 Tesla MR system. Signal intensities of hematomas were carefully evaluated and were compared with white matter intensity. A 9-hour-old hematoma was mildly hypointense on T1-weighted images, and was mildly hyperintense on T2-weighted images, suggesting a reflection of the high water content. On T2-weighted images, thin peripheral hypointense rim, probably due to deoxyhemoglobin, was also observed. Both of 15-hour-old hematoma and 21-hour-old hematoma had peripheral hypointensity on T2-weighted images. Both of 39-hour-old hematoma and 43-hour-old hematoma had central hyper-intensity on T1-weighted images and iso-to-mild central hypointensity on T2-weighted images, suggesting a reflection of decreased water content. A 3-day-old hematoma had thin peripheral iso-to-mild hyperintense rim on T1-weighted images, presumably due to intracellular methomoglobin. A 5-day-old hematoma had thin peripheral hyperintense rim on T2-weighted images, probably due to free methemoglobin. A 7-day-old hematoma was hyperintense on T1-weighted images and was mildly hypointense to hyperintense on T2-weighted images, presumably due to mixed intracellular methemoglobin and free methemoglobin.     

Fast MR imaging of liver metastasis using FLASH and FISP--optimal sequences for T1- and T2*-weighted images

This paper deals with a study to obtain the optimal sequence of gradient echo (GE) for T1- and T2*-weighted images similar to T1- and T2-weighted images of spin echo (SE). Two GE sequences, fast low angle shot (FLASH) and fast imaging with steady-state precession (FISP), were performed in 15 cases of liver metastasis in various combination of flip angle (FA), repetition time (TR), and echo time (TE). The optimal combinations were summarized as follows: 1) T1-weighted FLASH image with FA of 40 degrees, TR of 22 msec and TE of 10 msec, 2) T1-weighted FISP image with FA of 70 degrees, TR of 100 msec, TE of 10 msec, 3) both T2*-weighted FLASH and FISP images with FA of 10 degrees, TR of 100 msec and TE of 30 msec. Not only to provide the adequate T1- and T2*-weighted images but also to enable breath-holding MR imaging, GE sequences can optionally take place SE in cases of deteriorated images caused by moving artifacts. Other applications support the re-examination and further detailing when required, conveniently rather in short time.     

Double arterial phase dynamic MRI with sensitivity encoding (SENSE) for hypervascular hepatocellular carcinomas

PURPOSE: To investigate the efficacy of SENSE MRI, including the double arterial phase dynamic study, to detect hypervascular hepatocellular carcinomas (HCCs). MATERIALS AND METHODS: MRI of the liver was performed in 40 consecutive patients (20 by conventional MRI and 20 by SENSE MRI). The SENSE technique was used to obtain breath-hold T1-weighted FSE images (TR/TE = 556/12 msec), respiratory-triggered T2-weighted FSE images (TR/TE = 1800/90 msec) with and without fat suppression, and dynamic MR images (TR/TE/FA = 160-168/4.6 msec/70 degrees ). In each arterial dominant phase and portal dominant phase, two scans were consecutively performed with one breath-hold, leading to the double arterial phase and double portal phase images with SENSE. RESULTS: The sensitivity of SENSE MRI for HCCs diagnosed from all MR images, including dynamic study, T1-weighted images, and T2-weighted images, was 91.7%, while that of conventional MRI was 76.3%. The positive predictive value of SENSE MRI for HCCs was 91.7%, while that of conventional MRI was 87.9%. In terms of HCCs < or = 10 mm, the sensitivity and positive predictive values of SENSE MRI were 78.6% and 78.6%, respectively, while those of conventional MRI were 27.3% and 60.0%, respectively. The number of detected HCCs < or = 10 mm was significantly larger in SENSE MRI than in conventional MRI (P < 0.05). The cause of false-positive lesions on SENSE MR images was an arterioportal shunt. CONCLUSION: SENSE MRI with double arterial phase dynamic study showed higher sensitivity compared to the conventional technique. Therefore, SENSE MRI is a promising method for the detection of HCC.     

Giant cavernous hemangioma of the liver: CT and MR imaging in 10 cases

Ten giant cavernous hemangiomas of the liver in eight patients were examined with both MR imaging and dynamic bolus CT. The maximal diameters of the tumors were 6.5-19 cm (mean, 10.8 cm). MR imaging was done with a 2.0-T superconducting magnet and spin-echo imaging. CT was done with single-bolus dynamic scans. On MR images, all 10 hemangiomas had a heterogeneous appearance. The main part of the tumor comprised uniform, well-defined, high-intensity areas on T2-weighted images, with increasing intensity ratios with prolongation of TR and TE. Other parts of the tumor were cleftlike and were of lower intensity than the remainder of the tumor on T1-weighted images and of higher intensity on heavily T2-weighted images. These parts corresponded to the areas of the tumor that were of lower density on dynamic bolus CT scans. Internal septa in the tumor of low intensity were also noted on all MR pulse sequences. These parts corresponded to low-density areas on delayed contrast-enhanced CT. Familiarity with the characteristics of the internal architecture of giant cavernous hemangiomas on MR imaging or dynamic bolus CT might be useful in making the correct diagnosis of this tumor.     

Neck neoplasms: MR imaging. Part I. Initial evaluation

Untreated neoplasms of the neck (tumors of the oropharynx, supraglottic area, carotid body, and thyroid, in addition to malignant lymphadenopathy) were evaluated in 23 patients with magnetic resonance (MR) imaging. The results were compared with computed tomographic (CT) scans in 20 patients. Contrast between tumor and fat was best on relatively T1-weighted images (500/30-35 [TR msec/TE msec]), whereas separation of tumor and muscle was best with relatively T2-weighted pulse sequences (1,500/90). Balanced images (1,500/30-35) provided best overall image quality and best demonstrated vascular anatomy. MR imaging was usually superior to CT in showing the relationship of tumor mass to muscle. MR imaging and contrast material-enhanced CT were equivalent in most patients in defining vascular anatomy, but MR imaging was superior when intravenous contrast material was not administered. However, CT was more helpful in showing bone and cartilage anatomy, and in some patients CT also was better in showing airway abnormalities. Despite these limitations, MR imaging is a promising imaging technique for studying neoplasms of the neck.     

Improved tissue characterization of focal liver lesions with ferumoxide-enhanced T1 and T2-weighted MR imaging

The purpose of our study was to evaluate the potential value of ferumoxide-enhanced T1-weighted magnetic resonance (MR) imaging for tissue characterization of focal liver lesions when combined with T2-weighted sequences. Images were acquired within 30 minutes after the end of ferumoxide administration, when ferrite particles were not totally cleared from the intravascular compartment. Thirty-eight patients with 47 focal liver lesions underwent T1-weighted gradient-echo (TR/TE 150/4.1 msec) and T2-weighted fast spin-echo (3180-8638/90 msec) MR imaging at 1.5 T before and after intravenous administration of ferumoxides (10 micromol/kg body weight). A qualitative and quantitative analysis was performed. During the early phase after infusion of ferumoxide, blood vessels showed hypersignal intensity on T1-weighted fast low-angle shot (FLASH) images, while liver signal decreased. Hemangiomas showed both homogeneous and inhomogeneous enhancement patterns, and liver metastasis most typically showed ring enhancement. Hypervascular tumors (hepatocellular carcinomas and focal nodular hyperplasias) showed a slight degree of homogeneous enhancement. Quantitatively, the degree of enhancement and lesion-to-liver contrast on ferumoxide-enhanced images were significantly different among these tumors. Our results demonstrate that distinct enhancement patterns obtained on ferumoxide-enhanced T1-weighted MR imaging improve tissue characterization of focal liver lesions when combined with T2-weighted images.     

Magnetic resonance imaging of experimental renal hemorrhage

Experimental renal hemorrhage was induced by injecting autologous blood into the left kidney of 13 rats. To investigate the magnetic resonance (MR) characteristics of acute renal hemorrhage and subsequent stages of resolution, repetitive MR images were obtained using a 0.35 Tesla imager during a period of 21 days postinduction. A dual spin-echo imaging (TR 500 and 2,000 msec, TE 28 and 56 msec) was used to calculate the relaxation times and record the intensities in the renal medulla and cortex. Histologic examination (n = 9) indicated that blood was dispersed intrarenally, and no encapsulated hematoma developed. The signal intensity on the T1- and T2-weighted images, as well as the relaxation times in the hemorrhagic renal parenchyma were unchanged during 21 days when compared with intact kidney values. Subcapsular fresh blood had a high signal intensity on T2-weighted images. A marked overlap of the relaxation parameters between intact kidney parenchyma and diffuse intrarenal hemorrhage was observed. Detection of dispersed intrarenal blood using spin echo MR imaging may be difficult.     

MR imaging of brain abscesses

The MR images and CT scans of 14 patients with surgically verified pyogenic cerebral abscesses were reviewed. The MR findings correlated well with those seen on CT and were believed to be sufficiently characteristic to allow early and accurate diagnosis with MR alone. These features include (1) peripheral edema producing mild hypointensity on short TR/short TE and marked hyperintensity on long TR/intermediate to long TE scans; (2) central necrosis with abscess fluid hypointense relative to white matter and hyperintense relative to CSF on short TR/short TE scans and hyperintense relative to gray matter on long TR/intermediate to long TE scans (the fluid had concentric zones of varying intensity in seven cases, a finding not previously identified in other lesions); (3) extraparenchymal spread (intraventricular or subarachnoid), which was detected more easily on MR than on CT and was manifested by increased intensity relative to normal CSF on both short TR/short TE and long TR/intermediate TE scans; and (4) visualization of the abscess capsule, which was iso- to mildly hyperintense relative to brain on short TR/short TE scans and iso- to hypointense relative to white matter on long TR/intermediate to long TE scans. On the long TR scans, the relative hypointensity of the rim allowed for visualization of the typical morphologic features of the capsule, which in turn aided in differentiation of abscesses from other lesions (as it does on CT). To investigate the cause of the capsular intensity, pathologic studies of the capsules were reviewed when available (10 cases). Fibrosis was identified in all mature abscess capsules, but the combination of the intensities seen on short TR/short TE and long TR/intermediate to long TE scans as well as the temporal changes in intensity were believed to be incompatible with fibrosis as a cause of the capsular changes. Intensity patterns were suggestive of hemorrhage, but neither acute nor chronic hemorrhage was identified on routine H and E stains, while iron stain revealed scant hemorrhage in only two of the eight patients in whom these stains were used. We believe the capsular intensity (in particular the hypointense rims on long TR scans) may reflect paramagnetic T1, and to a greater extent T2, shortening, possibly due to the presence of heterogeneously distributed free radicals that are products of the respiratory burst produced by actively phagocytosing macrophages in the capsule wall. Distinctive MR features of pyogenic abscesses should afford early and accurate diagnosis.     

Visualization of brain iron by mid-field MR

Brain iron was visualized on a mid-field (0.5 T) scanner using a spin-echo pulse sequence. Methemoglobin was hyperintense on T1- and T2-weighted images. Deoxyhemoglobin, hemosiderin, and ferritin were seen as decreased intensity on T2-weighted images. The spin-echo pulse sequences were improved for identification of deoxyhemoglobin, hemosiderin, and ferritin by prolonging the TR to 3000 msec and the TE to 80-120 msec. Phase-encoding artifacts at the level of the sylvian fissures caused increased noise, obscuring the brain iron in the lentiform nuclei with the TE of 120 msec. This artifact was substantially reduced or eliminated by lowering the TE to 80 msec, changing the phase-encoding gradient to the Y axis, or using additional pulsing in the slice and read gradients. Use of either the improved spin-echo or gradient-echo pulse sequences on a mid-field MR scanner provides improved evaluation of brain iron.     

Central nodal necrosis and extracapsular neoplastic spread in cervical lymph nodes: MR imaging versus CT.

Computed tomographic (CT) scans and magnetic resonance (MR) images obtained in 24 patients with cervical lymphadenopathy were retrospectively and blindly evaluated by two readers for the presence of central nodal necrosis (CNN) and extracapsular nodal spread (ENS). The CT studies were all enhanced, and the MR images were obtained with short repetition time (TR)/echo time (TE), long TR/double echo, and enhanced short TR/TE fat-suppressed sequences. Each MR imaging sequence was interpreted separately and then collectively. Sixty lymph nodes were identified with CT. Sensitivity for CNN was 16%-67% with the unenhanced MR pulse sequences, 50% with enhanced sequences, and 83%-100% with CT. The most accurate reading of MR images for CNN was with the unenhanced T1-weighted and T2-weighted images (86%-87%); the accuracy of CT was 91%-96%. The accuracy of MR imaging for detecting ENS was maximal with T1-weighted images (78%-90%). Gadolinium-enhanced, fat-suppressed images did not improve accuracy in evaluating CNN or ENS. CT is currently more accurate than unenhanced or enhanced MR imaging in detecting CNN or ENS.     

Anterior and posterior lobes of the pituitary gland: assessment by 1.5 T MR imaging

Pituitary glands of 60 normal volunteers (30 men 20-36 years old, and 30 women 18-42 years old) were studied by 1.5 T magnetic resonance (MR) imaging. The T1-weighted images (T1WI) [repetition time (TR) = 400 ms; echo time (TE) = 25 ms] were obtained in the coronal, sagittal, and axial planes. Proton density (PD)/T2-weighted images (PDWI/T2WI) (TR = 2,000 ms; TE = 25/100 ms) were obtained in the sagittal plane using 3 mm slice thickness. On T1WIs of all subjects the posterior part (PP) of the pituitary fossa showed the highest signal, which was indistinguishable from fatty tissue. This study reveals that this region of high signal intensity (PP) corresponds to the posterior lobe and not intrasellar fat because its shape, size, and position are compatible with the posterior lobe; its signal intensity differs from that of fatty tissue on PDWI and T2WI; the absence of an intrinsic chemical shift artifact (CSA) characteristic of fat; and due to CSA, a dorsum with fatty marrow is shifted relative to the PP (or may be made to merge with it). Regarding the differentiation of the two lobes of the pituitary gland on MR, the morphology of the anterior and posterior lobes was evaluated and great variation found. Appreciation of normal is particularly important in evaluating coronal images for small pituitary lesions.     

Magnetic resonance in the study of cerebral gliomas

Forty surgically proved gliomas have been studied by MR: 21 were low-grade gliomas and 19 were anaplastic astrocytomas and glioblastomas. The examination involved the use of T1-weighted (IR and/or SE, TR 500 ms, TE 50 ms) and T2-weighted scans (SE, TR 1000 ms, TE 50 with multiple echos). All patients underwent CT and surgical removal or stereotactic biopsy whose findings were retrospectively compared with those of MR. MR findings were similar in low-grade and anaplastic astrocytomas, but quite different from imaging of glioblastomas. Tumoral cyst and areas of necrosis were recognized on MR studies and confirmed by surgical findings. Differentiation between tumor and oedema was difficult. The MR images did not allow a more specific diagnosis of nature or of malignancy than available with CT up to now, but offer major advantages, such as a superior depiction of tumor extent and anatomic relationships.     

MR imaging of infectious spondylitis

MR images of 14 patients with pyogenic and three patients with tuberculous infectious spondylitis were studied to develop criteria for diagnosis. T1-weighted scans, 800/20 (TR/TE), were obtained in 17 patients and T2-weighted scans, greater than 2000/30,80, were obtained in 14. In seven patients, T2*-weighted scans (gradient-recalled acquisition into steady state, 25/15/5-7 degrees [TR/TE/flip angle]) and short-T1 inversion-recovery scans (STIR), 1400/150/40 (TR/TI/TE), as well as fat and water images (using a suppression technique), were obtained. Unenhanced and gadopentetate-dimeglumine-enhanced scans were obtained in four patients. In all but two patients with pyogenic infectious spondylitis, the T1-weighted sagittal scan showed characteristic findings: narrowed disk space, low signal intensity in the marrow of at least two adjacent vertebrae, subligamentous or epidural soft-tissue masses, and erosion of cortical bone. In one patient the T1-weighted scan was normal and abnormalities could be detected only on the T2-weighted scan. The remaining patient had abnormal marrow signal on the T1-weighted scan but only in one vertebral body. On T2-weighted images the major findings were a narrowed disk space with variable signal changes, abnormal high signal in marrow of at least two adjacent vertebrae, high-signal subligamentous or epidural masses, and cortical bone erosion. The findings in the three patients with tuberculous spondylitis included areas of increased and decreased signal intensity in vertebrae on T1-weighted images. Disk spaces were relatively spared given the extent of disease. Extraosseous soft-tissue components could be large. Bone erosion was best seen on the first echo of a T2-weighted sequence and on a water image; the latter was most reliable since it had no chemical-shift artifact. The use of gadopentetate dimeglumine could obscure or clarify MR findings, depending on the situation. T1- and T2-weighted MR images should be obtained for assessment of infectious spondylitis. STIR scans are particularly helpful. Fat images can be useful in subtle presentations, since they are very sensitive to marrow replacement, and gadopentetate dimeglumine may be helpful for epidural delineation of disease.