The b value is a parameter selected for the performance of DWI and accounts for multiple gradient terms describing how diffusion affects signal intensity in the following equation:

where γ is the gyromagnetic ratio, G is gradient strength, δ is the diffusion gradient duration, and Δ is the time between diffusion gradient pulses.

The b value reflects the acquisition parameters and is expressed as seconds per square millimeter. ADC values are expressed in the reciprocal square millimeters per second (although there is no consensus on applied units in DWI, it is always expressed as a ratio of area and time). At least two b values are needed to calculate the ADC, assuming a linear model for loss of signal intensity, but the optimal number and scale of b value selection are controversial. It has been recommended [8] that at least three b values be used for DWI with ADC quantification to help mitigate the fact that the lower signal-to-noise ratio (SNR) with higher b values introduces a higher SD in the calculation of the ADC map. Nevertheless, Bogner et al. [15] found that ADC calculations based on various combinations of 10 different b values ranging from 0 to 1250 s/mm² were not significantly more precise than those performed with only two values. Our approach is to use b values of 50, 400, and 800 s/mm². We avoid higher b values because of accompanying degradations in SNR and lower b values to reduce the contribution of blood perfusion to the ADC measurement. These values are in line with empirical evidence in favor of selection of b values with a minimum of 270 s/mm² and maximum of 800 s/mm² to minimize perfusion effects on ADC calculations [14].

A typical 3-T DWI protocol is outlined in Table 1. DWI is an unenhanced sequence that can be performed in an acquisition time of less than 5 minutes. Although manufacturers provide software for in-line pixel-by-pixel calculation of the ADC map, there is evidence suggesting approximately 10% variability among vendors for ADC values in the brain [16]. Imaging is performed in the axial plane to minimize distortion; distortion also increases with larger FOVs. Because DWI is performed with EPI and requires high magnetic-field-strength homogeneity, geometric distortion as a result of small local inhomogeneity may be problematic, particularly in the musculoskeletal system with its many tissue types and interfaces that cause susceptibility artifacts. One approach to addressing this issue is a phase map that delineates the B₀ field homogeneity and thereby allows certain distortions to be “unwrapped” [17]. The use of parallel imaging can reduce the echo-train length and thus decrease the overall amount of phase shift and minimize geometric distortion. Increasing the bandwidth may reduce geometric artifacts but can also be accompanied by increases in ghosting artifact and decreased SNR [18]. Finally, shorter TEs must be used in musculoskeletal tissues than in the CNS because musculoskeletal tissues usually have a shorter T2 than CNS tissues do. This mandates the use of shorter TEs and hence a lower range of b values that can be examined. Unfortunately, these factors (lower b values and shorter T2 times) adversely affect SNR.

View Larger Version

TABLE 1: Sample Diffusion-Weighted MRI Protocol

DW images and ADC maps can be analyzed both qualitatively and quantitatively. In general, free water causes higher signal intensity on ADC maps than more cellular tissues do. This is seen qualitatively as a steeper decrease in signal intensity on successively higher-b-value DW images for tissues containing more free water. Steeper decreases in signal intensity manifest themselves as low signal intensity on the corresponding ADC map. For quantitative purposes, there is no standard method for measuring ADC values in terms of how large the ROI should be, where it should be within the lesion (e.g., an area that appears qualitatively to have the lowest signal intensity vs the most representative area of the entire lesion), or whether mean, minimum, or maximum ADC values should be reported. The minimum ADC value [19] may more accurately reveal the nature of the lesion because it theoretically reflects the area of highest cellularity, whereas the partially necrotic portion of a tumor would elevate the overall mean ADC of a lesion and lead to underestimation of tumor cellularity in the remaining viable tissue. On the other hand, limiting ROI placement to arbitrarily small portions of a lesion could lead to image-selection bias and lower interreader and intrareader reliability. In light of these considerations, we recommend placement of ROIs to encompass as much of the lesion as possible in the region thought to have the lowest ADC (and presumably most cellular tissue) and measurement of minimum and mean ADC values (Fig. 2).

View larger version (462K)

Fig. 2A —39-year-old man with myxoid liposarcoma. A, Axial fat-saturated T2-weighted MR image (TR/TE, 5150/75) shows large hyperintense mass in medial and posterior left thigh with only thin fibrous or fatty bands, reflecting underlying predominantly myxoid stroma.

View larger version (514K)

Fig. 2B —39-year-old man with myxoid liposarcoma. B, Apparent diffusion coefficient (ADC) map shows relatively high free water content, as is common in myxoid tumors. Different techniques result in slightly different measurements. In larger ROI (solid circle), mean and minimum ADC values are 2.31 and 1.45 × 10−3 mm2/s. In smaller ROI (dashed circle) centered over more hypointense region, mean and minimum ADC values are 2.10 and 1.49 × 10−3 mm2/s.

The advent of parallel imaging techniques, stronger gradients, multichannel coils, and moving tables has enabled performance of DWI over the whole body. The development of a free-breathing technique, termed diffusion-weighted whole-body MRI with background body signal suppression, has made DWI available with multiple improved features, including increased SNR, acquisition unconstrained by breath-hold time limits or a particular breathing phase cycle, and multiple signal averaging. DWI can be performed with thin axial sections (≈ 4 mm) that enable volumetric image processing, including maximum intensity projections and multi-planar reformatting [20, 21]. Hence, DW images can be acquired in a convenient plane (typically coronal) and subsequently reconstructed into another plane of choice for optimal viewing. However, there are increased time requirements for use of the background body signal suppression method compared with other whole-body DWI methods. Moreover, there is a limited FOV if the users choose a single coil instead of a body coil [20]. Other whole-body DWI methods are performed with a set of coils placed on the body that result in excellent SNR and tissue contrast. In addition, the use of a rolling bed technology reduces imaging time and covers the entire body [22].

DWI is at least as effective as or more effective than a STIR-based protocol for detecting osseous metastasis from prostate and breast cancers and multiple myeloma [22, 23] (Fig. 3). A meta-analysis of the records of 495 patients who underwent whole-body MRI [24] showed that DWI was a sensitive but not specific technique for detecting bone lesions. Sommer et al. [25] found not only that DWI with background body signal suppression was helpful in the qualitative depiction of myeloma deposits but also that quantifiable differences in ADC values existed between patient subgroups. Those with a high serum M component melanoma (> 20 g/dL) had lower ADC values than those with a low serum M component myeloma (< 20 g/L), suggesting a role for DWI in risk stratification as well as in lesion detection. Whether the increased sensitivity attained with whole-body DWI will translate into changes in therapeutic strategies and better patient outcomes is a subject for future investigation.

View larger version (675K)

Fig. 3 —64-year-old man with metastatic prostate cancer. Axial diffusion-weighted images (DWI) from whole-body MRI study (left) show numerous metastatic foci (arrows) in axial and appendicular skeleton and corresponding level from stitched sagittal scout image. Axial DWI and apparent diffusion coefficient (ADC) map (center) show decreasing signal intensity (but stronger diffusion weighting) with successively higher b values. Lesions (arrows) are shown to particular advantage in vertebral bodies where ADC mapping confirms elevated values in metastatic lesions (1.11 × 10−3 mm2/s) compared with regions of normal marrow (0.40 × 10−3 mm2/s). Sagittal T2-weighted (TR/TE, 4000/96) and STIR (TR/TE, 2818/10) images (right) show whole-spine coverage corroborating DWI results. Arrow indicates lesion.

Tumor characterization—Anatomic sequences alone are usually sufficient to diagnose a small subset of soft-tissue masses with characteristic MRI findings, such as simple lipomas (homogeneous fat signal intensity in all pulse sequences) and juxtaarticular ganglia or cysts (absence of internal contrast enhancement or presence in a characteristic location, such as a Baker cyst). Because they are composed primarily of free water, ganglia and juxtaarticular cysts have high ADC values (Fig. 4). In cases in which a cyst is suspected, ADC mapping could prove a useful complement to anatomic imaging. It has been found [26] that use of a mean ADC value greater than 2.5 × 10⁻³ mm²/s yielded a sensitivity of 80% and a specificity of 100% in the diagnosis of cysts and benign cystic lesions, indicating that no soft-tissue neoplasms are missed with DWI and ADC mapping. ADC mapping may be particularly helpful in suggesting the true neoplastic nature of tumors, such as myxomas, that mimic cysts by displaying homogeneous T2 hyperintensity (Fig. 5). Nevertheless, other institutions using different DWI protocols have obtained mean ADC values greater than 2.5 × 10⁻³ mm²/s in soft-tissue myxomas [27]. Because of the frequency with which cystic-appearing soft-tissue masses are encountered, it would be helpful to establish a robust ADC cutoff that would be valid across institutions.

View larger version (224K)

Fig. 4A —74-year-old woman with right inguinal soft-tissue mass that has been slowly growing for 6 months. A, Axial T2-weighted MR image shows juxtaarticular round T2-hyperintense lesion (arrow) with internal septa.

View larger version (295K)

Fig. 4B —74-year-old woman with right inguinal soft-tissue mass that has been slowly growing for 6 months. B, Axial contrast-enhanced T1-weighted MR image shows no central enhancement within lesion.

View larger version (256K)

Fig. 4C —74-year-old woman with right inguinal soft-tissue mass that has been slowly growing for 6 months. C, Apparent diffusion coefficient (ADC) map shows lesion (solid circle) with mean and minimum ADC values of 2.63 and 2.16 × 10−3 mm2/s and, for comparison, ROI in urinary bladder (dashed circle) with mean and minimum ADC values of 3.0 and 3.54 × 10−3 mm2/s. Because of interval growth, biopsy was performed, and results confirmed presence of acellular myxoid material consistent with imaging characteristics of complex cyst.

View larger version (283K)

Fig. 5A —59-year-old woman with myxoma. A, Axial T2-weighted fat-suppressed MR image shows homogeneously hyperintense mass with thin septations in right axilla involving subscapularis muscle and abutting chest wall.

View larger version (408K)

Fig. 5B —59-year-old woman with myxoma. B, Contrast-enhanced fat-suppressed T1-weighted MR image shows relative lack of internal enhancement. Taken alone, this finding could falsely lead to suspicion of cystic lesion, although size and intramuscular location in this case would be strong evidence against this conclusion.

View larger version (252K)

Fig. 5C —59-year-old woman with myxoma. C, Apparent diffusion coefficient (ADC) map shows mean ADC of 2.03 × 10−3 mm2/s, which is below suggested threshold for reliable diagnosis of cystic lesions and thus would mandate biopsy to diagnose soft-tissue tumor.

Aside from lesions such as lipomas and ganglia, it is challenging to reliably characterize soft-tissue neoplasms as benign or malignant on the basis of imaging features alone. The hypothesis that malignant tumors would have lower ADC values due to increased cellularity has been explored in numerous studies with mixed results. Maeda et al. [28] found no significant difference in ADC values between malignant and benign soft-tissue tumors. Einarsdóttir et al. [27] found substantial overlap in the ADC values of 16 benign and 13 malignant soft-tissue masses (mean ADC, 1.8 and 1.7 × 10⁻³ mm²/s). In the study by Einarsdóttir et al., the highest ADC values for sarcoma were found in a myxoid liposarcoma. As such, some authors advocate considering myxoid and nonmyxoid tumors separately, because a myxoid matrix will generate higher ADC values [28]. Nagata et al. [29] found that although there was no significant difference between benign and malignant myxoid tumors, nonmyxoid malignant tumors had a lower mean ADC (0.94 × 10⁻³ mm²/s) than nonmyxoid benign lesions (1.31 × 10⁻³ mm²/s).

Other investigations have found a difference in ADC values between benign and malignant lesions. A study of eight desmoid tumors [30] showed that the mean ADC value of desmoids was higher (1.36 ± 0.48 × 10⁻³ mm²/s) than that of sarcomas (0.88 ± 0.20 × 10⁻³ mm²/s). With a wider range of pathologic findings, Razek et al. [31] reported that malignant tumors tend to have a lower mean ADC value than benign soft-tissue tumors. Those authors suggested using a threshold ADC value of 1.34 × 10⁻³ mm²/s to help in discriminating benignity from malignancy with an overall accuracy of 91%. Furthermore, within the group of malignant neoplasms, ADC values inversely correlated with histologic tumor grade. Figures 6 and 7 depict benign and malignant tumors studied with ADC mapping. They show how ADC maps delineate areas of higher cellularity even within benign tumors and that benign tumors tend to have a higher ADC value than malignant neoplasms.

View larger version (275K)

Fig. 6A —17-year-old boy with neurofibromatosis type 1. A, Coronal STIR MR image (TR/TE, 3000/17; inversion time, 150 ms) shows elongated T2-hyperintense mass along course of common fibular nerve in lateral aspect of upper left leg.

View larger version (247K)

Fig. 6B —17-year-old boy with neurofibromatosis type 1. B, Apparent diffusion coefficient (ADC) map shows mean value of 1.7 × 10−3 mm2/s with diffusion-weighted recapitulation of target sign (circle) commonly seen in these lesions on images obtained with fluid-sensitive sequences. Lower ADC values in center of lesion reflect more cellular and fibrocollagenous tissue, whereas higher ADC values in periphery are caused by less restricted water diffusion in myxomatous, less cellular component.

View larger version (386K)

Fig. 7A —61-year-old woman with anterior lower left leg soft-tissue mass subsequently proved to be dermatofibrosarcoma protuberans. A, Axial T2-weighted MR image shows superficial mass (arrows) over anteromedial lower left leg, abutting tibia, with heterogeneous T2 hyperintensity.

View larger version (538K)

Fig. 7B —61-year-old woman with anterior lower left leg soft-tissue mass subsequently proved to be dermatofibrosarcoma protuberans. B, Axial contrast-enhanced T1-weighted image shows heterogeneous enhancement.

View larger version (460K)

Fig. 7C —61-year-old woman with anterior lower left leg soft-tissue mass subsequently proved to be dermatofibrosarcoma protuberans. C, Apparent diffusion coefficient (ADC) map shows relatively homogeneous low ADC values (mean, 0.86 × 10−3 mm2/s) (arrow) reflecting high lesion cellularity in this malignant sarcoma. ADC values are low in uninvolved bone marrow of adjacent tibia (arrowhead) owing to presence of normal fatty marrow.

The discrepancies in the literature likely stem from the many factors that influence ADC values in addition to lesion cellularity, such as the composition of the tumor matrix, the presence of spontaneous necrosis, and nonuniformity in establishing imaging protocols for performing DWI with ADC mapping and analyzing the images.

Treatment response—The chemosensitivity of a sarcoma is best assessed histologically. However, being able to assess tumor response in vivo with MRI after only a few treatment cycles could provide important prognostic information and potentially shorten the duration of side effects of the prolonged administration of ineffectual agents. However, imaging assessment of tumor response is challenging, particularly in sarcomas, because a change in tumor size is not as useful a criterion as it is for other solid tumors. Enhancement patterns on both CT and MR images are more important in predicting pathologic response [32]. Morphologic changes that occur within sarcomas after a treatment response reflect the development of hyaline fibrosis and granulation tissue in addition to necrosis [33]. Because both granulation and scar tissue are enhancing after contrast administration, differentiation of these features from viable tumor after treatment is challenging on conventional images. Dynamic contrast enhancement and DWI are expected to improve in this discrimination [34].

Although response can be measured with PET/CT early in the course of treatment [35], MRI with DWI has the potential to provide both functional information and superior anatomic detail, particularly advantageous in surgical planning (Fig. 8). Because cellular changes are expected to precede morphologic changes in tumor volume, it has been suggested that DWI may show evidence of good treatment response earlier than conventional imaging [36]. In animal models, DWI findings correlate with changes in tumor perfusion [37], and minimum ADC values inversely correlate with tumor cellularity, whether imaging is performed de novo or after delivery of neoadjuvant therapy [38].

View larger version (248K)

Fig. 8A —7-year-old girl with malignant peripheral nerve sheath tumor in anterior forearm treated by chemotherapy followed by surgical resection. A, Sagittal contrast-enhanced fat-suppressed T1-weighted MR image shows solidly enhancing intramuscular mass.

View larger version (185K)

Fig. 8B —7-year-old girl with malignant peripheral nerve sheath tumor in anterior forearm treated by chemotherapy followed by surgical resection. B, Axial fat-suppressed T2-weighted MR image shows large T2-hyperintense mass in forearm.

View larger version (299K)

Fig. 8C —7-year-old girl with malignant peripheral nerve sheath tumor in anterior forearm treated by chemotherapy followed by surgical resection. C, Axial fat-suppressed T2-weighted image obtained 5 months after chemotherapy shows mass is markedly smaller than in A and B.

View larger version (316K)

Fig. 8D —7-year-old girl with malignant peripheral nerve sheath tumor in anterior forearm treated by chemotherapy followed by surgical resection. D, Diffusion-weighted image shows relatively high apparent diffusion coefficient value of 1.82 × 10−3 mm2/s within mass (arrows), indicative of good treatment response. (Diffusion-weighted imaging was not performed at original pretreatment examination at outside institution).

View larger version (291K)

Fig. 8E —7-year-old girl with malignant peripheral nerve sheath tumor in anterior forearm treated by chemotherapy followed by surgical resection. E, Contrast-enhanced image shows only thin rim enhancement surrounding central area of nonenhancement consistent with necrosis. At surgery 9 days after this MR study, site of original tumor was excised along with skin and deep dermal tissue. Tiny focus of residual malignant peripheral nerve sheath tumor was found at edge of tumor margin, presumably corresponding to peripheral rim of enhancing tissue found at MRI.

Bone marrow—The interpretation of DW images in the evaluation of bones differs considerably from that in evaluation of soft tissues in that malignant lesions tend to have higher ADC values than normal marrow [39–41]. Further complicating matters is that bone can be thought of as having two different states of normal: red and yellow marrow, each with a different range for normal ADC values. Yellow marrow consistently has lower ADC values than red marrow. This phenomenon is thought to occur because yellow marrow has a smaller fraction of free water and little extracellular matrix and because the larger lipid-laden cells in yellow marrow impede water movement to a greater degree than smaller hematopoietic cells of red marrow [9] (Fig. 7). The higher intramedullary blood flow in hematopoietic marrow increases the perfusion-weighted component of the ADC derived from lower b values [40]. These are particularly important considerations in evaluation of the pediatric population, because normal skeletal maturation, with an attendant increase in yellow marrow conversion, naturally decreases the ADC values in bone marrow [42]. Padhani et al. [43], however, found that normal marrow, whether predominantly red or yellow, still has lower ADC values than tumor. Those authors found that the 95th percentile for ADC values in osseous metastasis at a cutoff greater than 0.77 × 10⁻³ mm²/s resulted in a sensitivity of 85% and specificity of 90% for discriminating neoplastic marrow infiltration from normal marrow.

In work on the spine, Baur et al. [44] first proposed using DWI to discriminate pathologic from benign compression fractures on the assumption that water mobility would be more restricted in the presence of accumulated tumor cells and that trabecular disruption, attendant edema, and an increase in the inter-stitial space would all contribute to a high free water fraction. Results of a meta-analysis of 265 cases from eight studies in which ADC values were used [45] confirmed the initial observations by Baur et al. Despite ongoing controversy regarding the role of DWI in assessing vertebral body lesions [46], Dietrich et al. [47] summarized the ADC value ranges in vertebrae as follows: normal, 0.2–0.5 × 10⁻³ mm²/s; marrow infiltration, 0.7–1.0 × 10⁻³ mm²/s; and benign traumatic and osteoporotic fractures, 1.0–2.0 × 10⁻³ mm²/s. In our experience, similar values have been found for lesions in the appendicular skeleton.

Primary bone tumors—Compared with osseous metastasis, primary malignant bone tumors are rare and are traditionally best assessed with conventional radiography and MRI for initial characterization and determination of the need for and location of biopsy. There is a paucity of information on the role of DWI in characterizing untreated primary bone tumors. Hayashida et al. [48] found that in a small sample (n = 20) of T2-hyperintense bone lesions (bone cysts, fibrous dysplasia, and chondrosarcoma), ADC maps were not helpful for differentiating malignant from benign lesions. Those authors did find, however, that solitary bone cysts had higher mean ADC values than fibrous dysplasia and chondrosarcoma. In another series, Yakushiji et al. [49] found that minimum ADC values could aid in discriminating between chondroblastic osteosarcoma and chondrosarcoma despite the similarities in histologic features and the chondroid-type matrix enhancement pattern. Figures 9 and 10 show examples of malignant tumors involving bone.

View larger version (337K)

Fig. 9A —17-year-old boy with Ewing sarcoma metastatic to right sacrum. A, Axial T1-weighted MR image shows low-signal-intensity lesion (arrow) in right sacrum.

View larger version (263K)

Fig. 9B —17-year-old boy with Ewing sarcoma metastatic to right sacrum. B, Apparent diffusion coefficient (ADC) map shows area of intermediate signal intensity (mean, 1.92 × 10−3 mm2/s) (arrow) higher in adjacent uninvolved bone marrow (ADC value, < 0.5 mm2/s), indicative of tumor infiltration. Tumors infiltrating bone cause higher than normal ADC values due to disruption and loss of bony trabeculae and replacement of normal fatty marrow, which is hypointense on fat-suppressed diffusion-weighted image.

View larger version (305K)

Fig. 10A —36-year-old woman with breast cancer metastatic to right humerus and axilla. A, Coronal T1-weighted MR image shows confluent area of low signal intensity in right humeral metaphysis consistent with marrow replacement (metastasis).

View larger version (464K)

Fig. 10B —36-year-old woman with breast cancer metastatic to right humerus and axilla. B, Apparent diffusion coefficient (ADC) map shows restricted diffusion (mean ADC, 0.9 × 10−3 mm2/s) within tumor (arrow) in humeral metaphysis.

View larger version (533K)

Fig. 10C —36-year-old woman with breast cancer metastatic to right humerus and axilla. C, ADC map through right axilla shows multiple enlarged lymph nodes (arrow) with ADC values ranging between 0.8 and 1.1 × 10−3 mm2/s, suggestive of metastatic lymphadenopathy.

Treatment response—Although functional imaging with DWI and ADC mapping alone may not provide substantial insight into the diagnosis of primary bone tumors beyond that already afforded by radiography and conventional MRI, early results on the use of DWI for monitoring treatment are more promising. In a series of 18 patients with osteosarcoma and Ewing sarcoma [50], ADC values increased approximately 95% in tumors that had greater than 90% necrosis (indicative of good response), whereas changes in tumor volume and contrast-to-noise ratio on both T2-weighted and contrast-enhanced images were not significant. DWI assessment of treatment response in osteosarcoma has similarly shown that increased ADC values, but not overall tumor size, correlate with tumor necrosis. Uhl et al. [51] found that necrotic areas of eight treated osteosarcomas had increases in mean ADC ranging from 0.4 to 0.7 mm²/s, whereas ADC values in tumors responding poorly to neoadjuvant therapy had increased only up to 0.3 mm²/s. These results were echoed in a subsequent study of 22 patients with osteosarcoma, in which Oka et al. [19] found that ADC values increased in tumors after chemotherapy. They emphasized the importance of using minimum ADC values, which were significantly higher (1.01 ± 0.22 × 10⁻³ mm²/s) in patients with responding tumors (defined as > 90% tumor necrosis at resection) than in patients with a poor response (0.55 ± 0.29 × 10⁻³ mm²/s). The performance of DWI has compared favorably with results derived from PET/CT in terms of prediction of good tumor response (mean ADC increase > 13% resulted in an accuracy of 78%) [52]. Some authors also contend that corrections of ADC values for tumor volume result in further refinements that enhance the predictive power of DWI [53]. Finally, susceptibility artifacts at bone-soft tissue boundaries in osteogenic tumors may limit the accuracy and reproducibility of echo-planar DWI. Alternative non-echo-planar DWI techniques, such as periodically rotated overlapping parallel lines with enhanced reconstruction (PROPELLER) DWI, are associated with reduced susceptibility artifacts in regions of dense bone and may be better suited to musculoskeletal applications [47, 54].

The higher SNR at 3 T than at lower field strengths comes at the expense of more severe susceptibility artifacts of all types due to increased B₀. In addition, although single-shot EPI sequences are most commonly used for DWI, their long gradient-echo trains make them particularly sensitive to susceptibility artifacts induced at tissue boundaries, such as those occurring at the many fat, water, and bone interfaces in the musculoskeletal system. Use of parallel imaging, autoshimming, correction algorithms, and modified radiofrequency pulses (monopolar or bipolar) can mitigate these effects. Use of a segmented echo-planar readout (multishot technique) allows a shorter echo-train length and reduced susceptibility artifact, but at the expense of longer acquisition time [47]. A variety of other DWI sequences less vulnerable to such limitations include spin-echo, stimulated echo, single-shot with multiple spin echoes, HASTE or single-shot fast spin-echo diffusion imaging, turbo spin-echo imaging, PROPELLER, and line scan diffusion imaging [55].

DWI sequences are less sensitive to patient motion than other MRI sequences [17], but enough patient motion can result in the large phase shifts that are acquired view to view and can add to or subtract from the phase-encoding process. The net result of motion is a ghosting artifact on DW images. This artifact can be mitigated by the adoption of single-shot EPI and motion correction (navigation), but interacquisition motion generally mandates repeat acquisition [56]. In addition, single-shot EPI DWI requires fat suppression, and images and ADC values derived from this type of sequence may be unreliable when fat suppression fails or is inhomogeneous.

In terms of interpretation pitfalls, the DWI novice may misconstrue the inherent T2 weighting of the diffusion-sensitive sequence as bona fide restricted diffusion [34]. This T2 shine-through error can be avoided with careful assessment of the higher-b-value images and the corresponding ADC map, which depict genuine restricted diffusion as low signal intensity. Another pitfall in DWI interpretation is the lowered sensitivity of DWI to sclerotic osseous lesions in the evaluation of bone lesions [57]. This can result in false-negative results because the sclerosis manifests itself as low signal intensity and remains inconspicuous against normal low-signal-intensity marrow on DW images [58]. Such a problem may account for some of the controversy in the literature regarding the accuracy of DWI in characterizing vertebral body lesions [47].

The ADC profile of hematomas may deviate from the principle that solid soft-tissue masses have lower ADC values than fluid collections. Oka et al. [59] found that the mean ADC value of chronic expanding hematomas (1.55 ± 0.12 mm²/s) was higher than that of malignant soft-tissue tumors (0.92 ± 0.14 mm²/s) (p < 0.01). However, there were only six chronic expanding hematomas in that small series, and it is unclear how ADC values in such hematomas would differ from those in benign neoplasms. That hematomas can have low ADC values is a known pitfall in body DWI [60]. DWI, like T2-weighted MRI, shows varying signal intensity as intracranial hematomas evolve over time. ADC values, on the other hand, have been found to be consistently low throughout the course of hematoma evolution [61]. Similar studies in the body are needed because the imaging characteristics of hemorrhage in the musculoskeletal system are unlike those in the brain and may not follow a predictable pattern of evolution.

Soft-tissue abscesses occasionally mimic soft-tissue neoplasms [62], particularly when thick or nodular rim enhancement is present after contrast administration. Although clinical context is frequently sufficient to ascertain whether a mass is inflammatory or neoplastic, it is important to be aware that there is substantial overlap in the DWI characteristics of abscesses and tumors. Because the high-viscosity pus within abscesses contains inflammatory cells, cellular debris, bacteria, and proteins, water diffusion may be slowed. In their review of eight cases of soft-tissue abscesses, Harish et al. [63] found that DWI improved the conspicuity of abscesses in two cases. In a third case, IV contrast medium could not be administered because of renal failure, emphasizing the utility of DWI as an unenhanced technique. In a review of the records of 50 patients with apparent soft-tissue cystic lesions [64], DWI was found to have 92% sensitivity and 80% specificity for the diagnosis of abscess (with aspiration as the reference standard). In that study, two cases were prospectively interpreted as abscesses but later found histologically to be superinfected neoplasms. Ultimately, the overlap in DWI characteristics underscores the need for use of a comprehensive set of MRI sequences and careful attention to clinical context when there is high pretest probability of infection.

Finally, because fat has a diffusion coefficient approximately two orders of magnitude less than that of water, lipid contamination in the ROI will artifactually lower the measured ADC [65]. Thus, the more lipomatous component of a liposarcoma will have low ADC values, which can result in misinterpretation of these areas as hypercellular regions of tumor (Fig. 11). Although evidence from other prostate biopsies shows that use of DWI increases the chance of obtaining representative tissue [66], such an error could lead to suboptimal targeting during imaging-guided biopsy and reinforces the necessity of correlating ADC maps with images obtained with other anatomic sequences.

View larger version (436K)

Fig. 11A —39-year-old man with myxoid liposarcoma in left thigh (same patient as in Fig. 2). A, Apparent diffusion coefficient (ADC) map shows subtle area of lower signal intensity (circle) within mass (mean value, 2.01 × 10−3 mm2/s; overall tumor mean, 2.31 × 10−3 mm2/s) that can be mistaken for region of more cellular tumor.

View larger version (504K)

Fig. 11B —39-year-old man with myxoid liposarcoma in left thigh (same patient as in Fig. 2). B, Axial contrast-enhanced fat-suppressed T1-weighted (B) and T1-weighted (C) images show area in A is simply focus of lipomatous tissue (arrow) in myxoid liposarcoma. ADC maps should always be interpreted alongside anatomic images to avoid misinterpretation.

View larger version (470K)

Fig. 11C —39-year-old man with myxoid liposarcoma in left thigh (same patient as in Fig. 2). C, Axial contrast-enhanced fat-suppressed T1-weighted (B) and T1-weighted (C) images show area in A is simply focus of lipomatous tissue (arrow) in myxoid liposarcoma. ADC maps should always be interpreted alongside anatomic images to avoid misinterpretation.