Vascular and Interventional Radiology
Original Research
Utility of Balanced Steady-State Free Precession MR Venography in the Diagnosis of Lower Extremity Deep Venous Thrombosis
Audio Available | 
Share
OBJECTIVE. The purpose of this study was to determine the sensitivity and specificity of balanced steady-state free precession MR venography in the diagnosis of lower extremity deep venous thrombosis.
SUBJECTS AND METHODS. After undergoing lower extremity ultrasound because of suspicion of deep venous thrombosis, 64 patients were prospectively recruited to undergo balanced steady-state free precession MR venography with ultrasound as the reference standard. Ultrasound images were independently interpreted by two blinded ultrasound radiologists, and MR venograms were independently interpreted by two blinded MRI radiologists. The sensitivity, specificity, positive predictive value, and negative predictive value of MR venography were calculated for the diagnoses of all deep venous thrombosis, acute thrombi, and thrombosis of the popliteal, femoral, and common femoral veins individually. Proximal extent, thrombus age, ancillary findings, and interobserver agreement calculated with the Cohen kappa test were evaluated for ultrasound and MRI. The McNemar test was used to evaluate for statistical differences in diagnostic accuracy.
RESULTS. MR venography had a sensitivity of 94.7%, specificity of 100%, positive predictive value of 100%, and negative predictive value of 97.7% for the diagnosis of all thrombi. For acute thrombi, the MR venography and ultrasound results were completely concordant. MR venography depicted greater proximal extent in five of 18 cases in which thrombosis was found. The MR venographic findings agreed completely with the ultrasound findings in determination of thrombus age. For both ultrasound and MR venography, interobserver agreement was 100% on a per-patient basis. No statistical difference was identified in the diagnostic performance of the two techniques.
CONCLUSION. Balanced steady-state free precession MR venography is highly accurate in the diagnosis of lower extremity deep venous thrombosis.
Keywords: balanced steady-state free precession, deep venous thrombosis, MR venography, trueFISP, ultrasound
| Introduction |   |
|---|
Ultrasound has replaced conventional venography as the test most widely used for the diagnosis of lower extremity deep venous thrombosis (DVT). The sensitivity and specificity of ultrasound are in the range of 95–99% in the popliteal vein and the deep veins of the thigh, and these values are comparable to those of conventional venography [1, 2]. However, ultrasound has limitations. Because compression is required, it is difficult or impossible to perform on persons in casts and persons who are obese or have marked leg swelling [3–9]. The use of ultrasound is limited by scars, wounds, and immobilization devices [4, 8, 10], and its sensitivity and specificity are highly operator dependent [5]. A large portion (as much as 46%) of the population has duplicated leg veins. If a thrombus exists in only one branch of a duplication, the examination findings may be reported as normal if only the nonthrombosed branch is visualized [2, 4, 11–13]. The iliac veins and the femoral vein where it passes through the adductor hiatus are especially difficult to image with compression sonography [4–6, 14, 15]. Ultrasound has been a successful technique for DVT assessment because it is convenient for referring physicians, allows comprehensive assessment, and eliminates both the nephrotoxicity of contrast material and radiation [16].
To have a place in the evaluation of DVT, MRI should offer all of the advantages of ultrasound and act as an alternative test in situations in which ultrasound is not ideal. Balanced steady-state free precession (SSFP) pulse sequences are commonly used in cardiac MRI to achieve high contrast between myocardium and blood without gadolinium administration [17, 18]. In several studies [19–21], these sequences have been used to assess the pulmonary arteries, aorta, and coronary arteries. Gradients in balanced SSFP sequences are balanced by others of opposite polarity to eliminate gradient-induced dephasing, and when combined with short TR and TE acquisitions, the steady-state sequences are associated with high signal-to-noise ratio. Image contrast is related to the T2/T1 ratio, and blood has intrinsically high signal intensity unrelated to flow characteristics. This characteristic minimizes the substantial artifact associated with slow-flowing blood in other unenhanced MR venographic protocols, such as gradient-recalled echo. Balanced SSFP sequences are widely available, and the short TR allows rapid motion-resistant acquisition and acquisition times less than 10 minutes for extremity MR venography [22–24]. The balanced SSFP pulse sequences have been found highly accurate in evaluation of the abdominal veins (Radulovic D et al., presented at the 2006 annual meeting of the International Society for Magnetic Resonance in Medicine).
The properties of balanced SSFP sequences lead to the hypothesis that balanced SSFP MR venography is a quickly performed alternative method of diagnosis of DVT from the bifurcation of the inferior vena cava to the popliteal vein. We hypothesize that the diagnostic accuracy of balanced SSFP MR venography is comparable to that of the reference standard, ultrasound, in terms of clot detection and characterization of acute as opposed to chronic thrombosis.
| Subjects and Methods | |
|---|
Institutional ethics board approval was obtained before initiation of the study protocol. The study protocol ran during the months of May to August in 2007 and 2008. Because of limited availability of MRI, patients were not recruited consecutively. Sixty to 70 patients were expected to be recruited into the study during this time period. The inclusion criteria were adult age and undergoing lower extremity ultrasound for suspicion of DVT. When an MRI time slot was available, all patients undergoing sonography for suspicion of lower extremity DVT were approached by a study coordinator about undergoing MRI. Informed consent was obtained from all patients, and patients who agreed to participate were screened for contraindications to MRI. Those who agreed had given informed consent, and had no contraindications underwent balanced-SSFP MRI. The interpreting radiologists played no role in patient recruitment, and the study coordinator played no role in image interpretation. Patients were excluded if they had contraindications to MRI or could not provide informed consent or if study participation might result in a clinically significant delay in care. Sixty-four patients were eventually recruited into this prospective study.
Ultrasound examinations were performed with standard techniques of real-time gray-scale evaluation and graded compression, color Doppler analysis, and pulsed Doppler waveform acquisition with respiration and augmentation. An experienced vascular ultrasound technologist performed the scans (Sequoia system, Acuson) from the common femoral vein to the popliteal vein trifurcation only in the leg with symptoms. The examination was extended to the inferior vena cava bifurcation when possible depending on patient size. Examinations took 15–30 minutes per leg. The images obtained by technologists were interpreted by two specialized ultrasound radiologists. The decision to initiate anticoagulation therapy was based on the ultrasound findings. All studies were checked by the radiologist before patient discharge from the department.
The MRI examinations were performed with a 32-channel 1.5-T system (Avanto, Siemens Healthcare) (45 mT/m maximum amplitude gradients, 200 T/m/s maximum slew rate) with a combination of a 6-channel body matrix array coil and an 8-channel extremity coil. After automated 3D shimming, transverse balanced-SSFP images (true fast imaging with steady-state precession; TR/TE, 3.7/1.8; flip angle, 70°; matrix size, 256 × 192; field of view, 30 cm2; slice thickness, 5-mm interleaved with a 20% gap) were obtained from the bifurcation of the inferior vena cava to the trifurcation of the popliteal vein of only the leg with symptoms. Examinations took 7–10 minutes of imaging time, for a maximum total of 15 minutes including all setup. No postprocessing of the images was needed.
If the patient was fit to undergo MRI, imaging was performed within 24 hours after the ultrasound examination. Although 24 hours was the official cutoff, this extended timeframe was necessary only for two patients owing to lack of availability of MRI. All other patients underwent imaging within 6 hours. The MRI did not interfere with patient therapy if DVT was found with ultrasound.
For both ultrasound and MRI, each examination was recorded as having positive or negative results for DVT. If the findings were positive, the proximal extent, thrombus age (acute or chronic), and involved venous segments were recorded. In all cases, ancillary findings were recorded.
Two ultrasound radiologists blinded to the MRI results independently evaluated the ultrasound images for evidence of DVT. The diagnosis of DVT was based on lack of compressibility and the absence of flow in the region on color Doppler images. Absent or abnormal pulsed Doppler waveforms, although considered supportive evidence, were not used as the sole criterion for a diagnosis of DVT. Evidence of chronic DVT was considered either the presence of venous wall thickening with or without weblike filling defects or a reduction in venous diameter or atresia.
The diagnostic criterion for DVT at balanced-SSFP MR venography were a definite filling defect within one of the visualized veins or, in the case of completely occlusive thrombus, complete lack of signal intensity within the vein. The MRI definition of chronic DVT was similar to the ultrasound definition (Figs. 1A, 1B, 1C, and 1D). All images were read independently by two fellowship-trained MRI radiologists blinded to the ultrasound results.
In the case of a discrepancy between ultrasound and MRI findings, the study coordinator asked the readers to jointly review the case. If a consensus was reached through joint review of both the ultrasound and MR images, the finding was counted accordingly as true- or false-positive or true- or false-negative. Consensus review resolved three discordant interpretations. If no consensus was reached, the patient underwent a repeated ultrasound examination with all radiologists in attendance. This step was necessary for one patient.
Statistical analysis was performed in consultation with the department statistician. Sensitivity, specificity, positive predictive value, and negative predictive value were calculated on a per-patient basis with regard to the diagnosis of all (acute and chronic) DVT and to acute DVT only. The popliteal vein, femoral vein, and common femoral vein were analyzed independently for the diagnosis of DVT with balanced-SSFP MRI. The Cohen kappa test was used to calculate interobserver agreement between the two readers for each technique and between MRI and ultrasound. The McNemar test for marginal homogeneity was used to evaluate for statistical differences in diagnostic accuracy between ultrasound and MRI. Statistical significance was defined as p < 0.05. The external iliac and common iliac veins and inferior vena cava were not evaluated independently because ultrasound is not reliable as a reference standard for these veins. Thrombus age and proximal extent were compared in cases of positive findings between ultrasound and balanced-SSFP MRI. Ancillary findings were compared between the two techniques, with particular attention being paid to duplication of venous segments.
| Results | |
|---|
The one indeterminate ultrasound examination was of a patient with a suspected vein wall hematoma, and the ultrasound findings did not definitively exclude DVT. MRI showed no evidence of DVT, and a hematoma around the common femoral vein was directly visualized. This case was excluded from statistical analysis because ultrasound was not used as the reference standard.
In another patient, balanced-SSFP MRI showed abnormal signal intensity suggesting the presence of chronic popliteal DVT. Owing to patient size and discomfort, sonography did not depict the popliteal vein in enough detail to confirm this finding. The patient did not return for repeated ultrasound, and because consensus review did not establish a diagnosis, this patient was excluded from analysis.
 View larger version (176K) | Fig. 1A —57-year-old woman with right leg pain and swelling. Transverse balanced steady-state free precession MR image obtained at level of femoral vein shows uniformly high intravascular signal intensity (arrow) in vein. There is no evidence of filling defect. |
 View larger version (160K) | Fig. 1B —57-year-old woman with right leg pain and swelling. Transverse balanced steady-state free precession MR image obtained at level of femoral vein shows completely occlusive intravascular filing defect (arrow). Serial slices showed that this filling defect was present at multiple levels in vein. Filling defect is solid and has uniformly low signal intensity consistent with acute thrombosis of femoral vein. |
 View larger version (174K) | Fig. 1C —57-year-old woman with right leg pain and swelling. Transverse balanced steady-state free precession MR image at level of femoral vein shows no convincing filling defect in vein but does show heterogeneous intravascular signal intensity (arrow). This heterogeneity disappeared on serial slices and has higher signal intensity and less definition than actual thrombus. Finding is highly indicative of artifact and is not diagnostic of deep venous thrombosis. |
 View larger version (154K) | Fig. 1D —57-year-old woman with right leg pain and swelling. Transverse balanced steady-state free precession MR image at level of femoral vein shows eccentric intravascular filling defect with associated wall thickening (arrow). This appearance is indicative of chronic thrombus. |
 View larger version (204K) | Fig. 2A —50-year-old man with chronic popliteal deep venous thrombosis. Ultrasound image of left popliteal vein shows wall thickening (arrow) consistent with chronic deep venous thrombosis confirmed at real-time acquisition. |
 View larger version (179K) | Fig. 2B —50-year-old man with chronic popliteal deep venous thrombosis. Transverse balanced steady-state free precession MR venogram shows heterogeneous signal intensity along outer contour of popliteal vein (arrow) dismissed as artifact by both MRI readers but in fact representing chronic thrombus. |
 View larger version (238K) | Fig. 3A —88-year-old woman with chronic popliteal deep venous thrombosis. Ultrasound images before (A) and after (B) compression show complete compression (arrow) of popliteal vein at this level. Sonographic findings were reported as normal. |
 View larger version (242K) | Fig. 3B —88-year-old woman with chronic popliteal deep venous thrombosis. Ultrasound images before (A) and after (B) compression show complete compression (arrow) of popliteal vein at this level. Sonographic findings were reported as normal. |
 View larger version (198K) | Fig. 3C —88-year-old woman with chronic popliteal deep venous thrombosis. Transverse balanced steady-state free precession MR image of popliteal vein shows intravascular filling defect (arrow) within vein. Finding is highly suggestive of chronic thrombus. |
 View larger version (210K) | Fig. 3D —88-year-old woman with chronic popliteal deep venous thrombosis. Focused repeated sonogram shows initially missed evidence of chronic thrombus (arrow) lying on cusp of venous valve. |
In the 62 remaining cases, the correct diagnosis was made with balanced-SSFP MRI in comparison with ultrasound in all cases but one. In that case, ultrasound showed wall thickening suggesting the presence of a chronic femoropopliteal thrombus not seen on MRI (Fig. 2A). At consensus review, it was determined that chronic thrombus was indeed present on both the ultrasound and MR images, resulting in a false-negative MRI interpretation (Fig. 2B). Balanced-SSFP images of another patient depicted a strandlike filling defect suggestive of chronic thrombus not detected with ultrasound. Consensus review of both sets of images indicated a need for repeated sonography because the balanced-SSFP MR venographic findings appeared real. Focused sonography of the popliteal vein depicted a remote thrombus on a venous valve, resulting in a true-positive MRI finding (Figs. 3A, 3B, 3C, and 3D). This finding was consistent with patient history of previous DVT in the region. In another case, common femoral vein thrombus was seen with MRI and ultrasound. However, one MRI reader detected femoral vein thrombus not reported on ultrasound. At consensus review of both sets of images the readers concluded that femoral vein thrombus was indeed present at sonography, making this a true-positive MRI finding.
In comparison with ultrasound, balanced-SSFP MRI had a sensitivity of 94.7%, specificity of 100%, positive predictive value of 100%, and negative predictive value of 97.7% in the diagnosis of all (acute and chronic) DVT (Table 1). In the 62 cases, interobserver agreement was 100% for both MRI and ultrasound (κ = 1.000). MRI and ultrasound findings agreed in the cases of 61 of 62 patients (98.4% agreement; κ = 0.9615; 95% CI, 0.8867–1.000). The discordant case was the case of chronic femoropopliteal thrombus described earlier. The McNemar test result showed no statistical difference between MRI and ultrasound in the diagnosis of DVT on a per-patient basis.
For the diagnosis of acute clinically relevant thrombi that would require anticoagulation therapy, balanced-SSFP MR venography had a sensitivity, specificity, positive predictive value, and negative predictive value of 100%. Interobserver agreement in the 62 cases was 100% between ultrasound readers, between MRI readers, and between the two techniques (κ = 1.000). The McNemar test result showed no statistical difference between MRI and ultrasound in the diagnosis of acute DVT.
The diagnostic accuracy of the MRI readers with respect to the popliteal, femoral, and common femoral veins is shown in Table 2. The McNemar test result showed no statistical difference between the ultrasound and MRI readers in diagnosis of DVT in these veins. In one ultrasound examination the popliteal vein was not visualized because open wounds and leg pain precluded compression. However, a patent popliteal vein was clearly visualized with MRI. This patient was excluded from statistical analysis of the popliteal vein only because ultrasound did not depict this region. In the 61 remaining cases, interobserver agreement was 100% for both MRI and ultrasound in the diagnosis of all (acute and chronic) popliteal DVT (κ = 1.000). MRI and ultrasound findings agreed in the diagnosis of 60 of 61 cases of popliteal DVT (98.4% agreement; κ = 0.9585; 95% CI, 0.8778–1.000).
For all femoral vein DVT, the two MRI readers agreed on the diagnosis in 61 of 62 cases (98.4% agreement; κ = 0.9550; 95% CI, 0.8676–1.000). The two ultrasound readers agreed in all 62 cases (100% agreement; κ = 1.000). The first MRI reader agreed with the ultrasound findings in the diagnosis of 61 of 62 cases (98.4% agreement; κ = 0.9570; 95% CI, 0.8375–1.000). The second MRI reader agreed with the ultrasound findings in the diagnosis of 60 of 62 cases (96.8% agreement; κ = 0.9122; 95% CI, 0.7929–1.000).
Regarding the common femoral vein, the two MRI readers agreed in the diagnosis of 61 of 62 cases (98.4% agreement; κ = 0.9379; 95% CI, 0.8173–1.000). The two ultrasound readers agreed in all 62 cases (100% agreement; κ = 1.000). The first MRI reader agreed with the ultrasound findings in all 62 cases (100% agreement; κ = 1.000). The second MRI reader agreed with the ultrasound findings in the diagnosis of 61 of 62 cases (98.4% agreement; κ = 0.9379; 95% CI, 0.8173–1.000).
 View larger version (170K) | Fig. 4A —57-year-old obese woman with right leg pain and swelling. Ultrasound image at femoral vein level has poor image quality owing to patient size. It is difficult to visualize full compression (arrow, B) between precompression (A) and postcompression (B) images. |
 View larger version (198K) | Fig. 4B —57-year-old obese woman with right leg pain and swelling. Ultrasound image at femoral vein level has poor image quality owing to patient size. It is difficult to visualize full compression (arrow, B) between precompression (A) and postcompression (B) images. |
 View larger version (184K) | Fig. 4C —57-year-old obese woman with right leg pain and swelling. Transverse balanced steady-state free precession MR image at same level as A and B shows large amount of subcutaneous fat that does not affect image quality. Uniformly high signal intensity (arrow) is evident in femoral vein. No filling defect is evident, and vein is thrombus-free at this level. |
All decreased diagnostic accuracy of balanced-SSFP MR venography resulted from the evaluation of two patients. In one patient, chronic femoropopliteal thrombus was visualized with ultrasound but not with MRI, resulting in false-negative findings in the popliteal and femoral veins for both MRI readers. In the other patient, the first MRI reader correctly identified thrombus with two foci: a thrombus in the common iliac and external iliac veins that was discontinuous from another thrombus in the common femoral and femoral veins. The second reader identified thrombus only in the common iliac vein, leading to a false-negative result in the common femoral and femoral veins. The second case constituted all interobserver disagreement on balanced-SSFP MRI findings.
In 13 of the 18 cases of DVT detected with both techniques, balanced-SSFP MRI and ultrasound showed the same proximal extent. In the other five cases, MRI depicted greater proximal extent, consisting of two thrombi extending into the inferior vena cava and three into the common iliac vein. In no case did ultrasound show greater proximal extent than balanced-SSFP MRI.
In the 18 cases of thrombosis that were seen with both imaging techniques, MRI and ultrasound agreed completely, showing 13 acute thrombi, three chronic thrombi, and two acute on chronic thrombi.
MRI depicted significantly more incidental findings than ultrasound. The additional findings with ultrasound were limited to air in the calf tissues, knee joint effusion, lower extremity edema, Baker's cyst, varicose veins, and enlarged uterus. MRI reliably depicted all of these findings and revealed additional clinically relevant findings in some patients. These findings included an intus-susception, two abdominal aortic aneurysms, ascites in two patients, a dysmorphic kidney, cholelithiasis, a splenic infarct, a bilateral hydrocele, two hernias, small-bowel thickening, and iliac and inguinal adenopathy.
MRI showed duplication of the femoral vein in 12 of 64 patients (18.8%) and duplication of the popliteal veins in seven of 64 patients (10.9%). Only one duplicated femoral vein was seen with ultrasound, and no popliteal duplications were seen. Three of the 12 patients with duplicated femoral veins detected with balanced-SSFP MRI had femoral vein thrombosis. In two of these patients, the thrombus did not extend to the level of duplication, and both limbs of the duplicated segment were patent. The third patient had thrombus completely occluding both limbs of the duplication. The ultrasound images of this patient depicted only one thrombosed venous segment and did not depict the other thrombosed branch of the femoral vein. In the patient with femoral vein duplication seen with both ultrasound and MRI, there was no thrombus.
 View larger version (164K) | Fig. 5 —67 year old man with recent right total hip arthroplasty. Transverse balanced steady-state free precession MR image shows common femoral vein (arrow) at level of arthroplasty. Although some artifact is present, it did not prevent venous assessment. This case was the worst example of susceptibility in the series, but any hypointensity in vein was clearly artifactual, extending beyond margins of vein and not hindering assessment. |
Only one of the seven patients in this study with duplication of the popliteal vein on balanced-SSFP MRI had popliteal thrombus. In this patient, the thrombus did not involve the duplicated segment.
| Discussion | |
|---|
Although many MR venographic protocols have high diagnostic accuracy for DVT, to date, to our knowledge, no single protocol has been widely adopted into clinical practice. Unenhanced techniques such as spin-echo, time-of-flight, and gradient-recalled echo are not in routine clinical use because of their long imaging times [12, 16, 25–28]. Some techniques described have imaging times as rapid as 4 minutes, but many of them require gadolinium administration [7, 12, 22, 28].
The ideal MR venographic technique for DVT would be rapid, highly accurate, and avoid contrast administration. Because MR venography generally allows superior visualization of pelvic veins and incidental findings, these benefits should be documented in any proposed protocol [3, 7, 14, 28–30]. We believe balanced-SSFP MR venography may be unique in that it satisfies all of these criteria and is readily performed with most commercially available 1.5-T systems. Although ultrasound is a useful test for most patients with clinical suspicion of DVT, a large patient population exists for whom the test is exceedingly painful or technically challenging. For these patients, a painless imaging technique that is unaffected by patient size, leg edema, surgical scars, and immobilization devices would be beneficial. This is especially true because of the increasing prevalence of obesity, the fact that pain and edema are the most common symptoms of DVT, and the frequency with which patients experience postoperative thrombotic complications.
In this study, the imaging time for balanced-SSFP MRI was 7–10 minutes, and the average time for an ultrasound examination was approximately 20 minutes. This difference represents another benefit of balanced-SSFP MRI: that it takes approximately one half the time of an ultrasound examination. The absence of contrast administration decreases cost substantially and eliminates the risk of nephrogenic systemic fibrosis [31]. Although the cost of balanced-SSFP MRI likely exceeds that of ultrasound, this rapid MRI protocol can be performed entirely by technologists. Thus the cost of staffing is less than that of fluoroscopic contrast venography, which requires radiologist involvement. Because balanced-SSFP MRI requires neither contrast administration nor nursing supervision, these features impart cost savings over contrast-enhanced CT and MR venography. It is estimated that balanced-SSFP MRI may become the least expensive test for evaluation after nondiagnostic ultrasound. Balanced-SSFP MRI is painless, relatively operator independent, and not hindered by wounds, surgical scars, or edema. These factors were important in this study because 18 of 64 patients had clinically significant leg edema. Obesity was not a problem if the patient's weight was less than the weight limits of the MRI unit (Figs. 4A, 4B, and 4C). Three ultrasound examinations were reported as being technically difficult to perform, and three ultrasound reports indicated poor visualization of venous segments owing to obesity, edema, position within the adductor hiatus, or overlying bowel. There were no instances of technically challenging MR venography.
We found balanced-SSFP MRI highly accurate in the diagnosis of DVT, the findings agreeing with the ultrasound findings in 61 of 62 cases. No MRI examinations were nondiagnostic, and the MRI findings resolved the diagnosis in the one indeterminate ultrasound examination. Balanced-SSFP MRI correctly depicted a chronic thrombus missed with ultrasound. It also clearly depicted the popliteal vein of a patient who did not tolerate popliteal compression. The one discordant case on a per-patient basis was a chronic femoropopliteal thrombus, thus the missed detection of this clot on MRI would not have affected treatment. Other than this case, MRI and ultrasound findings agreed completely in the differentiation of acute and chronic thrombi. Balanced-SSFP MRI had a sensitivity of 94.7%, specificity of 100%, positive predictive value of 100%, and negative predictive value of 97.7% for the diagnosis of DVT. For acute clinically relevant thrombi, which would necessitate anticoagulation therapy, the diagnostic accuracy was equal to that of ultrasound. Balanced-SSFP MRI was highly accurate in the popliteal vein, femoral vein, and common femoral vein. No statistical difference was identified between the diagnostic accuracy of ultrasound and that of balanced-SSFP MRI for any of these categories. Balanced-SSFP MRI also was superior in pelvic imaging and depicting the extent of proximal thrombus. This information is beneficial because the likelihood of pulmonary embolism increases as the thrombus extends proximally [27].
Balanced-SSFP MRI outperformed ultrasound for ancillary findings. In particular, the technique was far superior to ultrasound in the detection of duplicated venous segments, representing another advantage of balanced-SSFP MRI. Balanced-SSFP MR venography depicted 12 femoral vein duplications and seven popliteal vein duplications. Ultrasound imaging depicted only one femoral vein duplication, that was also seen with balanced-SSFP MRI, and did not depict any popliteal duplications. Because of the superior detection of duplicated veins with balanced-SSFP MRI, it is less likely that a thrombosed duplication would be missed with that technique than it would with ultrasound. Balanced-SSFP MRI did depict one duplicated femoral vein in which both limbs were thrombosed. The ultrasound images of this patient depicted thrombus in only one limb of the femoral vein and did not depict the duplication. The detection of both thrombosed segments in this patient supports the notion that use of balanced-SSFP MRI decreases the chance of visualizing only the patent limb of a duplicated segment. Had only one of the duplicated branches contained thrombus, with ultrasound the femoral vein could have been incorrectly identified as being free of thrombus if only the patent branch had been visualized.
An article published during the course of our study compared balanced-SSFP MRI with contrast venography in the diagnosis of DVT, reporting a sensitivity of 87%, specificity of 98%, and superior pelvic imaging [32]. However, the sample in that study consisted of only 24 patients, only six of whom had positive findings of DVT. A retrospective study [22] of balanced-SSFP MRI in the diagnosis of central vein thrombosis showed that in one case a thrombus was isointense to the normal surrounding veins. Although the possibility that a thrombus would have identical signal intensity to surrounding blood is real, no cases of isointense thrombus were encountered in our study.
A limitation of our study was that no third imaging technique was used to confirm the greater proximal extent of several thrombi seen with MRI. However, because these thrombi were clear continuations of thrombi found more distally with ultrasound, it is assumed that the findings were true-positive. Without a direct comparison with those of contrast venography, these results cannot conclusively show the superior performance of balanced-SSFP MRI in assessing proximal thrombus extent. However, the clear continuity of these thrombi on the MR images is supporting evidence that balanced-SSFP MRI is at least equal to ultrasound in this regard, and likely superior. Another limitation was that some authors [8, 10] still consider contrast venography the reference standard in the diagnosis of DVT. However, ultrasound has comparable diagnostic accuracy to contrast venography in the popliteal and femoral veins examined in our study and is therefore reliable as an alternative reference standard [5]. Furthermore, ultrasound has become the standard examination for DVT assessment and is more accessible as a reference standard.
A limitation of balanced-SSFP MRI is that it is prone to magnetic susceptibility artifact. However, three patients in our study had orthopedic hardware, and all of them received diagnostic-quality balanced-SSFP MRI examinations (Fig. 5). A limitation of the study was that ultrasound had 100% interobserver agreement in all categories of DVT diagnosis but that discrepancies occurred between MRI readers regarding the femoral and common femoral veins. These differences, however, arose entirely from the case of one patient with two distinct foci of thrombi. Both MRI readers identified a common iliac thrombus, but one reader did not identify thrombus in the femoral and common femoral veins. This limitation may be related to satisfaction of search, and because both readers did diagnose DVT in this patient, this discrepancy would not have affected treatment.
We believe that balanced-SSFP MR venography holds promise as a highly accurate alternative in the diagnosis of DVT. It should be particularly effective in situations in which ultrasound cannot be performed or the findings are nondiagnostic. Results of future studies evaluating the performance of this technique in these situations would be of interest, as would an analysis of its cost-effectiveness relative to ultrasound.
Address correspondence to C. Lindquist ([email protected]).
| References | |
|---|
