By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
MR angiography (MRA) of the peripheral vasculature requires the acquisition of an extended field of view (FOV), which exceeds the imaging volume of typical whole-body MR systems in the longitudinal direction. The commonly used method for the coverage of an extended FOV is the multistation approach, for which data are acquired in smaller sub-FOVs at several stations. For each station, the patient table is moved to position the next sub-FOV in the isocenter of the magnet prior to data acquisition. For optimal scan efficiency and volumetric coverage, each sub-FOV is typically as large as possible with respect to magnetic field homogeneity and gradient linearity (1–9).
However, the need to assign a sufficient overlap of the sub-FOVs to obtain a seamless undistorted volumetric coverage and the need to apply dummy scans to maintain the steady state of the MR signal reduce the efficiency of multistation techniques (10). Furthermore, discontinuities at the intersections of stations represent an additional limitation. In the case of contrast-enhanced (CE) MRA, the need to synchronize data acquisition with the arterial passage of the contrast agent bolus introduces additional constraints on total scan time per station, which can result in limited spatial resolution and/or volumetric coverage. In some cases, additional contrast agent injection may be required for some stations (11, 12).
An alternative method for covering an extended FOV, which has already been used for CE MRA of the peripheral vasculature (10, 13–20), is based on a continuously moving table (CMT) during data acquisition (21–25). CMT acquisitions provide a seamless coverage of the extended FOV without the need for additional time for repositioning, dummy scans, or station overlap. In the case of CE MRA, data acquisition time and patient table speed have to match the intra-arterial position of the contrast bolus. To optimize anatomic coverage and scan efficiency, CE MRA data are typically acquired in a coronal orientation, providing an excellent overview of the whole body vasculature, particularly after maximum intensity projection (MIP) processing. However, evaluation of vascular pathologies such as stenosis is typically performed by inspecting the vessel lumen normal to the vessel orientation, i.e., in axial image orientation for most vascular systems. In this context, coronal acquisitions may provide only limited spatial resolution in reformatted axial orientations, thereby limiting their diagnostic value. Furthermore, there is a current association of gadolinium-based contrast agents with nephrogenic systemic fibrosis for patients with decreased renal function (26, 27). Thus, non-contrast-enhanced (NCE) MRA techniques have recently received increased attention.
Suggested NCE techniques for imaging of the peripheral vasculature include electrocardiographically (ECG)-gated three-dimensional partial-Fourier turbo spin echo sequences and axial two-dimensional (2D) time-of-flight (TOF) acquisitions (28). TOF is regarded as the most common NCE technique and permits imaging in axial slice orientation without time constraints imposed by contrast agent administration.
In TOF MRA, the MR signal of static tissue is suppressed by short pulse repetition times (TR) combined with high flip angles. The blood flow carries unsaturated spins from outside into the imaging slice, resulting in high blood tissue contrast by inflow enhancement. Blood signal and contrast depend on TR, blood flow velocity, flip angle, the T1 of blood, slice thickness, and direction of the blood flow with respect to the orientation of the imaging slice (28). Selective saturation of the MR signal in slices parallel to the imaging slice can be used to selectively saturate venous or arterial inflow, thereby providing angiographic images of the arterial or venous system, respectively.
In the present study, the feasibility of axial 2D TOF venography in combination with the CMT approach was investigated to seamlessly image the venous system in the lower limbs. For MR venography, the suppression of the bright fat signal is essential to provide artifact-free angiograms. However, the continuous change of the imaging volume during CMT acquisitions results in changing field inhomogeneities. Since a dynamically updated shim procedure was not available on our MR system, typical fat saturation pulses based on the chemical shift effect could not provide reliable fat suppression for the entire peripheral anatomy. As an alternative approach, a subtraction technique was employed. Two different acquisition strategies for performing a subtraction CMT TOF venography are investigated: a simple two-step approach (29) and an advanced one-step approach providing a reduced sensitivity to patient motion and a reduced scan time (30).
Resulting image quality of the CMT TOF approaches was qualitatively and quantitatively evaluated in a study with 10 healthy volunteers and compared to a standard stationary 2D TOF multistation acquisition.
From a clinical point of view, the presented CMT TOF techniques might be of interest to detect pathologies in the deep and superficial veins of the lower limb. Although conventional phlebography is still considered the gold standard, ultrasound is generally applied in this respect because it does not rely on radiation and contrast agent injection. To illustrate the feasibility of CMT TOF imaging for the assessment of the venous status in patients, three patients suffering from postthrombotic changes and varicosities were included in the study. Ultrasound served as a standard of reference, which is the current clinical standard at our institution.
MATERIALS AND METHODS
To achieve stable fat saturation in combination with CMT MRI, each axial slice was acquired twice, once with saturated arterial blood signal (referred to in the following section as image set A) and once with saturated arterial and venous blood signal (image set B). Consequently, voxel-by-voxel subtraction of two corresponding magnitude images for each slice position resulted in a difference data set containing only venous blood signal and, ideally, almost no signal from stationary tissue. Two subtraction CMT TOF methods, a two-step and a one-step technique, were developed and applied as schematically illustrated in Figs. 1 and 2.
Two-Step Method for Subtraction CMT TOF Venography
An axial 2D single-slice CMT acquisition of the whole extended FOV was performed twice in two separate scans and table sweeps (Fig. 1a). During the first acquisition, a 50 mm-wide saturation slice was placed cranial to the axial imaging slice, with a 10 mm gap to saturate signal from arterial blood (image set A). Next, the patient table was repositioned and the measurement of the extended FOV was repeated using the identical sequence parameters but with two symmetrically arranged saturation slices to cancel the signal from both venous and arterial blood (image set B). Subtraction of two images of sets A and B yielded an image of the venous structure.
Slice-by-slice image acquisition was performed in the opposite direction to the venous blood flow in the patient coordinate system, as illustrated in Fig. 1b.
One-Step Method for Subtraction CMT TOF Venography
The one-step method was designed to allow for simultaneous acquisition of image sets A and B during a single table sweep by arranging two saturation slices around two imaging slices (Fig. 2a). Both imaging slices were acquired in an interleaved manner and symmetrically arranged around the isocenter.
The interslice gap included one saturation slice with a sufficient distance to the two adjacent imaging slices and was minimized to avoid distortions in the image sets and errors in the subtracted images. Two parameters determined the interslice gap: the distance between saturation and imaging slice and the thickness of the saturation slice. To evaluate the optimal arrangement of imaging and saturation slices, image quality in terms of artifact level was visually inspected while the gap between saturation and imaging slices was reduced. The minimal distance between saturation and imaging slices that provided artifact-free images was found at 9 mm. Thickness optimization resulted in a minimal width of 10 mm for saturation slice thickness, which provided sufficient cancellation of inflowing blood for the presented acquisition pattern. Since the imaging slice thickness was 4 mm, the evaluated minimal interslice gap (= (9 + 10 + 9) mm = 28 mm) corresponded to the 7-fold imaging slice thickness.
The saturation slice located between the two imaging slices cancelled the arterial blood signal in the caudal imaging slice (image set A) and the venous blood signal in the cranial imaging slice (image set B). The second, cranial, saturation slice affected only image set B and cancelled arterial inflow. Thus, image set B is free of any vessel signal (Fig. 2a) and the sets A and B provide the same contrast properties as in the two-step approach. Subtraction of two images of sets A and B, which are assigned to the same slice position in the patient coordinate system, yielded an image of the venous structure.
Since the subject was continuously moved through the scanner during data acquisition, the temporal gap between the acquisitions of two spatially corresponding slices of image sets A and B was equal to the time needed for the acquisition of seven imaging slice pairs (Fig. 2b). The first and the last seven acquired slices did not contribute to the final difference data set since images of either set A or set B are missing for subtraction.
For all experiments, 2D axial scanning was performed on a 1.5T whole-body system (Magnetom Avanto; Siemens Medical Solutions, Erlangen, Germany) using a radiofrequency spoiled gradient echo sequence. MR signal was received by a peripheral phased-array six-element coil and a spine coil, which was integrated in the patient table. The motion of the imaging volume through the scanner was taken into account by adapting the frequency of the excitation radiofrequency pulses according to the patient table velocity (25). Slice thickness was 4 mm for all sequences. In the case of the two-step method, a minimal TR of 18.1 ms could be achieved, which had to be increased to 25 ms for the one-step method due to the interleaved data acquisition pattern. Further sequence parameters for the two-step (one-step) method to acquire 128 (256) axial slices were echo time = 3.95 ms, bandwidth = 180 Hz/pixel, generalized autocalibrating partially parallel acquisition (GRAPPA) acceleration factor R = 2, partial Fourier factor = 6/8, image matrix = 320 × 208 (320 × 181), FOV = 400 × 260 mm2, voxel size = 1.3 × 1.3 × 4 mm3 (1.3 × 1.4 × 4 mm3), and total coverage = 51.2 cm (45.6 cm).
These parameters were used for all CMT TOF volunteer and patient measurements. The sole exceptions were the CMT TOF and the stationary multistation data acquired for the quantitative and qualitative analysis, where the TR and the image resolution had to be set to 25 ms and 1.3 × 1.4 × 4 mm3 (image matrix = 320 × 181, FOV = 400 × 260 mm2) for sake of comparability.
Since both CMT TOF methods presented rely on difference data, the maximal blood-to-tissue contrast can be obtained by maximizing the venous blood signal in image set A. For a fixed combination of TR and slice thickness, this is, in theory, fulfilled by a single flip angle for a certain vessel segment. To optimize the flip angle to yield the best venous signal, a stack of 10 axial slices was acquired repeatedly with varying flip angles during table motion and with arterial saturation in one healthy volunteer. Flip angles from 5° to 90° in 5° increments were employed to measure the signal in the femoral vein in one representative axial slice for each angle.
Assessment of Image Quality of the Subtraction CMT TOF Methods
In 10 healthy volunteers, imaging of the venous system between the proximal thigh and the distal calf was conducted using the two-step and one-step CMT approaches. All studies were approved by the ethics committee of our institution, and informed consent was obtained from all subjects.
Voxel-by-voxel subtraction of the two image sets and the application of a MIP resulted in a venogram for both techniques, which consists of 128 axial slices in the case of the two-step method and 114 slices the in case of the one-step method, respectively.
Two radiologists assessed the diagnostic quality of the difference data sets of both CMT TOF methods, blinded to the results of the other reader and the applied method. Different parts of the venous system were graded on a four-point scale, as follows: 0, not visible; 1, partially visible or impaired by major artifacts; 2, largely visible or impaired by minor artifacts; 3, fully visible without artifacts. The two CMT TOF methods were presented in a random order. Paired Wilcoxon's signed rank test was applied to compare the quality scores.
Primary analysis was conducted to compare difference data sets of both methods on a per-observer basis, with the vessels being combined to represent the proximal veins (great saphenous vein, superficial femoral vein, popliteal vein) and the distal veins (anterior and posterior tibial veins and fibular veins). In the case of subgroup analysis on a single-vessel basis or for direct comparison of different vessels, the significance values were adjusted by the Holm-Bonferroni method. Veins from both legs were considered as independent units of analysis. Results of both observers were pooled for subgroup analysis.
Quantitative Analysis of the Subtraction CMT TOF Methods and Comparison to Stationary Measurements
Two different analyses were performed for quantitative comparison of the two CMT TOF methods, one intraindividual region-of-interest (ROI) analysis evaluated in difference data sets and the other one was evaluated in image set A.
First, the signal-to-noise ratio (SNR) and the contrast-to-noise ratio (CNR) between vein and muscle (CNRmuscle), as well as between vein and surrounding fat tissue (CNRfat), were calculated for the difference data sets to compare the two CMT TOF methods between each other.
Second, SNR and CNRmuscle in image set A of both CMT TOF methods were compared to a stationary 2D TOF multistation acquisition with arterial blood saturation.
For all quantitative assessments, ROIs were placed on three reference slice positions at the level of the middle thigh, the knee, and the middle calf (Fig. 3, slices 1-3) to encompass the superficial femoral vein (on slice 1), the popliteal vein (on slice 2), one branch each of the paired anterior and posterior tibial veins, and the fibular veins (all on slice 3). The great saphenous vein was analyzed on all three slices.
The paired Student's t test was applied to determine whether differences between SNR, CNRmuscle, and CNRfat values were statistically significant. Throughout the study, P values <0.05 were considered to indicate a statistically significant difference.
The one-step method for CMT TOF venography was performed in three patients with postthrombotic changes and varicose veins to determine whether clinically relevant features of venous disease can be detected by the proposed CMT TOF method. Relevant clinical information for these patients is presented in Table 1. CMT TOF venography was compared to ultrasound, which is the current clinical standard at our institution. Two radiologists, who were blinded to the clinical data and ultrasound results, classified the varicosis discernible on the CMT TOF data according to Hach (31).
Table 1. Relevant Clinical Information for the Three Patients Enrolled in the Study Provided by Ultrasound. Hach Classification of the Severity of Detected Varicosities Based on Ultrasound and on CMT TOF Imaging
Severity of varicosities based on ultrasound
Severity of varicosities based on CMT TOF
Patient 1, male, 62 years
Deep venous system
Postthrombotic syndrome on both sides
Respiratory-dependent flow on both sides
Postthrombotic residues below the distal popliteal vein on the right side
Postthrombotic residues below the pelvis popliteal vein on the left side
Superficial venous system
Incomplete varicosis of the right great saphenous vein
From distal upper leg to the ankle
From distal upper leg to the ankle
Varicosis of the left great saphenous vein
Varicosis of the small saphenous vein on both sides
Patient 2, female, 60 years
Venous reflux in the right small saphenous vein
Patient 3, female, 58 years
Varicosis of the left great saphenous vein
The flip angle variation for parameter optimization resulted in an optimum distribution for venous blood signal, with a variation of signal intensity between 148.1 and 750.0 arbitrary units. The mean signal intensities and the respective standard deviations indicate a plateau between 35° and 60° (Fig. 4). According to this observation, a flip angle of 45° was selected for all subsequent experiments.
Coronal MIP images from the difference data sets acquired with the two-step and one-step method in one volunteer demonstrate the feasibility of CMT TOF venography (Fig. 5). For this preliminary in vivo experiment, the shortest possible TR was chosen, resulting in a table velocity of 3 mm/sec and a total acquisition time (TA) = 2 × 3:17 min = 6:34 min for the two-step method and 2 mm/sec table velocity and TA = 4:34 min for the one-step method, respectively.
The resulting images exhibit good suppression of the surrounding stationary tissues, including the subcutaneous fat, thus leading to good delineation of the superficial and deep venous structures.
Independent and blinded assessment of diagnostic image quality in 10 healthy volunteers was conducted by two radiologists to further investigate to what extent each technique is capable of depicting different parts of the venous system. Mean scores for proximal and distal veins are presented in Table 2. For both observers, the difference between both CMT TOF methods was statistically significant for the distal vessels in favor of the one-step method. For the proximal veins, the quality score was only significantly different for observer 2. The score distribution of the image grading for both CMT TOF techniques and vessel groups is summarized in Fig. 6. On a per-vein subgroup analysis, the quality scores of the one-step method were again superior to those of the two-step method for all three distal veins (anterior and posterior tibial veins and fibular veins). When the scores achieved by the one-step method of the distal veins were compared among each other, significantly lower scores were found for the posterior tibial veins compared to anterior tibial veins and fibular veins. No significant intervein differences were found for the two-step algorithm. Overall assessment of interobserver agreement showed a weighted κ of 0.71 (standard error, 0.037), reflecting good agreement.
Table 2. Mean Scores of the Image Grading for Both Observers*
Image quality of difference data sets of both CMT TOF methods was assessed on a four-point scale.
indicates significantly superior results for the one-step method.
Table 3 lists the average SNR, CNRmuscle, and CNRfat values evaluated in the difference data sets of both CMT TOF techniques. Statistical analysis demonstrated no statistically significant differences in terms of SNR, CNRmuscle, and CNRfat values (P > 0.05) for the comparison of the two CMT TOF techniques. Similar values for CNRmuscle and CNRfat in the difference data confirmed excellent performance of fat suppression via subtraction.
Table 3. Comparison of Average SNR, CNRmuscle, and CNRfat Values in Difference Data Sets in 10 Volunteers, Obtained With the Two Presented Methods for CMT TOF Venography
Great saphenous vein, slice 1
Great saphenous vein, slice 2
Great saphenous vein, slice 3
Anterior tibial vein
Posterior tibial vein
When a conventional stationary measurement and the image set A of both CMT TOF methods are compared, no statistical differences could be observed for most of the investigated veins in terms of SNR and CNRmuscle values (P > 0.05). This is illustrated by Fig. 7, which shows average values of SNR and CNRmuscle for all evaluated leg veins. Only the SNR evaluated in the great saphenous vein (slice 1) and the CNRmuscle evaluated in the politeal vein demonstrated significantly superior results for the stationary measurements compared to one of the CMT TOF acquisitions.
The application of the one-step method for CMT TOF venography in patient measurements clearly permitted the detection of pathologic changes in the deeper and in the superficial venous system. The scan time for all presented patient data was TA = 4:34 min.
Coronal and sagittal views of the venogram allow for the evaluation of the deep venous system in patient 1 (Fig. 8) with CMT TOF imaging. Only the veins in the proximal part of the patient's right leg could be depicted. Due to the flow dependency of TOF MRA, these observations were associated with low flow to the point of no flow in the distal vessels, which is in good agreement with the postthrombotic residues found by ultrasound. Furthermore, in the proximal right femoral vein signal variations across the slice position were clearly visible in the MR venogram, reflecting varying blood flow velocity over time. This observation can therefore be correlated with the respiratory-dependent flow conditions revealed by ultrasound.
In the superficial venous system, CMT TOF imaging detected the dilatation of the great saphenous vein on both sides, with a more severe degree in the patient's left leg. Both radiologists graded the severity of varicosis on the patient's left side as grade IV according to Hach (31) and in good agreement with ultrasound findings, both listed in Table 1. The vessel dilatation ranged from the proximal thigh to the foot due to valve failure at the saphenofemoral junction. The two radiologists assessed the severity of incomplete varicosis of the right great saphenous vein, ranging from the distal upper leg to the ankle. As additional information, in comparison to the ultrasound the radiologists could classify the incomplete varicosis as perforator type (Table 1). Due to severe dilatation and probably increased flow, one typically small perforating vein connecting the superficial and deep venous system becomes visible in the MR venogram and marks the proximal point of insufficiency. Furthermore, CMT TOF detected the dilatation of the posterior arch veins on both sides, which is clearly visible on the sagittal view in Fig. 8.
The MIP of patient 2, which was processed based on a freehand user-defined ROI in CMT TOF data (Fig. 9a), showed varicose changes in terms of a curved small saphenous vein, representing the insufficiency of this superficial vein in accordance with findings by ultrasound (Table 1).
Side-by-side comparison of the vessel diameter in the coronal MIP (Fig. 9b) revealed pathologic changes of the left great saphenous vein of patient 3, as demonstrated by ultrasound. The dilatation of the vessel extended to the foot and therefore suggests a more severe finding with a higher grade according to Hach (31) (grade IV) than detected in the ultrasound investigation (grade II), which was performed 3 months prior to CMT TOF imaging (Table 1).
DISCUSSION AND CONCLUSIONS
Two different imaging strategies for TOF vein imaging with extended FOVs during continuous patient table motion have been presented. Both techniques are based on a subtraction method requiring the acquisition of two image data sets, one with cancelled arterial blood signal and one with cancelled venous and arterial blood signal. The subtraction of these two sets provides images with well-suppressed signal from static tissue and fat. The deep and the superficial venous system can be well depicted in a reasonable total scan time.
However, perfect suppression of background signal in subtracted images is only possible if completely artifact-free images can be acquired for each slice position. Even artifacts with only small signal intensity can become visible in subtracted and processed MIP images if they manifest differently for image set A and B and for different slice positions. A potential reason for these artifacts might be the application of saturation pulses, which was performed for both CMT TOF methods presented, and might affect the signal evaluation in the steady state in the imaging slices. The result of such variations of artifact level becomes visible as horizontal banding in processed MIP images, as visible in Fig. 8 and Fig. 9. Since the arrangement of imaging and saturation slices has already been optimized in this study, image filtering can further improve image presentation.
The feasibility of CMT TOF techniques for peripheral vein imaging was demonstrated in a study with 10 healthy volunteers. In terms of TOF SNR and CNRmuscle values, no statistical difference could be observed in comparison to stationary measurements. These results reflect the precisely adapted frequency of the excitation radiofrequency pulses to the patient table velocity (25), which freezes the motion in the CMT experiments. Similar values for CNRmuscle and CNRfat in the difference data revealed good suppression of the static background signal, including fat.
In the case of the two-step method, the temporal gap between the acquisitions of an image assigned to set A and an image of the same slice position assigned to set B is equal to the time needed for the acquisition of the whole extended FOV. Therefore, patient motion during data acquisition can affect the image quality of large regions. In contrast, the introduced one-step method considerably reduces this temporal gap to the time needed for the acquisition of seven imaging slice pairs. The improved correspondence of two subtracted images of set A and B for the one-step method is reflected in the image grading results, which demonstrated better visibility of the small distal veins.
Additionally, using the one-step method, total scan time is decreased by approximately 30% due to the interleaved acquisition pattern requiring only a reduced number of radiofrequency pulses for signal saturation of inflowing blood.
The total scan times of both CMT TOF techniques have also to be compared to conventionally fat-saturated stationary TOF acquisitions, for which a minimal TR of 35 ms would be achievable with identical echo time, image resolution, and saturation of inflowing arterial blood. With these parameters, a putative fat-saturated CMT TOF technique would, however, still require an almost identical total scan time compared to the two-step method and a significantly prolonged total scan time compared to the one-step method. Nonetheless, fat-saturated CMT TOF techniques might be a desirable goal because they would not require image subtraction, but stable shim and fat suppression techniques in moving table MRI are not yet available.
The easiest way to reduce scan time would be to omit the step of data subtraction and to only rely on image set A for interpretation. This would result in an approximately 30% shorter scan time compared to the one-step approach. As considerable parts of the venous vasculature are, however, directly embedded in subcutaneous and intermuscular fat, the diagnostic value of these images would not be sufficient. Furthermore, MIP processing would be impeded by ubiquitous subcutaneous fat and additional postprocessing of the data would be required. Even the available fat saturation techniques were not sufficiently stable in this respect.
Despite the general goal to reduce acquisition time, there is no temporal constraint for TOF acquisitions, whereas in the case of CE MRA contrast bolus injection, table speed and data acquisition have to be synchronized. If the acquisition of one station in CE-MRA exceeds or falls below the bolus travel time, the bolus has already left or has not yet entered the investigated region and the visualization of the veins is reduced. Even if those requirements are met, full suppression of arterial signal is not achievable because the contrast agent first passes through the arteries before it enters the venous system, resulting in residual arterial enhancement. Even saline chaser boli cannot be expected to sufficiently reduce arterial contrast while preserving venous enhancement. CE MR venography without arterial contamination can therefore only be achieved by contrast agent injection directly into the extremity of interest, combined with the use of a tourniquet, as shown in (32, 33). Avoiding arterial signal contamination, however, can easily be solved in TOF acquisitions by the application of saturation slices.
The temporal flexibility of TOF acquisitions is also an advantage over ECG-gated three-dimensional partial-Fourier turbo spin echo methods, which have also been proposed for application to the peripheral vasculature (28) and strongly depend on the exact timing of the triggered acquisitions.
In general, techniques relying on ECG data are not compatible with CMT since table motion cannot yet be synchronized with the ECG signal. Phase-contrast MRA would be another alternative for NCE venography, but since phase contrast MRA requires the acquisition of several images per slice position, a potential CMT phase-contrast technique would also require further acquisition acceleration in order to render it feasible for the coverage of extended FOVs.
Thus, this TOF study is the first successful attempt to combine a standard, well-established NCE method with CMT.
Recently, a new NCE method for vessel visualization based on a balanced steady-state free precession Dixon method has been introduced (34). Its feasibility was demonstrated visualizing the peripheral vessels, using a multistation approach. This method still suffers from some low-signal areas in the reconstructed MIP images due to magnetic field inhomogeneities. In contrast, the image quality achievable with the presented CMT TOF methods for venography is less affected by magnetic field inhomogeneity or gradient nonlinearity since data are always acquired near the isocenter of the magnet. The one-step method, however, is somewhat more sensitive to variations of the static magnetic field strength or gradient linearity since images used for subtraction are not acquired at the same physical location in the scanner. Nonsymmetric field deviations with respect to the isocenter can affect slices from data sets A and B differently, thereby introducing errors in the subtracted images. To reduce the influence of nonlinearities in the main magnetic field and the gradients on image quality, the thickness of the applied saturation slices and their distance to the imaged slices were chosen as small as possible. Although the effects of inhomogeneity are even more pronounced at wide-bore systems, the feasibility of the two-step method has already been demonstrated for that scanner type (29).
A limitation of axial 2D TOF imaging is the dependence of the TOF signal on the orientation of the imaging slice relative to the imaged vessel. TOF signal in vessel parts with nonperpendicular orientation is decreased and vessel parts running parallel to the imaging slice suffer from reduced visibility in TOF angiograms. This drawback of the TOF technique was also observed in the volunteer and patient CMT acquisitions performed here. Small sections of curved veins formed in varicosis could therefore not be depicted due to in-plane flow saturation, as would also be the case for stationary TOF acquisitions (Fig. 9a). Nevertheless, the affected vessels themselves could still be identified, despite the signal loss in these tortuous segments.
A further drawback of the presented method is the limited spatial resolution, which may not be sufficient to reliably detect smaller venous branches. Small-diameter veins such as perforating veins, small muscle veins, or sections of the deep system below the midcalf may be difficult to visualize on reconstructed venograms. In addition, low inflow velocities in smaller veins may further reduce SNR and CNR and thus impair the depiction of small vessels. As varicose changes often result in considerably dilatated veins with increased flow, these vessels could clearly be detected, as demonstrated for the perforating vein and the small posterior arch veins in patient 1 and the small saphenous vein in patient 2.
A potential solution for general improvement of vessel visibility could be the transition of the proposed CMT TOF venography from 1.5T to 3T systems, which promises increased image quality, as already observed for intracranial and cervical imaging (35–37). The intrinsically higher SNR of 3T acquisitions in combination with prolonged T1 of tissue has the potential to enhance TOF signal and improve background suppression. The additional vessel signal may be exploited to increase image resolution.
Other possible improvements include automatic coil switching. If receiver coil elements are switched on and off during data acquisition according to the currently acquired body region, the presented CMT acquisition strategies can be extended to a larger FOV and global SNR can be increased (29).
Initial patient measurements indicate the clinical potential of the presented CMT TOF methods. Results were compared to ultrasound, which is generally used for detection of venous pathologies. However, it can be difficult to apply ultrasound measurements to the region of the middle leg, and especially the lower leg (38), for which the feasibility of the CMT TOF methods has been demonstrated. All CMT TOF findings were in good agreement with ultrasound diagnostics. Since data of large parts of the peripheral venous system on both sides are available in the same data set, side-by-side comparison can easily be used to elicit additional medical information, as shown for the varicose great saphenous veins in patients 1 and 3. Dilatated varicose vessels were clearly highlighted, with an improved visualization in comparison to nonpathologic vessels.
In our feasibility study, no patients with acute deep vein thrombosis were available. However, the ability of stationary TOF imaging to detect thrombi in 2D axial images has already been proven in several studies (39, 40) with submillimeter in-plane resolution. As indicated by the successful identification of the postthrombotic residues for patient 1 (Fig. 8), CMT TOF may be promising for the detection of acute deep vein thrombosis, which needs to be evaluated in future studies.
Future developments also include the application of CMT TOF for NCE arteriography of the peripheral vasculature. To achieve a stable TOF signal, the acquisition must be synchronized with the pulsatile arterial flow profile, thus requiring considerable acceleration of image acquisition. Future work will therefore focus on acceleration techniques with partial k-space acquisition, which take advantage of only small deviations in image content in z-direction and the k-space sparsity in difference data. Possible acquisition strategies may be view sharing techniques (41) or compressed sensing (42). Recently, another promising method for parallel imaging, termed z-GRAPPA (43), has been introduced, which uses k-space information from spatially adjacent 2D axial slices.
In summary, the feasibility of axial 2D CMT TOF acquisitions for application to the peripheral veins has been demonstrated. Two different acquisition strategies for CMT TOF provided a stable fat suppression. Using the one-step method, total scan time could be significantly decreased and image quality was less sensitive to patient motion during data acquisition, which is most important for imaging the small distal veins. Preliminary patient data confirm the potential of this CMT TOF approach to successfully detect venous pathologies. The diagnostic potential of the CMT TOF methods, however, has to be further evaluated in future studies.
The authors appreciate the expert assistance of Philipp Franke in evaluating the image quality.