Thrombosis, in its various forms, accounts for more deaths in the western world than any other disease. Major conditions in which thrombosis plays a principal role include myocardial infarction, stroke, peripheral vascular disease, deep vein thrombosis (DVT) and pulmonary embolism (PE). Medical management of these conditions relies on accurate diagnosis not only of the resultant tissue or organ damage but also of the causative vascular lesion. The development of cross-sectional imaging techniques (ultrasound, computed tomography, magnetic resonance imaging [MRI]) in recent years has resulted in huge advances in end organ imaging (brain and heart). Imaging of vascular disease, despite similar advances in sectional imaging techniques in the form of magnetic resonance and
computed tomography angiography, have tended to mimic conventional angiographic techniques by visualizing flowing blood within the vessel lumen, rather than the causative obstructing lesion.
Working on the premise that the acute primary event resulting vascular obstruction is thrombosis, the ability to directly visualize vascular thromboses should provide a means of accurately assessing the culprit lesions, resulting in more accurate diagnosis and improved patient management, along a better understanding of the disease process and providing effective research tool.
MRI is rapidly establishing itself as a first-line investigation many clinical settings. Most recently MR angiography has
made significant advances allowing accurate angiographic images comparable to conventional studies [1]. The lack of ionizing radiation makes MRI an attractive technique in both clinical and research setting. Multiplanar, 3-dimensional image acquisitions provide comprehensive imaging strategies.
Above all the ability of MRI to exploit the differences in tissue make-up and display these as alterations in image contrast provides a unique opportunity to selectively discriminate between different tissues, both normal and pathological. is this unique capability that is exploited in magnetic resonance direct thrombus imaging (MRDTI). It has long been known that as blood undergoes clotting it passes through a number of well defined stages reflecting the state of oxygenation of red blood cells trapped within the clot and these are reflected in alterations in their MRI contrast characteristics [2].
When hemoglobin is removed from the normally high oxygen environment of the circulating blood it undergoes oxidative denaturation to the ferric (Fe3 þ) form. In this form methemoglobin will cause shortening of T1, akin to MRI contrast agents, will therefore result in high signal on a T1-weighted acquisition. This is due to the fact that methemoglobin is paramagnetic, containing 5 unpaired electrons. Potentially therefore this high signal generating derivative could be used an endogenous contrast agent within clotting blood and therefore act a marker for thrombosis. For this to occur a number of important provisos must be met:

• adequate methemoglobin must be generated to produce sufficient signal;
• methemoglobin must be formed reliably within a thrombus;
• methemoglobin must be formed sufficiently quickly to allow identification of acute thrombosis (hours);
• methemoglobin must persist long enough to provide a suitable imaging window of opportunity (days).

Potential pitfalls in the use of methemoglobin as a marker of thrombosis will occur if any of these are not fulfilled. For instance the difference in constituents between arterial (platelet rich) and venous (red blood cell rich) thrombi could produce different imaging characteristics.
In order to detect methemoglobin the MRI sequence parameters used must not only be adjusted to provide a T1-weighted sequence but also remove or depress high signal arising from any other tissue that may obscure the thrombus signal. On a T1- weighted scan this high signal most commonly arises from tissue containing fat. To overcome this potentially obscuring and confusing effect the sequence must employ some form of fat suppression or saturation. While blood on a T1-weighted sequence is of low signal this can be artefactually increased as it
flows into the imaging field. Dark blood can be generated by the application of inversion recovery prepulses to cause blood to generate zero signal. The remainder of the pulse sequence design can be selected on the basis of how quickly the data is to be acquired. Regions of the body that are stationary and undisturbed by physiological motion (i.e. breathing) can be acquired by longer sequences allowing high resolution 3-dimensional sequences. More rapid acquisition may require far shorter sequences acquired slice by slice in a 2-dimensional acquisition thus allowing acquisition in a breath-hold.
Generation of methemoglobin in vitro has demonstrated that there is a linear relationship between the concentration of methemoglobin and T1 shortening [3]. Using a 3-dimensional fat suppressed, nulled blood sequence the first application of MRDTI for the detection of in vivo human deep vein thrombosis (DVT) [4] confirmed that this technique had the potential to detect recent thrombosis in patients with known DVT.
Various environments exist in which thrombus maturationmay occur that potentially can give different imaging appearances.
The formation of methemoglobin will obviously require sufficient hemoglobin, and therefore red blood cells, to allow
detection. The theoretical difference in make-up between arterial and venous thrombus, one being platelet-rich, the other red blood cell-rich, could have significant effects. The ease with which hemoglobin can change from one state to another should also be considered. Formation of methemoglobin is an oxidative process. The availability of local oxygen in all parts of the thrombus could therefore lead to differential contrast generation.
The relationship of the thrombus to the vessel wall is also important. Hemorrhage and thrombus formation within the vessel wall, either due to vessel dissection or complication of atheromatous plaque, will expose the thrombus to a completely different environment than if in the vessel lumen itself.
One species that has particular affinity for hemoglobin, with the rapid formation of methemoglobin, and potentially found within the vessel wall, is nitric oxide. During the process of thrombus organization invasion of the thrombus by cellular constituents of blood (macrophages and white cells) may also alter image appearances. The exact effect of all of these components in vivo on MRI signal generated from thrombus has yet to be elucidated.
Following the promising pilot study in patients with known DVT the utility of this technique in the setting of DVTwas more rigorously tested against the gold standard of ascending contrast venography (ACV) [5]. In 103 patients suspected of suffering DVT the overall sensitivity and specificity for MRDTI was 96% and 90%. Unlike other imaging techniques that struggle to maintain diagnostic accuracy throughout the lower limb venous system, MRDTI had no apparent blind spots, being just as sensitive and specific below the knee, above the knee or in the pelvis. The agreement between two readers for these areas was very good with kappa values varying between 0.89 and 0.98.
Knowing that ACVis in fact a ‘fuzzy’ gold standard, incorporation of results from ultrasound in discordant cases was found to increase the sensitivity and specificity further (98% and 96%).
Two of the major perceived drawbacks of using MRDTI as a first-line investigation of DVTare cost and availability. There is no doubt that with the current provision of MRI scanners in many countries this is problematic. However, as this situation improves a case can be made to include DVT diagnosis as part of the service provided by MRI. The technique appears to be accurate and is an ‘end test’ with few unresolved diagnoses. It can be combined with chest imaging for a comprehensive investigation of thromboembolic disease (TED) (see below).
As there is no clinician input during acquisition making, availability of the test is only dependent upon scanner time.
The images are easily interpreted and could be electronically transported and read from a central facility. The time required for scanning is short (15–20 min) allowing ease of inclusion during busy scanning schedules and maximizing scanner utilization.
We are at present trialing (Trial of OutPatient MRDTI of DVT – TOP MD) an outpatient service using clinical scoring
and D-dimer testing as screening tests prior to MRI which acts as the definitive end test.
For those centers unable to provide this level of service a number of specific applications warrant consideration. Investigation of TED during or just after pregnancy can be difficult with standard techniques. The pelvis is often difficult to examine, but it is in this group that isolated pelvic thrombosis is likely to occur. Radiation and intravenous contrast agents are to be avoided wherever possible. In a small trial using MRDTI in pregnancy, all thromboses were detected that had been diagnosed by alternative diagnostic methods, but in addition the extent of thrombosis was found to be more extensive in five cases, and progression of disease was confidently diagnosed in three patients leading to alteration in management (Fig. 1).
Despite repeated studies in a number of these patients the technique was well tolerated [6].
Because the contrast generation using MRDTI relies on the formation of methemoglobin, the time course of signal appearance/ disappearance is similarly dependent. The accuracy of this technique for detecting DVT suggests that in the venous environment methemoglobin is formed sufficiently rapidly to be detected in the acute setting. The earliest reliable timing of formation in our experience is within 8 h [7]. Just as important as the rapidity of formation is the duration of contrast. Followup of patients with proven DVT has shown that maximum signal is achieved at approximately 3 weeks, after which time the signal intensity plateaus. In the lead up to this point the development of signal within the thrombus is also indicative of its maturity. The high signal is initially seen in the peripery of the thrombus giving a characteristic target sign [8]. This may reflect the interaction of the thrombus with components derived from the vessel wall, as outlined above. As more methemoglobin is generated the high signal is seen to migrate centripetally until the whole thrombus is of high signal. In a vessel the size of the femoral or iliac vein this process takes up to 3 weeks to complete. After 3 weeks the signal will persist up to, but not beyond, 6 months. During this phase the appearances of the thrombus are also characteristic. As the thrombus becomes organized and there is further oxidative denaturation of methemoglobin to hemosiderin, signal is lost. This does not tend to be a uniform process throughout the length of the thrombus but rather occurs at intervals, with the resultant appearance of islands of high signal, which disappear over the ensuing weeks.
Caution however, should be taken in equating disappearance of high signal with recanalization as many of the vessels were found to recanalize incompletely, if at all. The major advantage of being able to image the maturation of thrombus in this manner is that aging of thrombus is possible. Even if this is a crude estimation the ability to discriminate between thrombus which is recent (within the last few weeks) and old (over 6 months) allows the ruling in or ruling out of recurrent DVT, within this time frame, with its resultant management implications.