The diagnosis of SVMs was made on the basis of an enlarged and tortuous perimedullary vein together with abnormal arteriovenous communication in the spinal canal[7, 9, 10]. For the former, we can use conventional T2 MR imaging for detection, while for the latter, we need angiography (MRA, CTA or DSA) to diagnose. In order to facilitate clinical diagnosis and treatment for SVMs, lesion type was based on the locations of the fistulas such as spinal dural arteriovenous fistulas (SDAVF) (Fig. 1), spinal cord perimedullary arteriovenous fistula (PMAVF) (Fig. 2) and spinal cord arteriovenous malformations (SCAVM). AVF of spinal cord is the most common spinal vascular malformation. In this study, it accounted for 94.1%. SDAVF commonly involve the thoracolumbar spine, and shunt lesions located along a nerve root sleeve within a neural foramen and venous congestion of the spinal cord  .The shunt of PMAVF is located in the epidural space and drains into epidural veins. Microsurgery and transarterial embolization are the mainstays of treatment for SVMs, but the associated success rates depend upon the disease type and blood flow status. Stereotactic radiosurgery and fractionated radiotherapy may have the relative safety and effectiveness which can bring significant promise for rare and complex lesions, but the optimal treatment parameters are required. So it is very important to find feeding arteries and fistulas accurately.
DSA is the diagnostic gold standard for SVMs. It can accurately dynastic observe fistula, feeding arteries and draining vein, but it is invasive, high radiation doses, large volumes of iodinated contrast material, complex high technical requirement, and may occur paraplegia and other serious complications. The morbidity of 0.03% and morality of 0.06% have been reported for patients undergoing diagnostic cerebral angiography[17, 18]. DCA-MRA is sensitive, no radiation exposure, shorter scaning time, convenient operation, can be repeated to operate and be able to observe the change about spinal cord etc advantages. So it can be widely applied to clinical diagnosis. The main advantages of DCA-MRA are good image quality, high temporal resolution, and can get three-phase images(arterial, venous and delay phase)(Fig. 3) to observe small blood vessels and the difference between spinal cord artery and vein.
Noninvasive imaging modalities such as DCA-MRA reliably detect SVMs and may predict the level of their location[6, 8].In 1995,Bowen described that spinal CE-MRA can provide valuable information for SDAVF, but they were not use the dynamic phase to classification of the vascular malformations. With technical improvements in 3.0TMR, Spinal MRI using 3D sequences (CE-3D fast spoiled gradient echo, Dynamic CE‑3D MRA) can provide more information than conventional 2D sequences in the evaluation of spinal. Such as the use of parallel imaging, compressed sensing or other parallel imaging techniques at 3T, DCA-MRA have allowed to get a bigger FOV of scan and greatly accelerate a high temporal and spatial resolution, which can help to better locate feeding arteries and fistula as well as visualization of arterial and venous separation, thus allowing improved classification of vascular malformations. For avoid the side effects of renal fibrosis, the dose of contrast is only need 20ml on 3.0T which less the reported in the literature of 25-45ml on 1.5T MR [4, 21].
DSA examination demands high technology. In case 2 and 15, DCA-MRA showed feeding artery and fistula clearly, while DSA was negative. The reason may be that the pressure of the blood supply artery and the drainage vein resist each other, so that the iodine contrast agent cannot reach the drainage vein of the fistula. Although DSA is the gold standard, but sometime the results were not entirely reliable. Due to iodine contrast agent allergy(Case 12),some patients only underwent DCA-MRA to find the feeding arteries and fistulas before surgery. The diagnosis of DCA-MRA was proved to be correct after operation and follow-up. Case 16༌DCA-MRA mistake to predict the localize of fistula. The possible reason maybe the dilated feeding artery has bigger turning angle on the level of T12, leading to the discontinuous DCA-MRA appearance.
It is often difficult to locate the fistula of PMAVF, in our research 4 cases did not inconsistent with DSA. Based on the characteristics of angioarchitecture, PMAVF is the direct communication between root pulp artery and perimedullary vein, the diameter is relatively smaller. For example, the feeding artery of case 1can be clearly determined, but the fistula needs to be carefully searched. According to the analysis of multiple cases and the comparison of surgical results, gadolinium contrast agent concentration is generally considered to be the location of the fistula(Fig. 2). The reason may be the pressure difference between the supplying artery and the draining vein on both sides of the fistula. In case 6༈Fig. 1༉,DCA-MRA results was consistent with DSA and surgery, the patient's clinical symptoms was not completely alleviated after surgery. DCA-MRA found abnormal vascular still exist after 3 months of surgery, but the blood supply artery was not found. The feeding artery from T10 may be the Adamkiewicz (AKA)in case 11. We seem to found the level of the fistula about case 20 (Fig. 3) and 21, but did not find the feeding artery by DCA-MRA and DSA also was negative. This case may classified it as PMAVF.
In recent years, there are some new MR sequences for the diagnosis of spinal vascular malformations, such as volumetric T2 MRI and first-pass contrast-enhanced MRA, which can improve the sensitivity of finding SVMs or help the accuracy of DCA-MRA scanning. But in our experience, with the guidance of convention T2 and the support of test bolus technology, the imaging success rate of DCA-MRA is very high, which can resolve the evaluation of SVMs,identifying the location of the feeding artery to guide DSA and assist in treatment planning. The MRA technique used in the study is three-phase MRA technique and 43 seconds for one phase༌but it can provide more temporal information for better image post-processing.