At the surface of the cord, two arterial systems can be described (Fig. 1):

The spinal cord parenchyma is supplied by the intrinsic arterial system, which is subdivided into a central (centrifugal) system and a peripheral (centripetal or vasocorona) system [13]. The central system is comprised of the central arteries (sulcal or sulcocommisural arteries; diameter 0.06–0.40 mm), which originate from the ASAs and travel into the anterior median fissure, penetrate into the cord either on the left or right, and branch centrifugally, mainly within the gray matter [19]. The peripheral system (vasocorona; diameter 0.1–0.2 mm) [19] consists of small perforators (rami perforantes) that originate from the pial plexus at the periphery of the spinal cord and course into the white matter centripetally.

Although radicular arteries accompany the spinal nerves at each level, the radiculomedullary branches supplying the spinal cord are present only at few levels. The average number of anterior radiculomedullary arteries is 6 (range 2–14) [18, 19], whereas the number of posterior radiculomedullary arteries varies from 11 to 16 [15]. The superior cord segment involving cervical and upper thoracic cord (up to T2/T3) is usually supplied by vertebral arteries and 2–3 anterior radiculomedullary arteries, including the artery of cervical enlargement [20], often a branch of the deep cervical artery, and usually accompanies the C6 nerve root. Similarly 2–3 anterior radiculomedullary arteries, including the artery of Adamkiewicz (AKA), usually supply the inferior cord segment involving the lower thoracic and lumbar cord. The intermediate cord segment involving the mid-thoracic cord (T4–T8) is a poorly vascularized segment with paucity of the radicular arteries, although occasional radicular artery is present at T7 level. In addition to the overlapping areas between the upper, mid, and lower cord segments, the entire mid-thoracic cord due to its paucity of the radiculomedullary arteries is considered as an anatomical “watershed territory” of the spinal cord.

This is the most common type of spinal cord infarction. The common etiological factors in this group are included in two distinct groups: hypoperfusion and embolism. The individual risk factors are discussed in detail below:

Spinal subarachnoid hemorrhage (SAH) accounts for 1 % of all SAH. However, this is the most common form of hemorrhagic spinal cord infarction (Fig. 8). Cervical spinal cord ischemia has been described in association with spinal subarachnoid hemorrhage in patients following dissection of the vertebrobasilar system or ruptured vascular malformations of the spinal cord [77]. Disruption of the blood supply due to the involvement of radiculomedullary arteries is thought to be the underlying mechanism in the acute setting [78]. Micro- and macrocirculatory failure secondary to subarachnoid hemorrhage may result in delayed ischemic deficits. Vasospasm has shown to be the most likely mechanism causing secondary injury in experimental animal models of spinal cord injury [79].

Although different lesions predispose to hemorrhage in different compartments (Table 2), their strict localization is not possible. The hemorrhage tends to extend across multiple compartments depending on the amount and site of hemorrhage and may result in subdural and epidural hematoma in addition to the subarachnoid hemorrhage. Prolonged mass effect from the extra-axial hematoma in addition to vascular disruption related to involvement of the radiculomedullary feeders, circulatory failure, and vasospasm might result in spinal cord ischemia in these scenarios [78]. Conditions such as anticoagulation and bleeding diathesis may cause hemorrhage in either compartments.

This less common etiology is usually associated with spinal dural arteriovenous malformations (Fig. 9) secondary to venous congestion [80] and is discussed in multiple subsequent sections of this chapter. It may also be a complication of deep sea diving, in which gaseous occlusion of the spinal venous plexus has been postulated as a mechanism of injury in spinal decompression myelopathy [9]. Early recompression reverses symptoms in about three-fourths of patients [81, 82]. Kim et al. [83] stated that the hemorrhagic, nonhemorrhagic, and embolic venous infarction could not be differentiated with certainty on clinical grounds. Hemorrhagic venous infarctions were associated with marked pain at the time of onset and progressed rapidly; survival time was relatively short. Nonhemorrhagic infarctions were less often painful and evolved slowly. Survival time was longer than in hemorrhagic infarction. Embolic venous infarction was associated with pain of abrupt onset. As the initial presentation does not correlate with severity of the disease, treatment with hyperbaric chamber should be initiated promptly in decompression sickness, even in those with mild symptoms [82]. Also, the potential association of malignancy and venous spinal cord infarction is well recognized, particularly in the case of pancreatic neoplasm [30].

Fehlings proposed a two-step process involving primary and secondary mechanisms in acute spinal cord injury [85]. Allen first postulated the concept of a secondary mechanism in 1911 based on canine studies [86, 87].

Spinal cord veins are valveless. However, an anti-reflux mechanism has been described at the dural exit of radicular veins. It consists of an increased smooth muscle fibers and a Z-shaped course of the radicular vein when it crosses the dura to drain spinal cord venous blood in the epidural veins [14]. In case of spinal dural arteriovenous fistulas, the arterialized vein drains the fistula blood flow toward the spinal cord, and the increased pressure is transmitted to the intrinsic cord veins in retrograde fashion. This results in decreased intramedullary arteriovenous gradient and hypoxia, loss of autoregulation, and disruption of blood-cord barrier and, if left untreated, results in subacute necrotic myelopathy.

Ischemia in the distribution of the anterior spinal artery rapidly initiates vasogenic and cytotoxic edema in the gray matter with subsequent increase in mobile water. This is manifested by a focal increase in proton density and a prolongation of the T2 relaxation time, which appears as distinct increase of signal intensity on the long TR pulse sequences.

The ischemia/signal abnormalities start in the anterior horns of the gray matter and, with increasing severity, spread posteriorly to involve the posterior horns. Ultimately, the ischemic changes and corresponding MR signal abnormalities extend laterally to the posterolateral funiculi of the spinal cord, including the crossed corticospinal tracts. In severe cases, the whole cross section of the spinal cord is infarcted. There seems to be a correlation between the distribution of the MR signal abnormalities and the severity of the clinical symptomatology seen in spinal cord ischemia [46] (Table 3).

Based on international standards for neurological and functional classification of spinal cord injury proposed by the American Spinal Injury Association (ASIA) and International Medical Society of Paraplegia (IMSOP), ASCIS can be categorized into four groups [137]. The clinical syndromes are:

Hemorrhagic infarction presents with sudden onset of pain, usually at the level of hemorrhage. Meningeal irritation is a typical finding. Patients may have associated headache, cranial neuropathies, and decreased level of consciousness depending on extent and distribution of the hemorrhage. The evaluation for spinal SAH typically follows negative intracranial workup.

This is the most common presentation in SDAVF. Patients present with progressive lower extremity weakness (>50 %), sensory deficits (30 %) including paresthesia, sensory loss and hyperesthesia, gait disturbance, back pain (24 %), bowel and bladder disturbance, and impotence. Once motor or sensory deficits appear, severe disability ensues over 6 months to 2 years.

Low signal-to-noise ratio (SNR) and prolonged imaging times while scanning the entire spine are the main challenges of the spine imaging. Introduction of the phased-array coil technology [141] and increased field strength has increased the baseline SNR, a major improvement in spine imaging, which has inherently limited SNR. Similarly introduction of the parallel imaging [142, 143] and improvement of multielement radiofrequency (RF) coils have not only enhanced the SNR of high-resolution MR imaging over a large craniocaudal extent but also dramatically shortened the imaging times.

Regular structural MR imaging sequences for diagnosis of the cord infarction include multiplanar T1- and T2-WI fast spin-echo (FSE), gradient-echo (GRE), and contrast-enhanced (CE) acquisitions. Diffusion-weighted imaging (DWI) has increasingly shown as a promising technique in the evaluation of the acute spinal cord syndrome [144, 145] and used as a part of the standard protocol in the evaluation of acute non-compressive myelopathy.

The sensitivity of conventional MRI is limited, particularly in the first several hours. The reported sensitivity of T2-WI cord signal abnormality in SCI ranges between 45 % and 73 % depending in part on timing of the scan [1, 9, 10, 146, 147]. Hence the initial scan may either confirm the diagnosis or be a normal study and serve as a baseline for future follow-up imaging. Isolated finding of T2 signal abnormality, even with restricted diffusion, is not specific for ischemia and can be seen in transverse myelitis and other intrinsic cord pathologies. These findings, confined to a vascular territory, are more specific for SCI.

The typical MRI findings in spinal cord infarction include isolated pencillike area of T2 hyperintensity involving the centromedullary region with sparing of the anterior rim, often encompassing more than two vertebral segments on sagittal images [2, 156], and bilateral hyperintensities that are mostly confined to the anterior horn area, leading to the typical “snake eyes or owl’s eyes” configuration on the axial acquisitions [156, 157]. Although the T2 hyperintensities are seen in 90 % of patients with SCI, these are nonspecific [156]. Further in the acute phase, only 50 % of the patients show a demarcation of T2 hyperintensity of the spine during the first 24 h which limits the sensitivity of this finding [80].

A finding of restricted diffusion on diffusion-weighted images (DWI-MRI) reported to be significantly more sensitive than standard T2 images, appearing as early as 3–8 h after the ictus, but the limited evidence precludes a numerical estimate of its sensitivity [144, 151, 152, 158]. Studies have shown that restricted diffusion can be seen in SCI as early as 3–8 h following an ictus [152–154]. Moderate swelling of the cord is noted in the acute phase with onset of T1-WI hypointensity in the cord from days 5–7 [80]. Contrast enhancement is noted in the subacute stage, typically after 5 days, and persists for up to 3 weeks after onset [159]. Similar to the cerebral edema, venous hypertension from spinal DAVF without cord infarction gives pure extracellular edema, thus increasing T2-WI signal intensity but with normal findings on DWI.

A finding of vertebral body infarction on MRI is a useful indicator of vascular nature of the acute spinal cord syndrome [160–163]. However, this finding is not specific, seen in only 4–35 % of patients with spinal cord infarction [1, 146, 164]. SCI may not always be accompanied by vertebral body infarction, because arterial occlusion may be located distal to those vessels supplying the vertebral body or because of the good collateral blood supply of the vertebra [160]. Cheng et al. [164] showed that concomitant spinal cord and vertebral body infarction is highly associated with aortic pathology, which highlights the common blood supply segmental arteries supplying the metamere and radiculomedullary branches of the radicular artery (branch of spinal trunk of the segmental artery) feeding the spinal cord [13].

In a patient with suspected SCI, the MRI signal abnormalities alone are nonspecific. Care should be taken to interpret these findings in conjunction with history, clinical findings, and CSF analysis. Although CSF analysis in spinal cord infarction is typically normal, but there can be pleocytosis (rarely more than 100 WBC) and an elevated protein (usually less than 119 mg/dl) [7, 9, 10, 166].

Advanced techniques such as diffusion tensor imaging (DTI) and fiber tractography (FT) are increasingly investigated in the cord infarction. These techniques have technical challenges limiting the image quality, related to the highly magnetically inhomogeneous material surrounding the spinal canal, small size of the spinal structures, long craniocaudal extent of the spinal canal, pulsation effects of cerebrospinal fluid (CSF) and blood, and motion degradation from respiration, swallowing, and patient movement. In patients with cervical cord injury, fractional anisotropy (FA) of DTI and FA showed better correlation with motor function and clinical findings such as neurological impairments compared to the conventional MRI [167]. DTI imaging of the spinal cord is currently only a research tool, but preliminary studies have shown that it holds considerable promise in predicting the severity of spinal cord injury [168].

Unlike cerebral infarct, onset of T2 hyperintense signal abnormality in the spinal cord infarction is variable from 8 to 72 h, and studies reported normal MRI findings at 48 h following ictus [157, 169]. High index of clinical suspicion warrants follow-up imaging. However, up to 14 % of patients with suspected SCI have normal follow-up MRI scans [10, 166].

Repeated or follow-up DWI may also help in differentiating diagnosis. There seems to be more sensibility on DW images than on T2-WI images during signal changes within the course of disease, and DWI may be useful in monitoring the progress of the disease [155]. Serial MR studies with contrast showed sequential changes of enhancement similar to those seen in cerebral infarcts [169]. Understanding of characteristic MR findings and evolutional changes at different stages of SCIs and follow-up using serial MRI with contrast are vital in the early diagnosis and more effective in the management of the patients.

Multiple sclerosis is a chronic inflammatory demyelinating disease of the central nervous system (CNS). The frequent involvement of spinal cord in MS is shown up to 99 % in autopsy series [170]. 70–80 % of the patients with MS have T2-WI cord signal abnormalities [171]. MS plaques are best demonstrated on the T2-WI acquisitions, with signal characteristics of T2-WI hyperintensity and T1-WI iso-hypointensity [172], similar to cord infarction. Cord swelling is a usual feature of the relapsing-remitting form of MS [173, 174], and similarly cord atrophy with axonal loss is a recognized form of MS [175]. Post-contrast enhancement of the demyelination lesions is described due to the associated breakdown of the blood-brain barrier, although significantly lower than their brain counterpart [176].

However, the MS lesions differ from SCI in pattern of cord involvement and extent of CNS distribution. MS plaques present as well-circumscribed foci of T2-WI hyperintensity with asymmetric involvement of the cord. They are characteristically peripherally located, are less than two vertebral segments in length, and occupy less than half the cross-sectional area of the cord. On sagittal acquisitions, they are located centrally, anteriorly, and dorsally. On axial images, plaques are located in the lateral segments and have wedge shape with the basis at the cord surface or round shape if there is no contact with the cord surface [172].

Unlike SCI, MS predominantly involves cervical cord (62 %) and usually demonstrates relapsing-remitting disease course [172]. Simultaneous imaging of the brain to detect synchronous lesions helps to differentiate MS from cord infarction. Up to 92 % of MS patients with cord lesions have associated brain lesions [177] (Fig. 14). Detection of synchronous lesions in brain and spinal cord is useful in differentiating the MS from other inflammatory and cerebrovascular diseases with an accuracy of 93–95 % [178–180].

Transverse myelitis (TM) is predominantly immune system mediated [181]. This may exist as an isolated idiopathic entity or as a part of the systemic process with reported incidence in patients with vasculitis, neurosarcoidosis, multiple sclerosis, and systemic lupus erythematosus [182] and following vaccination [183, 184] and acute gastrointestinal and respiratory illness [181].

ITM has reported a prevalence of 15.6 % in a retrospective study involving 288 patients with transverse myelitis [185]. ITM predominantly affects the middle-aged adults. The term “transverse” is self-explanatory describing the position of the inflammation, that is, across the width of the spinal cord. Thoracic cord is most commonly involved [182]. The criteria proposed by the Transverse Myelitis Consortium Working Group for idiopathic transverse myelitis [186] include:

Typical MR findings of ITM [182] include focal centrally located T2-WI hyperintensity, usually occupying more than two-thirds of the spinal cord cross-sectional area (88 %), signal abnormality that extends for more than three to four vertebral segments in length (53 %), variable presence of cord swelling or expansion (47 %), and patchy or diffuse post-contrast enhancement (47–53 %) [182, 185]. The variability in the duration of clinical onset, extent, and pattern of cord signal abnormality and given up to 40 % of ITM cases can have normal MR imaging [187] make it quite challenging to differentiate ITM from cord infarction (Fig. 15). A detailed history of associated systemic conditions, through clinical examination and close monitoring of the clinical condition in combination with CSF analysis, may be helpful in differentiating these two entities. A history of recent vaccination or viral illness may suggest ITM, but these are often absent. Symptoms isolated to the anterior spinal artery territory suggest infarction over inflammation, but this is not definitive. Progression of symptoms over at least a few hours, symptoms and MRI lesions that extend beyond a vascular territory, gadolinium enhancement on the initial presentation, and cerebrospinal fluid findings of pleocytosis or elevated IgG levels favor transverse myelitis over infarction [188].