Acute traumatic spinal cord injury (SCI) is a devastating condition with an annual incidence ranging from 8.0 to 57.8 per million worldwide [1, 2]. For many years, magnetic resonance imaging (MRI) has routinely been used in the acute setting to characterize the injury and assist in surgical planning. More recently, there has been much interest in utilizing objective measures derived from MRI scans to characterize the severity of neurologic impairment and to predict outcome. The large heterogeneity amongst SCI patients, especially with regard to extent of neurological deficit, injury pattern, overall injury severity, comorbidities, surgical strategies, and rehabilitation care, makes prediction of neurological outcome very difficult [3]. Current methods for predicting outcome based on the initial functional neurologic examination are limited, and there is a great need for objective measures to better classify injury severity and predict outcome.

MRI is currently the gold standard for imaging of SCI and the radiographic characteristics of this injury on MRI are associated with neurological outcome [4]. Since 1988, it has been identified that the presence of spinal cord edema and hematoma portend a worse neurological outcome [5, 6]. It has been suggested that the extent of spinal cord edema is dynamic in the first days after SCI, based on experimental findings [6]. The first indirect clinical evidence was presented in 2008; it was demonstrated that increased time to MRI lead to an increased length in spinal cord edema [7]. Direct evidence that spinal cord edema is dynamic was shown in 2012 when the expansion rate of spinal cord edema between two consecutive MRIs in the first 72 h after injury was described [8]. However, how spinal cord edema evolves after this initial expansion phase remained unclear.

The work of Aarabi et al. has revealed that to utilize objective MRI features as measures of injury, it is important to acknowledge that the processes captured by MRI are not static; rather, these changes are dynamic and evolve rapidly during the acute period after SCI [8]. When interpreting cord edema on MRI, for example to classify injury severity, or as a surrogate measure of a novel intervention’s therapeutic effect, how the extent of cord edema changes over time must be considered. To our knowledge, there are no studies that systematically evaluate the MRI characteristics of traumatic SCI over the first 3 week post-injury. The objective of this study was to characterize the dynamic nature of these MRI characteristics during the first three weeks following SCI using serial MRI scans and to assess the feasibility of conducting such a study in acute SCI patients.

This prospective study, for the first time, describes the temporal changes of MRI characteristics in SCI during the first 3 weeks after injury. We identified an increase in vertical length of spinal cord edema in the first 48 h after SCI, followed by a gradual decrease in the 3 weeks after injury. Not surprisingly, the greatest lengths of spinal cord edema and hematoma were found in those with the most severe (AIS-A and B) injuries. Hematoma in the spinal cord was seen in all AIS-A and B patients, in 50% of the AIS-C patients, and in none of the AIS-D patients.

The first serial MRI study in SCI was performed by Shimada et al. and was published in 1999 [12]. This study reports MRI characteristics of SCI in the first year after injury and showed a decrease in length of spinal cord edema between the initial MRI and the scan performed 3 weeks after the injury [12]. Since no serial MRI scans were performed in the first week after SCI, this study was not able to assess for any increase in spinal cord edema during the first 48 h after injury [12].

The first direct evidence that spinal cord edema increases in the first days after spinal cord injury was provided by Aarabi et al. in [8]. Two consecutive scans in the first days after SCI were performed in 42 patients and showed an increase in spinal cord edema during the time interval between the scans [8]. The described spinal edema expansion rate of 0.9 mm/h by Aarabi et al. is comparable to the mean 0.6 mm/h that we found in our 24–48 h interval [8]. As shown in the current study, the expansion rate steadily decreases during the first 48 h after SCI. Therefore, the lower value in the current study could be explained by the timing of the initial MRI. In Aarabi’s study, the initial MRI was performed a mean of 6.8 h after injury, whereas in the current study, this was 14.6 h [8].

Hematoma in the spinal cord has been associated with a worse neurological prognosis and is one of the most important predictors for neurological outcome [5, 13]. Spinal cord hematoma in SCI is predominantly seen in AIS-A and B patients, with reported ranges being 50–100% and 14–72%, respectively [9, 14, 15, 16]. However, hemorrhage in AIS-C patients and in combined patient cohorts of incomplete injuries (AIS-B-D) has also been reported [9, 14, 16]. These findings are all based on single MRI studies and in a heterogeneous time window after injury, ranging from <24 h to 12 days. This might explain why the percentage of spinal cord hematomas in the AIS-B and C patients in the current study was higher than reported by the previous studies, especially since 25% of the hematomas in this study were not identified on the initial MRI.

Length of hematoma is sparsely reported in literature. Boldin et al. report a mean length of 10.5 mm in AIS-A patients and a mean of 4.0 mm in incomplete injuries [14]. These findings are comparable with the hematoma length found in the current study.

MCC and MSCC as methods to quantify spinal canal compromise and spinal cord compression/swelling after traumatic SCI were introduced by Fehlings and Furlan et al. [11, 17]. Several studies have used these methods to analyse canal compromise and cord compression after traumatic SCI [9, 13, 16, 18]. MCC and MSCC reported in these studies range between 22–62% and 23–58%, respectively, for complete injuries and 14–38% and 20–52%, respectively, for incomplete injuries [9, 13, 16]. The MCC and MSCC reported in the current study fall within the lower percentile reported in literature. When compared to other studies, the current study consists of a relatively large number of injuries caused by low energy mechanisms (falls) that might result in a lower mean MCC and MSCC [9, 13, 16]. Recently, the level of pre-existing spinal cord compression in patients with SCI without bone injury, as measured by the MSCC, has been related to the severity of neurological injury [19]. In the current study, only 5 patients without bone injury were included, and therefore, no relation between MSCC and AIS grade could be identified; nevertheless, all these patients had pre-existing cord compromise with a mean MSCC of 16%.

The observed swelling of the spinal cord on MRI is a well-known qualitative variable in SCI and has been associated with more severe injuries and worse neurological prognosis [9, 18, 20]. The spinal cord swelling found in the current study was predominantly seen in the AIS-A and B patients and, therefore, in line with our current understanding.

A secondary aim of the current study was to evaluate if it was feasible to conduct a serial MRI study in the acute phase after SCI. In total, 78% of the planned MRIs were actually performed. The most commonly missed MRI was the scan that was planned at 48 h after injury. During the 24–48 h interval, the patients in our study were often undergoing surgery or just recovering from it, often in the ICU, intubated, sedated, and on vasopressors. Therefore, in some cases, transfer to the MRI scanner was deemed too risky for a research MRI. In addition, the MRI at 3 weeks after the injury was often missed, especially for the AIS-C and D patients. These patients were usually already discharged to a rehabilitation facility without resources to accommodate the transport to our hospital for the follow-up MRI.

There are very few serial MRI studies performed during the first weeks after a trauma or other acute event. One serial MRI study conducted in the first weeks after a stroke achieved a total of 88% completed scans [21]. Based on the fact that the patient population in the current study often required acute surgery, prolonged intubation, and ventilation, the 78% completed scans, in our opinion, should be considered as a reasonable achievement.

Clearly, missing a substantial part of our patients’ data is a limitation of this study. However, this might be an inevitable consequence of studying severely injured patients during the acute phase after their trauma. Another limitation of our study is the small sample size per AIS grade, especially the AIS-B group (n = 2). Nevertheless, the other incomplete (AIS-C and D) and the complete injuries (AIS-A) are well represented, and therefore, it is still reasonable to draw conclusions from this descriptive pilot study.

This study was designed as a descriptive pilot study with a small number of patients and this limits the statistical analysis that can be performed. This study was specifically designed to identify the change of MRI characteristics over time enabling us to design further studies with an appropriate sample size. For this study, a single observer with high ICC values is considered adequate to identify the dynamic nature of MRI characteristics in SCI; nevertheless, for future studies, multiple observers will be used.

The relation between the dynamics of MRI characteristics and neurological recovery is an important subject of future serial studies. Furthermore, it will be important to identify which time point for an MRI scan after SCI has the best prognostic value for neurological recovery. The recently introduced BASIC classification for SCI strongly correlates with potential neurological recovery [22]. Including the BASIC classification in future serial MRI studies is an important step to determine the time point at which an MRI has the best prognostic value [22].

This study demonstrates a clear pattern of temporal changes in the MRI characteristics that accompany the first 3 weeks after SCI on serial images. We have shown that serial MRI studies in traumatic SCIs are feasible; however, it should be anticipated that some of the planned scans may not be performed due to the clinical condition of the patient. Based on the current study, it is important to realise that spinal cord edema on MRI is dynamic. This should be taken into account when interpreting SCI on MRI for clinical or scientific purposes. This study provides further support for the current understanding that many morphologic changes occur in the injured spinal cord in the first 48–72 h after injury. Assuming that these morphologic changes are manifestations of the pathophysiologic processes that occur post-injury, the findings would support the notion that significant secondary damage is occurring in these first few days and it is a time period, well suited for neuroprotective agents. Importantly, for novel neuroprotective interventions being tested in acute SCI, the alteration in these measurable dynamic changes on MRI may prove to be a useful surrogate outcome measure of biological effect. Such biomarkers would be highly useful in the testing of novel therapies, but in the case of early MRI measures, there is definitely the need to understand the dynamic nature of the changes.