Spinal cord injury (SCI) is one of the most life-changing neurological conditions affecting individuals and the health care system [1]. In recent studies, the prognoses of patients with SCI have been predicted to some extent by using magnetic resonance imaging (MRI) findings, such as the intensity of the intramedullary hemorrhagic signal and the degree of cord compression [2,3].

Conventional MRI is routinely used to evaluate damage to spinal cord tissue. However, MRI provides little information on the health and integrity of the spinal cord tissue itself and the white matter; thus, it has limitations in reflecting neurological and functional statuses [4]. Diffusion tensor imaging (DTI) is an MRI technique for estimating the directional diffusivity of water that correlates with axonal integrity [5]. DTI has been used for the quantitative anatomical evaluation of white matter tracts, and the relationship between DTI and clinical statuses or outcomes has been previously studied [6,7]. However, DTI has important technical limitations that can result in images with low signals and artifacts due to the minute size of the spinal cord or motion artifacts due to swallowing, respiratory motion, and cerebrospinal fluid pulsation [8].

Surgical intervention is performed in many patients with SCI to re-establish and prevent secondary injury. These interventions may affect the image sensitivities of MRI and DTI due to artifacts, and their effects on the neurologic and functional prognoses are still unclear.

We aimed to demonstrate the utility of DTI and tractrography in predicting the outcomes of patients after cervical SCI and to determine whether the prediction accuracy differs between DTIs performed before and after surgical intervention.

In this study, we analyzed whether neurologic, functional, and DTI parameters changed after surgery compared to preoperative statuses. Between the preop and postop groups, there were no differences in age, sex, cause of trauma, initial NLI, or AIS scores. The most common NLI was C5, and the severity of most of the subjects’ impairments according to the ISNCSCI was AIS D. Patients underwent anterior cervical discectomy and fusion (ACDF), lateral mass fixation, laminectomy, or laminoplasty surgeries. In addition, ACDF was the most common treatment among patients, whereas there was no difference in surgical treatment approaches between the two groups (Table 1). Both groups showed significant improvements on the UEMS, K-MBI, the three motor subscales (self-care, sphincter control, and transfer/locomotion), and total FIM scores after 4 weeks. In addition, there was a significant improvement in the LT sensory score and the total sensory score of UE in the preop group. There were no significant differences between the preop and postop groups (Table 2).

When comparing the quantitative DTI parameters of the two groups, there were significant differences in several parameters. FA values at the C3 and C4 levels were significantly higher in the preop group than in the postop group, and ADC values at the C3, C4, C5, and C7 levels were significantly lower in the preop group than in the postop group. When comparing FA and ADC values at each patient’s level of injury, ADC values at the level of injury were significantly lower in the preop group than in the postop group. Moreover, fiber numbers at all levels (C3, C4, C5, C6, and C7 levels) were significantly higher in the preop group than in the postop group, but there were no differences in crossing fiber counts (Table 3).

Spearman correlation coefficients between preoperative DTI parameters and the patients’ baseline ISNCSCI motor/sensory scores and K-MBI and FIM scores (at 4 weeks after injury) demonstrated that FA at the C7 level correlated with LEMS (r=−0.601). In addition, the ADC values at the C4 level and fiber numbers at the C3 and C7 levels were significantly correlated with the baseline PP sensory score of the UE (r=−0.471, r=0.442, and r=0.417, respectively). Moreover, the FA and ADC values at the C7 level before surgery correlated with the LEMS (r=− 0.601 and r=0.722, respectively). The FA values at the C3 level were associated with the baseline total FIM score (r=0.476), and the fiber numbers at the C3 level were associated with the baseline sphincter control subscale of FIM (r=0.474).

When comparing the preoperative DTI parameters and the follow-up clinical statuses (at 8 weeks after the injury), FA values at the C4 level before surgery were associated with the LT sensory score of UE after 8 weeks. ADC values at the C4 level before surgery were also significantly correlated with LT and the total sensory score of UE (r=−0.622 and r=−0.561, respectively). In addition, ADC values at the C5 and C7 levels before surgery were associated with LEMS. Moreover, fiber counts at the C6 level were associated with the self-care subscale and total FIM scores, and fiber numbers at the C7 level were correlated with the same items and K-MBI (Table 4).

When comparing the postoperative DTI parameters and baseline clinical statuses, which were performed at similar periods (at 4 weeks after the injury), many values showed significant associations (Table 5). The FA values at the C3 level after surgery were highly associated with UEMS (r=0.736). Additionally, the ADC values at the C4 and C5 levels, fiber numbers at the C3, C4, C5, and C6 levels, and crossing fiber numbers from the C3 to C5 levels after surgery were associated with LEMS. Many of the postoperative fiber numbers and crossing fiber numbers were associated with the functional scores. Furthermore, the fiber numbers at the C4 to C6 levels and crossing fiber numbers from the C3 to C5 and C3 to C6 levels after surgery were correlated with the K-MBI score. Additionally, the crossing fiber numbers from C3 to C7 after surgery were related to the self-care subscale of the FIM score. The fiber numbers at the C3 to C6 levels (r=0.61, r=0.694, r=0.804, and r=0.632, respectively) and the crossing numbers from the C3 to C5 and C3 to C6 levels (r=−0.676 and r=0.628, respectively) correlated with the sphincter control subscale of the FIM score. Moreover, the fiber numbers at the C5 and C6 levels and crossing numbers from the C3 to C5 and C3 to C7 levels correlated with the transfer/locomotion subscale of the FIM score. The fiber number at the C4 and C5 levels and all of the crossing fiber numbers (C3 to C5, C3 to C6, and C3 to C7 levels) were associated with the total FIM scores.

In a comparison of the postoperative DTI parameters and the follow-up clinical scales (8 weeks after injury), there were a limited number of correlations. The ADC values at the C4 level (r=0.572) and fiber numbers at the C5 and C6 levels were significantly correlated with the follow-up LEMS (r=−0.658 and r=0.575, respectively).

In the follow-up group who underwent initial DTI and follow-up DTI, the UEMS and LT sensory scores of the UE and all of the functional scores were significantly improved after 4 weeks. When comparing initial and follow-up tractography data, the fiber numbers at C6 and C7 levels and the crossing fiber numbers from C3 to C6 levels were significantly higher in initial values than in follow-up values. Fig. 3 shows the fiber tractography of 10 patients in the follow-up group. Furthermore, imaginary fibers at the C3 and C6 levels and crossing fibers from the C3 to the C6 levels were compared before and after surgery. Four weeks after surgery, all of the patients showed improvements in UEMS, K-MBI, and FIM compared to the baseline values. However, the number of fibers at the C3 and C6 levels and the number of crossing fibers from the C3 to C6 levels in 8 patients were postoperatively decreased compared to pre-operation.

The UEMS and functional statuses of cervical SCI patients were significantly improved after 4 weeks, and there were no differences between the preop and postop groups. Preoperative FA values at the C3 and C4 levels and fiber numbers at all levels were significantly higher than the postoperative values. In contrast, the ADC values at the C3, C4, C5, and C7 levels were significantly lower than the postoperative values. Fiber numbers and crossing fiber numbers after surgery were significantly decreased compared to before surgery.

FA measures the “magnitude” of total diffusion, and this effect has been attributed to anisotropic diffusion, with 0 indicating isotropic diffusion, and 1 denoting infinite anisotropy; in addition, FA values closer to 1 indicate a high degree of anisotropy. Any damage to the axonal membrane results in diffusion becoming unrestricted and isotropic. Moreover, the ADC or mean diffusivity is a measure of the average diffusion in all directions, with a low ADC value indicating that the nerve fibers are organized, whereas a high value indicates disorganization within the fiber tracts [12]. Previous studies have shown that SCI patients have lower FA values and higher ADC values than controls [7,13]. Physical trauma can lead to the disruption of axons, thus causing disorganization of the spinal cord white matter and creating barriers to the diffusivity of the water molecules [14].

In this study, the postoperative FA values were lower than the preoperative FA values, and the postoperative ADC values were higher than the preoperative ADC values. Therefore, this indicates that postoperative DTI has demonstrated axonal membrane damage and disorganization within fiber tracts compared to preoperative DTI. This may be influenced by the difficulty in analyzing postoperative DTI. Metallic implants located within the spine can cause geometric distortions that may cause inaccuracies in the measurement of DTI parameters and tractography [15]. This was also demonstrated in our study, wherein the failure rate of DTI analysis was more than doubled after surgery than before surgery.

Several animal studies have examined changes in quantitative DTI parameters of spinal cord tissue during the acute period [16–18]. One of these studies reported a decrease in axial diffusivity (AD), which is a measure of diffusion parallel to the spinal cord following injury, which could be used to distinguish SCI according to its severity. The reduction in AD was shown to be correlated with locomotor recovery [18]. Similarly, in this study, FA and ADC values at the C3 level were associated with the LEMS and the locomotion/transfer subscale of FIM.

In a study examining the optimal timing for obtaining DTI in SCI, several studies have shown that FA reductions in DTI performed within 3 hours or at the middle time point of 24 hours after SCI is the best time for identifying the injury severity [18,19]. However, in this study (and unlike in previous studies), the preoperative DTI performed on the same day as the SCI was not significantly better than the postoperative DTI in predicting the prognosis after SCI.

In a previous study, motor and sensory scores of the ISNCSCI were positively correlated with FA and negatively correlated with ADC values in chronic SCI patients [2]. The correlation between the reduction in diffusivity and the severity of SCI could help to quantify the amount of spared neural tissue in the spinal cord and to provide an estimate of functional recovery after injury [14]. In another study, crossing fiber numbers and connection rates calculated from tractography showed strong correlations with motor scores from C5 to T1 levels in cervical spinal cord injury patients [7]. In this study, as in previous studies, both motor and sensory scores and functional scores demonstrated the same trend. Additionally, fiber numbers and crossing fiber numbers were also positively correlated with clinical status. In both preoperative and postoperative DTI, many parameters showed a correlation with neurologic and functional statuses. However, this correlation was different between the preoperative and postoperative DTIs, and it was also different according to the clinical status at each time point. The most significant correlation was found when the DTI and clinical status were performed simultaneously. In particular, the postoperative fiber numbers at the C3 to C6 levels were most strongly correlated with the baseline K-MBI and FIM scores at 4 weeks after the injury. The postoperative DTI parameters significantly reflected the clinical state at the time of the evaluation.

This study had several limitations. First, the initial time of performing DTI and tractography was different in the preop and postop groups. The initial DTI of the preop group was performed on the same day as SCI, but the initial DTI of the postop group was performed at 4 weeks after SCI. In addition to metal implants, the pathological process of the injured spinal cord or surrounding tissues over 4 weeks in the spine may have affected DTI and tractographic findings. Second, the types of surgery performed with patients are diverse. Spine surgeons can use various methods, such as closed or open reduction, decompression, and anterior or posterior fixations, depending on the injury condition of the cervical spine. There was no significant difference according to the type of surgery between the two groups, but the surgery methods may affect DTI and the tractographic findings. Third, the sample size of each group was small, and the follow-up period was not long enough to delineate the long-term neurological and functional prognostic values of the DTI and the tractographic findings. Last, there were no accurate statistical data regarding the epidemiology of patients with spinal cord injury in Korea. Therefore, based on statistics in the United States, the most common AIS in incomplete cervical SCI patients was AIS D (58%) [20]. In our study, the most common AIS grade was AIS D (76%). The difference in the ratios could be considered a limitation.

Nevertheless, the most meaningful significance of this study is (to our knowledge) that it is the first study to clarify the relationship and prognostic power of DTI and tractography by individually analyzing preoperative and postoperative DTI in cervical SCI patients.

In conclusion, we performed DTI and tractography before and after surgery in cervical SCI patients and found correlations with their neurological and functional statuses. DTI and tractographic findings before surgery showed better FA and ADC values, more fiber numbers, and a lower failure rate for interpretation than those taken after surgery. However, the preoperative imaging findings could not reflect the neurological and functional prognoses better than the postoperative findings. The postoperative images showed better correlations with the neurological and functional statuses when both were executed at similar time frames. Therefore, DTI and tractography in cervical SCI patients should be separately interpreted before and after spine surgery.