A New grading system based on T2-weighted axial MRI for Cervical Canal stenosis

Background: Cervical Canal Stenosis (CCS) is a common cervical disease that occurs during and after middle age. There are a variety of grading systems and Measurement methods based on X-ray, Computed Tomography(CT) for cervical canal stenosis. Although many authors measured and classied cervical canal stenosis on the basis of T2-weighted sagittal MRI after the appearance of magnetic resonance, the research on Cervical Canal Stenosis remains to be improved. To explore a more detailed grading system on MRI for cervical canal stenosis and could be used to the clinical diagnosis and treatment. Methods: We retrospectively studied 200 patients with cervical spinal stenosis. Cervical canal stenosis was classied according to the T2-weighted axial MRI with the most serious compression into the following grades: grade 0,normal; grade 1, mild canal stenosis with obliteration of anterior or posterior of cerebrospinal uid (CSF) space ; grade 2, moderate canal stenosis with obliteration of both sides of the CSF space but no signicant changes in spinal morphology; Grade 3, severe canal stenosis with spinal cord deformity, 3A the oppression ratio of cervical spinal cord is under 50%, 3B the angle of the protrusion less than 90°; Grade 4, serious canal stenosis with spinal cord deformity and the oppression ratio of cervical spinal cord is exceeding 50%, 4A the non-ossication protrusion, 4B the ossic protrusion. We collected the relevant clinical and imaging data of the patients and analyzed the relationship between them and our new grading system. Patients were followed up one year later. Results: Our new grading system was easy to learn and communication. The the ICC reliabilities between six readers was between 0.817-0.865. And the Effective sagittal spinal canal was signicantly different along with our grades at each disc level. In terms of clinical data, our grading system is correlated with the patient's preoperative mJOA score and had impact on the choice of surgical strategy. Conclusions: We suggest our new grading system can provided as a reliable assessment that can guide clinical treatment for Cervical Canal Stenosis.

stenosis compared to only the sagittal plane of canal height [12]. Yusuhn Kang et al described the CCS from T2-weighted sagittal MRI as the following grading system: grade 0, absence of canal stenosis; grade 1, subarachnoid space obliteration exceeding 50%; grade 2, spinal cord deformity; and grade 3, spinal cord signal change [13]. Karpova, A et al found quantitative magnetic resonance imaging methods, such as compression ratio (CR), maximum canal compromise (MCC), and maximum spinal cord compression (MSCC), can used to assess cervical canal stenosis and correlated well with clinical severity of cervical myelopathy [14].
As mentioned above, the MRI grading system of CCS is now centered on the T2-weighted sagittal image. However sagittal images usually show the compression of the entire cervical spine, and axial images can cover the morphology of protrusions and the detail of the stenosis [15][16][17]. Josef Bednarik et al have predicted the progression of spondylotic cervical myelopathy from T2-weighted axial MRI [18].
Therefore, we made a new grading system for cervical canal stenosis that is based on the T2-weighted axial MRI. The purpose of this study was to propose a more detailed grading system on MRI for cervical canal stenosis and could be used to the clinical diagnosis and treatment. Methods:

MRI Grading System for Cervical Canal Stenosis
Cervical canal stenosis was classi ed into the following grading system according to the MRI on T2-weighted axial images with the most serious oppression (Fig. 1): grade 0, absence of canal stenosis without obliteration of Cerebrospinal uid(CSF) space; grade 1, mild canal stenosis with obliteration of anterior or posterior of CSF space ; grade 2, moderate canal stenosis with obliteration of both sides of the CSF space but no signi cant changes in spinal morphology; grade3A: the degree of cervical spinal cord compression is below 50%(the cervical spinal cord compression ratio is between 0.4 − 0.2), grade 3B, severe canal stenosis with spinal cord deformity and the angle of the compressive mass less than 90°; grade 4, severe cervical spinal stenosis with cervical spinal cord malformation, cervical spinal cord compression is over 50% the cervical spinal cord compression ratio is less than 0.2 grade 4A, the nature of the herniation is non-bony (Fig. 2), grade 4B: the herniation is the ossify mass.

Case Selection
This study was approved by our Ethics Committee, and informed consent was not required. We retrospectively studied the MRI of 230 patients in our hospital from January 2016 to December 2017. Among the selected 230 patients, 30 patients with either history of trauma, other surgery, or serious heart and lung disease. As a result, 200 patients were included in the nal analysis (150 men and 50 women; mean age, 53.635 ± 7.599 years). Among the 200 patients,149 patients have received surgery nally (Table 1).

Image Acquisition
In our hospital, the MRI scans were performed with either a 3.0 T unit equipment (SIEMENS, Magnetom Verio 3.0, Germany) with a 3T Neck Matrix coil and two imaging protocols of the entire cervical spine: a sagittal T1-weighted (700/10[repetition time msec/echo time msec]) tubro spin echo and T2-weighted (3000/113.0) tubro spin echo. The slice thickness was 3.0 mm, with a 1 mm interslice gap.
To quantify the stenosis, The shortest anteroposterior diameter of the cervical spinal cord was set as AP, the widest transverse diameter was set as W, and AP/W was used to evaluate the degree of cervical spinal cord compression( Figure   3). The value of normal AP/W should be above 0.5. When AP/W < 0.4, cervical spinal stenosis is considered [18,19].

Image analysis
Six observers with different levels of experience, one radiology resident, three Orthopaedic surgeons, and two experienced musculoskeletal radiologists with 5 and 10 years of experience, independently read the images without knowing the patient's condition and age. We introduced our grading system to the readers in the form of words and sketches, some typical examples for a better understanding. The readers were asked to assess the degree of cervical canal stenosis according to above-mentioned grading system at ve sequential levels (C2/3, C3/4, C4/5, C5/6, C6/7), and measure the diameter of effective sagittal spinal canal(ESSC) on the T2W axial MRI. To minimize observer memory bias, one reader read the images twice with an interval of 2 months.

Clinical correlation
Clinical symptoms were collected and physical examinations were performed by a orthopedic resident with 4 years of experience and reviewed by a deputy chief physician with more than 10 years of experience. We collected the general information and positive clinical manifestations (weakness, parethesia, numbness and balance disorders) to calculate the preoperative mJOA (modi ed Japanese Orthopaedic Association Scores) score for each patient. For surgical patients, we also recorded their surgical procedures and treatment segments. After 1 year we contacted all the patients by phone, and collected their mJOA scores and asked them whether had an operation and the surgical produres.

Statistical Analysis
Intraobserver and interobserver reliability of the new MRI grading system were analyzed using intraclass correlation coe cient (ICC). According to Landis and Koch, the agreement was rated as follows: ICC values of 0-0.2 indicated slight agreement, 0.21-0.4 indicated fair agreement, 0.41-0.60 indicated moderate agreement, 0.61-0.8 indicated substantial agreement, and 0.81 or greater indicated excellent agreement [20]. For analysis of the relationship between our grading system and patients' characteristics, such as preoperative mJOA score and surgical procedures, and the MRI ndings we used categoric regression analysis and nonparametric correlation analysis (Spearman correlation) to calculate the correlation coe cients (R). An R between 0.7 and 0.9 indicated a relatively high correlation and R > 0.9 indicated a very high correlation. The level of correlation signi cance was 0.01. GraphPad Prism (GraphPad Software Inc, San Diego, CA, USA) was used for statistical analyses.

Interobserver Agreement
Primary evaluation showed an almost perfect ICC reliability between six readers. Except for the C4/5 disc level (0.781), the ICC reliabilities between six readers was almost perfect (0.817-0.865) at all disc levels ( Table 2). The second step, we asked all readers reclassify all the patients' images within 2 months. And the ICC ranged from 0.792 to 0.859, indicating a better agreement. ICC values were as follow: 0.859 at C2/3 disc level, 0.843 at C3/4 disc level, 0.792 at C4/5 disc level, 0.834 at C5/6 disc level, and 0.821 at C6/7 disc level and all of above were highly statistically signi cant (P < 0.05). To verify if there are differences between all groups according to our grading system, Effective sagittal spinal canal (ESSC) was measured in the next step and was compared between our grades at each disc level by one-way ANOVA. Table 3 indicated that not all disc levels had all of our grades except the disc C4/5. And ESSC was signi cantly different along with our grades at each disc level. In addition, in disc C3/4, C4/5, C5/6 there were a relatively high correlation between stenosis grades and ESSC. Along with the stenosis grades, the ESSC decreased. In the disc C3/4, the R score was 0.813, 0.780 in disc C4/5, 0.826 in disc C5/6, and 0.709 in disc C6/7. It indicated a relatively high correlation. -means there was no this grade in the disc level. The values were given as mean ± SD. Table 4 has shown the distribution of the stenosis grades in each cervical disc level based on our new grading system. In general, grade 3A was most frequently noted except grade 0. Then was grade 2 and grade 1. In the disc C2/3 and C3/4, grade 1 was the most frequently noted grade. And in the grade C4/5 and C5/6, grade 3A was the most. Grade 2 was found at most in the disc C6/7. Series stenosis grade, such as grade 4, was only found in the disc C4/5 and C5/6. And in the disc C4/5 and C5/6, the incidence of stenosis (stenosis grade was above grade 0) was signi cantly higher than any others, which indicated than these disc levels had a evident large burden of stenosis others. And in the disc C2/3, the overall burden of stenosis was lightest. Evaluation of Patient Symptoms To explore the relationship between our grading system and the symptoms of Cervical Canal Stenosis (CCS), we ranked patients according to the grade of the most serious stenosis and analyzed some clinical indicators such as mJOA score, age, gender and surgery (Table 5).  Among the cases we observed, patients with grade 3A stenosis were the most common (90/200). Preoperative mJOA score in all patients was signi cantly different along with our grades(P < 0.01). However, there was no high correlation that preoperative mJOA score decreased as the grading system increased(R = 0.5351). In surgical patients, the Pre-mJOA score was also decreased along with the grading system, and the Pearson R score was 0.

Discussion
Degenerative cervical myelopathy (DCM) is a common degenerative disorder, with the aging of the population getting worse its incidence is also increasing year by year [21], which is characterized by numbness, limb weakness, walking di culties, and serious paralysis [22]. CSS is an important cause of DCM, which bring a serious burden to the family and community [23][24][25].
Although several research groups have studied the relationship between the radiologic parameters and clinical manifestations or outcomes, radiological ndings still can not be used as a predictive factor in clinical severity or decide on treatment options [5,6,11,13,26]. At the beginning, most of the research on CCS focused on the X-ray and CT without considering the epidural fat between the dural sac and the bony spinal canal [5][6][7] [27,28]. After entering the 21th century with the rise of MRI technology, the assessment of CSS resumed in various studies. Laurencin et al contrasted the interobserver agreement among the qualitative evaluation of cervical spinal stenosis on CT myelograms and MR images [26]. It showed MRI fared slightly better than CT myelograms with regard to level of stenosis. But it didn't come up with a speci c grading system only with a degree (none, mild, moderate, or severe), and the observed kappa values were 0.31 and 0.22 respectively. The fair agreement also indicated the importance of a uni ed and speci c standard that can be accepted and used in practical application. The Mentioned previously grading system of Claus Muhle et al was based on the occlusion of the subarachnoid space around the spinal cord and Spinal compression or displacement at maximum exion and extension [11]. This classi cation indicated that cervical spinal motion may contribute to the development of cervical spondylotic myelopathy but had no use to the diagnosis of CCS. Yusuhn Kang et al classi ed CCS into 4 grades according to the T2-weighted sagittal images considering subarachnoid space, spinal compression and the change of the spinal cord signal [13]. However, although there are many scholars had assessed the cervical stenosis based on sagittal MRI, but they did not further verify if there were the differences between their grading systems, and the guiding function to clinical diagnosis and treatment. In addition to there is no consensus and universally adopted scheme of classi cation between radiologists and clinicians. This will hinder the communication between doctors, as well as further research on CCS. Therefore, it is still necessary to continue to explore a more practical and reliable grading system.
We applied our CCS grading system for T2-weighted axial images, decided by Cerebrospinal uid (CSF) space, spine cord form and the character of the protrusion. We assumed that the remaining CSF space of the dural sac and the character of the protrusion were an important factor in predicting the development of the CCS, and the spine cord form was the key to determine the patient's current treatment strategy.
We have repeatedly explored how to ensure our new grading system is consistent with the degree of stenosis and makes it easier to understand. According to our results, the Interobserver variability among readers with different levels was in a range of 75%-86%, which was a fairly high level of agreement. This nding implied that the grading system was su cient as a reliable method to assess the degree of CCS and could be learned and understood easily.
About the clinical severity of CCS and the effective sagittal spinal canal(ESSC), many studies have mentioned the relationship between the symptoms of spinal stenosis and ESSC [5,26]. Our purpose is not to prove the diagnostic value of ESSC for CCS, but rather to compare the differences of CCS among all levels according to our new grading system. Statistically signi cant differences were recorded for ESSC between CCS grades at C2-C7 disc levels ( Table 3). The difference between grade 1 and grade 2 is not obvious. We consider this may be because the cerebrospinal uid space (CSF) on one side is small, and the differences between individuals further narrow this difference. However, there was a difference between grade 1 and grade 2 in clinical data analysis. There is only 18.75% of patients with grade 1 nally had surgery by December, 2017, compared with nearly 70% of patients with grade 2. According to the Muhle [11], with the anterior or posterior of cerebrospinal uid space, when patients with grade 1 stenosis do exion and extension activities the cervical canal cord still has a full activity space. But patients with grade 2 with no cerebrospinal uid space of all cases, when patients do activities in exion and extension, anterior and posterior cord impingement can be observed, thus inducing DCM or other symptoms of spinal cord [29]. Therefore, we insist grade 1 and grade 2. In addition, due to the small number of cases in grade 4B, this result may be biased.
For a new classi cation system, the information to be conveyed should not only be concise, but also play a guiding role in clinical treatment. Some scholars thought that the severity of spinal cord compression is proportional to the severity of clinical symptoms and the high signal intensity on the T2-weighted MRI indicates a poor prognosis [30,31]. Once the oppression occurs, the condition of most patients will gradually deteriorate within a few years [18,32]. Grade 3 and 4 in our grading system are mainly based on the degree of compression of the spinal cord. According to our results (Table 5), once the stenosis reaches grade 4, the patient's mJOA score will drop rapidly, and always accompanied by a stenosis of the adjacent stage (the data is not shown). At this time a single-stage decompression cannot meet the surgical needs.
And there is evidence that about 5% of asymptomatic spinal cord compression patients begin to have symptoms each year [18]. In other words, patients with grade 2 stenosis are more likely to progress to higher grades of cervical spinal stenosis and myelopathy. Our results also show doctors will suggest that most patients with grade 2 stenosis to choose a surgical treatment. This is another reason why we still have to distinguish between grade 1 and 2. Therefore, we recommend active drug intervention for patients with grade 2 stenosis and surgical treatment if necessary. Although no evidence of the superiority of anterior versus posterior surgery can be drawn for patient with multilevel CCS, the role of surgery in stopping the prevention of DCM symptoms has been widely recognized [33].
In terms of 1-year improvement rate, patients with grade 3B had a low improvement rate. We considered that patients with 3B were more likely to have severe never damage. Because a sharp injury was more severe than blunt stresses. For grade 4B patient who undergo anterior surgery, bony compression was often di cult to remove and always accompanied by adjacent stenosis, simple anterior surgery was di cult to achieve a satisfactory improvement rate.
That's why we indicated a patients with 4B would receiver a posterior surgery.
There are also many limiting factors in our research. First, patients with grade 3B and 4B have a small sample size. This will reduce the credibility of our results. Second, this study was conducted at a single medical research center. And a multi-center research is needed, in order to ensure the universality of the classi cation. Third, Lack of forward-looking research. In order to verify the effect of our grading system on clinical practice, we will continue to track the outpatient, grade them by stenosis and collect them outcomes.

Conclusions
As suggested by Ganau et al. [34] the diagnosis and treatment of DCM has been evolving over the past 5 decades as a result of collaborations between clinicians and scientists. Spine surgeons are witnessing a fast-paced evolution in basic and clinical research which include advances in imaging and clinical diagnostic tools supporting new surgical techniques, and reparative/regenerative strategies. Our study suggests that a new grading system could improve the interobserver agreement and provide clinicians with a reliable assessment tool to guide clinical treatment for patients with DCM resulting from CCS.   A patient with serious canal stenosis. Axial T2-weighted MRI shows grade 4A stenosis with a 0.5/1.6≈0.3125 compression ratio at C3/4 level. This patient received a anterior decompression surgery.

Figure 3
The compression ratio of cervical spinal cord is evaluated by AP/W.AP is the shortest sagittal diameter of cervical spinal cord. And W is the longest transverse diameter of cervical spinal cord.