MRI criteria of developmental lumbar spinal stenosis revisited.
Abstract: Purpose. It is somewhat surprising that radiographic criteria for lumbar stenosis have been transposed from radiography and CT to MR without scientific validation. As these radiographic criteria were developed via population studies with criteria defined by two standard deviations from the mean, we sought to perform the same methodology via MR.

Methods. The study was approved by the institutional review board; the requirement for informed consent was waived. One-hundred patients referred for possible metastatic disease, aged 4 to 94 were studied. Measurements were obtained on a midline sagittal T2-weighted (6000/120) image at each disc level, as well as at the mid-vertebral level. The distributive mean, and standard deviations were calculated and -2 SD was used as a "cutoff" for spinal stenosis. To assess for interobserver variation, 20% of the measurements were repeated by a second observer. To assess for intraobserver variation, another 20% of the measurements were repeated a second time at a minimum of a two month interval.

Results. The spinal canal was narrowest at L5-S1 (mean: 1.16 cm), and widest at L1-L2 (mean: 1.56 cm). Overall the narrowest measurements were at the intervertebral disc space and were narrower at the lower disc spaces. In our population, the lowest cutoff limit (two standard deviations below the mean) had a range between 0.38 cm at the L3-L4 disc space and 0.9 cm at the L1 vertebral level. Notably at the L3 level the size range was from 0.77 to 1.75

Conclusion. Traditional measurements of canal diameters may be too large when applied to soft tissue analysis on MR. We suggest using a cutoff of smaller than 0.90 cm for developmental stenosis.
Subject: Spinal canal (Stenosis)
Magnetic resonance imaging
Authors: Chatha, Deep S.
Schweitzer, Mark E.
Pub Date: 10/01/2011
Publication: Name: Bulletin of the NYU Hospital for Joint Diseases Publisher: J. Michael Ryan Publishing Co. Audience: Academic Format: Magazine/Journal Subject: Health Copyright: COPYRIGHT 2011 J. Michael Ryan Publishing Co. ISSN: 1936-9719
Issue: Date: Oct, 2011 Source Volume: 69 Source Issue: 4
Accession Number: 289120464
Full Text: It has been over 50 years since primary lumbar spinal stenosis was first described. (1-4) Lumbar stenosis may be classified as acquired or developmental. (5) Acquired, or degenerative, stenosis is usually readily visible on imaging, since the changes are centered at the disc space. (6) Developmental stenosis is more difficult to diagnose, since it is diffuse by definition and does not include the readily visible hallmarks of disc-centered pathology. (7,8)

A large amount of interest in imaging has concentrated on degenerative lumbar stenosis and the grading of the stenosis of the central canal. Eisenstein's two large anatomic studies of skeletons found the lower anteroposterior diameter of the spinal canal in adults to be 12 mm and 13 mm. (9,10) Other studies have shown values closely resembling these values. (11-14) These criteria for stenosis were transposed, often without adaptation, from radiography to myelography, computed tomography (CT), and more recently magnetic resonance (MR) imaging. Myelographic and smaller CT scan measurements of the lumbar spine demonstrated a mean anteroposterior canal diameter between 12 mm and 14 mm, with a measurement of 11.5 mm considered small. (15-19) Less data is available using MRI to quantitate developmental stenosis (20) and no study (outside of the the pediatric population (21)) has sought to determine a normal distribution for canal diameter. Thus in clinical practice, developmental stenosis is often regarded as an anteroposterior spinal diameter of less than 12 mm. We sought to address this deficit in the literature.

Stenosis is a quantitative diagnosis that is made when the measurement of an individual is outside the range of normal. Therefore, the criteria for stenosis should be developed from an analysis of a normative distribution of measurements within a population. As has been shown in other areas of the musculoskeletal system, traditional radiographic measurements were often made because of the ease of identification of osseous landmarks. One advantage of MR is that direct measurements can be made of the soft tissue in question. This has led to modification of several skeletal measurements in the era of MR. (22,23) Therefore, criteria for spinal stenosis on CT and radiography cannot be assured to be directly applicable to MR. Consequently, we sought to study the "normative" distribution of spinal canal measurements at various anatomic points on MR. This was hypothesized as an attempt to develop a distribution of spinal canal measurements and potential MR criteria for developmental spinal stenosis.


Our retrospective study was approved by the hospital's institutional review board and was compliant with the Health Insurance Portability and Accountability Act (HIPAA). A waiver of informed consent was obtained.


One-hundred patients (36 males and 64 females) ranging in age from 4 to 94 (mean: 61.9 years of age) were studied. Patients who were referred for possible metastatic disease and who had no evidence of secondary spinal tumors were evaluated.

Selecting the patient population for a "normal" distribution is difficult. Healthy volunteers were not imaged, since they may not provide a broad distribution that is likely to be skewed to the left. Similarly, we chose not to evaluate patients referred with spinal symptoms, since that would skew the distribution to the right. We selected patients who were undergoing screening for possible metastatic disease, since they approximate the general population imaged in age and usually have non-specific symptoms. Although this is an imperfect group, we felt that it was similar to that used in prior, non-MR studies of spinal stenosis. (11-13,24)

MR Imaging and Measurements

Patients were scanned at 1.5 T (Siemens, Erlagen, Germany). The following measurements were obtained on sagittal T2-weighted (6000/120) images. At each disc level (L1-L2, L2-L3, L3-L4, L4-L5, and L5-S1) and at the mid-vertebral level (L1, L2, L3, L4, and L5), a perpendicular line was drawn on a midline image from the posterior aspect of the spinal canal to the anterior aspect of the canal using electronic calipers (Magic View, Siemens, Erlagen, Germany), and the diameter was measured. Measurements were performed in centimeters (cm), to the nearest hundredth (Fig. 1).


The distributive mean and standard deviations were calculated. The mathematical definition of a musculoskeletal spine anomaly is outside two standard deviations. (25) Consequently, -2 SD was used as a "cutoff" for spinal stenosis.

Interobserver and Intraobserver Variation

To assess for interobserver variation, 20% of the measurements were repeated by a second observer. To assess for intraobserver variation, another 20% of the measurements were repeated a second time at a minimum of a 2 month interval. The measurements were analyzed using the Pearson correlation, intra-class correlation coefficient, and 95% limits of agreement.


Measurements of the Lumbar Spine

Our normative measurements are presented in Table 1.

Figure 2 demonstrates the distribution of the anteroposterior dimensions of the lumbar spinal canal with accompanying upper and lower limits for each level.


As would be expected, the spinal canal was narrowest at L5-S1 (mean 1.16 cm) and widest at L1-L2 (mean 1.56 cm). Overall, the narrowest measurements were at the disc space level and were narrower at the lower disc spaces. This is also intuitive.

The greatest standard deviation was at L3-L4, with a standard deviation of 1.00 cm. The smallest standard deviation was found at L2, with a deviation of 0.44 cm.

The smallest cutoff of two standard deviations below normal was at L3-L4, with 0.38 cm. The widest measurement with two deviations above normal was at L2-L3, at 2.43 cm. The measurements of the lower limits of each level are compared to those of five other prior studies (Fig. 3). These measurements are presented in a normative distribution curve (Fig. 4).

In our population, the lowest cutoff limit (two standard deviations below the mean) had a range between 0.38 cm at the L3-L4 disc space and 0.93 cm at the L1 vertebral level.

Interobserver and Intraobserver Agreement

There was good interobserver and intraobserver agreement at the mid-vertebral levels where developmental stenosis would be a factor. The intraclass correlation ranged between 0.68 and 0.87. Conversely, there was relatively weak correlation and wide limits of agreement observed for the intervertebral disc levels where acquired stenosis is the predominant factor. Intraclass correlation was poor at these levels ranging between 0.18 and 0.84. There was strongest agreement at L5, and the least agreement was at L1-L2.


Spinal stenosis should not be considered an actual measurement but rather a manifestation of a continuum of various canal sizes. This same concept is applied to other areas of the musculoskeletal system, such as the height of the patella (26,27) and the length of the ulna, (28) among many others. In the situation where there is a continuum of measurements without an external reference, two standard deviations has been used as the definition of an anatomic outlier. This criteria of two standard deviations has been used in other investigations on the imaging of spinal stenosis. (21,30) Thus, we used two standard deviations as our criteria for outliers of our normative measurements.



Initially measurements were based on large anatomic studies of lumbar spine diameter in skeletons by Eisenstein. Examining 275 skeletons, with 1,340 lumbar vertebral levels, he found that 12 mm represented the lower limit of anteroposterior dimensions of the spinal canal in adults. In his study, only 1.7% of vertebral levels measured less than 12 mm, and none were less than 11 mm. (9) His larger follow-up study with 443 skeletons and 2,166 vertebral levels demonstrated slightly larger measurements with an average mid-sagittal diameter of 16 mm and a lower limit of normal of 13 mm. (10) On lumbar radiographs, the lower limit of normal was found to be 15 mm after correction for magnification factor. As these studies were performed on skeletons, they did not take into consideration the soft tissues that are readily seen on MR imaging.

The diameter of the spinal canal was relatively wider at the upper levels than at the lower levels. This is similar in the data of Wildermuth and colleagues (16) and Haig and associates. (30) The data of Hinck and coworkers (14) was constant from the upper levels to the lower levels. This was because their measurements were obtained via radiographs, which does not allow adequate soft tissue contrast.

There was significantly more agreement at the midvertebral level and wider measurements than at the disc levels. This is likely due to discogenic degenerative disease and the consequent difficulty in precisely determining the edges of the degenerated soft tissue.

The most variability was found at L1-L2, possibly because of signal dropoff at the end of the coil. This location is, however, not a common symptom generator. The least variability was at L4 and L5, which is fortunate, as this is an extremely important clinical level. Overall our interobserver and intraobserver correlation was good at these levels.

When the results of this study were compared against the results presented by Hinck and coworkers, (14) the mean of the diameter in this study was significantly lower at each spinal level. The same conclusion applies to the comparison with the studies presented by Larsen (17) and Haig. (30) The differences are explained by our use of the soft tissue canal measurement as opposed to the indirect osseous radiographic measurement.

When comparing our results to the measurements that Wildermuth and colleagues (16) made on myelography, our mean diameter was wider at the L1-L2, L2-L3, and L3-L4 levels. Our mean matched that of Wildermuth's at the L4-L5 level. However, since our standard deviations differed, the lower limit of our measurements were different. The mean of this study was narrower than Wildermuth's mean, at the L5-S1 level. This relatively small difference may be explained by radiographic magnification as well as potential gravitational effects during myelography. When we compared our results to the results of Wildermuth's study on MR, our results more closely match. Since Wildermuth and colleagues (16) used patients with symptoms, their measurements for the lower lumbar spine were narrower, as would be expected in a symptomatic population. In our population, the lowest cutoff limit (two standard deviations below the mean) had a range between 0.38 cm at the L3-L4 disc space and 0.93 cm at the L1 vertebral level, which is well below the traditional cutoff points using other modalities. Using the two standard deviations at the L1 level for an upper limit cutoff, we then would qualify stenosis in our population as those having an anteroposterior diameter of less than 0.93 cm, or roughly 0.9 cm.

The interval of two standard deviations is fairly constant across modalities, while the areas measured, and consequently, the absolute minimal criteria, varies significantly. In a prior radiographic series on spinal stenosis, Hinck and coworkers (14) found much larger spinal cord diameters than the ones in this study. However, their absolute number for two standard deviations was nearly identical. The details of the symptoms of the patients studied were not provided. Eisenstein performed two separate radiographic studies on spinal stenosis. In both of these investigations, he had developed criteria slightly larger than ours, but lower than that of Hinck and coworkers, (14) while evaluating ex-vivo skeletons. He used the previously described smallest radiographic measurement values as a measurement for stenosis (1.2 cm and 1.3 cm). A broad age range of patients was also used. Several studies have shown that although there is minimal increase in the dural sac diameter with age, it is not statistically significant, while other studies have shown that the mid-sagittal growth of the lumbar spine does not change significantly after 4 years of age. (21,23)

We realize that there are several limitations to this study. First, our population was somewhat limited. Eisenstein used 433 skeletons to make his measurements. In another study by Eisenstein, he used 11 skeletons. The population selected is always open to debate, and we would be the first to acknowledge that although not suboptimal, it was not ideal. Other than selecting a large number of random participants, this was thought to be a useful population. If volunteers were selected, many who might be interested would likely have a history of back problem and might participate because of the secondary gain. Our "clinical" population is, however, not dissimilar to that used in other studies of stenosis, or for that matter patella height or ulna variance. We realize that the population is also diverse with a large age range. Perhaps future studies with populations grouped into age ranges may be useful. However, we believe the eventual answer is related less to patient selection than to the concept of the canal as a three-dimensional construct and, in the future, some type of multi-dimensional shape analysis.

Secondly, evaluation of a much larger population of both symptomatic and asymptomatic patients with analysis of patient demographics and clinical history would be a greater undertaking but would prove useful. That is to say that there are patients with significant symptoms of spinal stenosis and normal sized canals and, conversely, patients with markedly stenotic canals who have little or no symptomatology. In addition, patient size itself, more so than age, may influence the size of the central canal and should be evaluated. Thus a measurement in and of itself, without reference to the patients' clinical picture, may be of limited use.


Accepting these limitations, we conclude that traditional measurement of canal diameters may be too large when applied to soft tissue analysis on MR. We suggest using a cutoff of smaller than .90 cm (9 mm) for developmental stenosis. We provide some preliminary data for acquired changes at the disc level, but we suggest that these be viewed with much more caution.

Disclosure Statement

The authors have no financial or proprietary interest in the subject matter or materials discussed, including, but not limited to, employment, consultancies, stock ownership, honoraria, and paid expert testimony.


(1.) Sarpyener MA. Congenital stricture of the spinal canal. J Bone Joint Surg. 1945;27:70-9.

(2.) Verbiest H. [Primary stenosis of the lumbar spinal canal in adults, a new syndrome]. Ned Tijdschr Geneeskd. 1950;94:2415-33.

(3.) Verbiest H. A radicular syndrome from developmental narrowing of the lumbar vertebral canal. J Bone Joint Surg Br. 1954;36(2):230-7.

(4.) Verbiest H. Further experiences on the pathological influence of a developmental narrowness of the bony lumbar vertebral canal. J Bone Joint Surg Br. 1955;37(4):576-83.

(5.) Arnoldi CC, Brodsky AE, Cauchoix J, et al. Lumbar spinal stenosis and nerve root entrapment syndromes: definition and classification. Clin Orthop Relat Res. 1976;(115):4-5.

(6.) Postacchini F. Lumbar spinal stenosis and pseudostenosis. Definition and classification of pathology. Ital J Orthop Traumatol. 1983;9(3):339-50.

(7.) Garfin SR, Rauschning W. Spinal stenosis. Instr Course Lect. 2001;50:145-52.

(8.) Resnick D. Degenerative disease of the spine. In: Resnick D (ed): Diagnosis of Bone and Joint Disorders (4th ed). Philadelphia: W.B. Saunders, 2002, pp. 1382-1475.

(9.) Eisenstein S. Measurements of the lumbar spinal canal in 2 racial groups. Clin Orthop Relat Res. 1976;(115):42-6.

(10.) Eisenstein S. The morphometry and pathological anatomy of the lumbar spine in South African negroes and caucasoids with specific reference to spinal stenosis. J Bone Joint Surg Br. 1977;59(2):173-80.

(11.) Amonoo-Kuofi HS. The sagittal diameter of the lumbar vertebral canal in normal adult Nigerians. J Anat. 1985;140(1):69-78.

(12.) Lee HM, Kim NH, Kim HJ, Chung IH. Morphometric study of the lumbar spinal canal in the Korean population. Spine. 1995;20(15):1679-84.

(13.) Epstein BS, Epstein JA, Jones MD. Lumbar spinal stenosis. Radiol Clin North Am. 1977;15:227-39.

(14.) Hinck VC, Hopkins CE, Clark WM. Sagittal diameter of the lumbar spinal canal in children and adults. Radiology. 1965;85(5):929-37.

(15.) Verbeist H. Neurogenic intermittent claudication in cases with absolute and relative stenosis of the lumbar vertebral canal (ASLC and RSLC), in cases with narroe lumbar intervertebral neuroforamina, and in cases with both entities. Clin Neuro surg. 1973;20:204-14.

(16.) Wildermuth S, Zanetti M, Duewell S et al. Lumbar spine: quantitative and qualitative assessment of positional (upright flexion and extension) MR imaging and myelography. Radiology. 1998;207(2):391-8. Erratum in: Radiology. 1998;208(3):834.

(17.) Larsen JL, Olsen KO. Radiographic anatomy of the distal dural sac. A myelographic investigation of dimensions and termination. Acta Radiol. 1991;32(3):214-9.

(18.) Schonstrom NS, Bolender NF, Spengler DM. The pathomorphology of spinal stenosis as seen on CT scans of the lumbar spine. Spine. 1985;10:806-11.

(19.) Ullrich CG, Binet MD, Sanecki MG, Kieffer SA. Quantitative assessment of the lumbar spinal canal by computed tomography. Radiology. 1980;134:137-43.

(20.) Singh K, Samartzis D, Vaccaro AR, et al. Congenital lumbar spinal stenosis : a prospective, control matched, cohort radiographic analysis. Spine J. 2005;5:615-22.

(21.) Knirsch W, Kurtz C, Haffner N, et al. Normal values of the sagittal diameter of the lumbar spine (vertebral body and dural sac) in children measured by MRI. Pediatr Radiol. 2005;35(4):419-24.

(22.) Shabshin N, Schweitzer ME, Morrison WB, Parker L. MRI criteria for patella alta and baja. Skeletal Radiol. 2004;33(8):445 50.

(23.) Papp T, Porter RW, Aspden RM. The growth of the lumbar vertebral canal. Spine. 1994;19(24):2770-3.

(24.) Appelgren L, Nordborg C, Sjoberg M, et al. Spinal epidural metastasis: implications for spinal analgesia to treat "refractory" cancer pain. J Pain Symptom Manage. 1997;13(1):25 42.

(25.) Taylor JR. Growth of human intervertebral discs and vertebral bodies. J Anat. 1975;120(Pt 1):49-68.

(26.) Hogan SE. Knee height as a predictor of recumbent length for individuals with mobility-impaired cerebral palsy. J Am Coll Nutr. 1999;18(2):201-5.

(27.) Han TS, Lean ME. Lower leg length as an index of stature in adults. Int J Obes Relat Metab Disord. 1996;20(1):21-7.

(28.) Unver B, Gocen Z, Sen A, Gunal I, Karatosun V. Normal ranges of ulnar and radial deviation with reference to ulnar variance. J Int Med Res. 2004;32(3):337-40.

(29.) Cohen ZA, Mow VC, Henry JH, et al. Templates of the cartilage layers of the patellofemoral joint and their use in the assessment of osteoarthritic cartilage damage. Osteoarthritis Cartilage. 2003;11(8):569-79.

(30.) Haig AJ, Weiner JB, Tew J, et al. The relation among spinal geometry on MRI, paraspinal electromyographic abnormalities, and age in persons referred for electrodiagnostic testing of low back symptoms. Spine. 2002;27(17):1918-25; discussion 1924-5.

Deep S. Chatha, M.D., F.R.C.P.C., is in the Department of Diagnostic Imaging, CML Healthcare, 790 Bay Street, Toronto, Ontario, Canada. Mark E. Schweitzer, M.D., is in the Department of Radiological Sciences, University of Ottawa, The Ottawa Hospital, Ottawa, Canada.

Correspondence: Deep S. Chatha, M.D., F.R.C.P.C., Diagnostic Imaging, CML Healthcare, 790 Bay Street, Suite 418, Toronto, Ontario M5G 1N8;
Table 1 Lumbar Spine Measurements of the Studied Population at the
Disc Level and the Mid Vertebral Body Level

Width (cm)              L1       L1-L2      L2       L2-L3      L3

Average                1.41      1.56      1.32      1.51      1.26
Widest                 1.89      2.32      1.76      2.43      1.75
Narrowest              0.93      0.80      0.88      0.58      0.77
Standard Deviation     0.48      0.76      0.44      0.92      0.49

Width (cm)             L3-L4      L4       L4-L5      L5       L5-S1

Average                1.38      1.24      1.29      1.24      1.16
Widest                 2.38      1.76      2.09      1.79      1.78
Narrowest              0.38      0.72      0.50      0.69      0.55
Standard Deviation     1.00      0.52      0.80      0.55      0.61
Gale Copyright: Copyright 2011 Gale, Cengage Learning. All rights reserved.