Computer Aided Analysis of Upper Cervical Radiographs

6th Annual Conservative Health Science Conference
Old Brookfield, NY, October 9-11, 1987

Authors: John D. Grostic, B.S., D.C., F.I.C.R.
Edward F. Owens, Jr., M.S., D.C.

Institution: Life Chiropractic College
Marietta, GA


Some doubt exists regarding the reliability of x-ray analysis in general, and of upper cervical x-ray analysis in particular [1-4]. Intra-examiner reliability has generally been found to be good, while inter-examiner reliability is poor. To help objectify the process of x-ray analysis and investigate the sources of error, a computed analysis has been developed at Life College.

The analysis system makes use of a sonic digitizer to locate structures visible on the cervical radiographs and an AppleII+ computer to calculate structural relationships. In all, the operator is instructed by the program to locate 144 points in a predetermined sequence on the cervical series consisting of Lateral, Nasium and Vertex views. Listings, such as Atlas laterality and rotation, head tilt and cervical curve are calculated from the coordinates of the skeletal landmarks.

This report contains the results of repeatability trials for the location of points and testing of the sensitivity of the listing computations to point location error.


Some doubt still exists regarding the use of radiographic analysis to determine the presence of skeletal misalignments. The reliability of techiques of x-ray marking and measurement, especially those of the upper cervical region, has recently received much attention in the literature [1-5]. The typical method of investigation has been to assess the reliabilty of a marking system by sending coded sets of radiographs to field practitioners for blind analysis. The ability of an assessor to repeat his own findings, called intra-examiner reliability, and the agreement between different assessor, or inter-examiner reliability, have both been tested.

In a study by Sigler and Howe [4] of an upper cervical x-ray analysis, the reported intra-examiner error in atlas laterality measurements was 57% while the inter-examiner error was found to be 41% and 71% [4]. It was concluded that upper cervical x-ray analysis was not acceptably reliable. More recently, Jackson et. al. found considerably different results from a similar study of atlas laterality measurements made with the Pettibon technique. Intra-examiner reliability in the latter study was found to be 93-98%, while reliability between different analyzers ranged from 90-98%. The great discrepancy in the results found by Sigler and Howe, and those by Jackson, suggests that further work needs to be done to clarify the sourses of error in upper cervical x-ray analysis.

The approach used at Life College has been to eliminate the errors associated with line drawing and angle measurement by digitizing bony landmarks on cervical x-rays and computing relative positions of vertebrae using the digitized coordinates. This essentially reduces the task of x-ray analysis to the correct location of the skeletal structures. This article will describe the X-ray analysis system, show the errors involved in point location and describe how those errors distort the computation of listings.


The analysis system is composed of a sonic digitizer (Science Accessories Corporation, GP-8) with resolution of .01 millimeters, a light box onto which the digitizer is mounted and an Apple II+ computer which records the coordinates of landmarks and calculates relative skeletal positions. The software for the system is composed of three parts: an input program, an error analysis program and a listing computation program.

The data input program prompts the operator of the system to choose the landmarks on the lateral, nasium and vertex cervical radiographs. Since several techniques of upper cervical X-ray analysis are used by field practitioners and each technique uses slightly different landmarks, we decided to digitize all those points used in each of three techniques with which we were familiar. The techniques used were the Life Upper Cervical technique, which is being developed and taught at Life College, the Grostic Technique and the Atlas Orthogonality Technique. The list of digitized landmarks can be found in table 1. The resulting data set, composed of 144 coordinates, is saved on a floppy disc for later analysis.

The error analysis program is used to calculate the mean and standard deviation of coordinate sets stored on the disc. The program can be run on files gathered by one investigator to test intra-examiner reliability, or used to test repeatability across different investigators.

The listing computation program is used to assess the relative positions of the skull and upper cervical vertebrae. Each radiograph in the cervical series contributes to the computed listings.

The lateral cervical radiograph is analyzed for head tilt, atlas inclination, and cervical curve. The nasium analysis makes use of an algorithm developed at the college to calculate the central axis of the skull from the coordinates of 24 points on each skull edge [6]. The upper angle (that between the atlas plane line and the central axis of the skull) and lower angle (between atlas and lower cervical spine) and the radius of curvatures of the occipital condyles and atlanto-axial joints are also calculated. The vertex film is used to calculate the rotation of atlas, using, in one case, the nasal structures to help define the central axis of the skull, and in another calculation, making use of the skull edge to define the axis.

The computation program results in a listing file that is stored on a floppy disc for later conversion to adjustment vectors, or statistical analysis.

As a first test of the accuracy digitizing system, one set of radiographs was marked with a pencil to show the locations of the landmarks. Five student assistants were then asked to digitize the points in order as prompted by the input program. Each student generated five different data files on the same set of xrays. The error analysis program was then used to analyze the error involved in the digitization process. Next, another set of films was chosen for digitization; this time without any marks on the films. This set of films was digitized by the students and also by three other doctors who were well trained in locating the skeletal landmarks used in cervical analysis. The resulting data was also subjected to error analysis.


  1. Sigler DC and Howe JW. Inter- and intra-examiner reliability of the upper cervical x-ray marking system. JMPT 1985: 8:75-80
  2. Aldis GK and Hill JM. Analysis of a chiropractor's data. Journal and Proceeding, Royal Soc. New South Wales. 1979: 112:93-99.
  3. Anderson RT. A radiographic test of upper cervical chiropractic theory. JMPT 1981: 4:129-33.
  4. Grostic JD and DeBoer KF. Roentgenographic measurement of atlas laterality and rotation: A retrospective pre- post- manipulation study. JMPT 1982: 5:63-71.
  5. Jackson, BL, et. al. Inter- and intra-examiner reliability of the upper cervical x-ray marking system: a second look. JMPT 1987: 10:157-163.
  6. Grostic JD. Vertical central skull line algorithn. FCER Conservative Health Science Research Conference. Palmer College of Chiropractic, Davenport, Iowa.