INTRODUCTION

Conventional and storage phosphor digital cephalometry generates a radiograph using pyramidal projection geometry. The patient is positioned between a stationary x-ray source and a flat image receptor, which can be a conventional film or a photostimulable storage phosphor imaging plate.1 The latter has a much wider dynamic range and exposure latitude than film-screen systems. This means that overexposed and underexposed images are recorded with equal “brightness.” Almost no retakes are necessary as the result of incorrect exposures.

Modern computed tomography (CT) enables the human body to be accurately visualized in any plane of space, including three-dimensional (3-D) reconstructions. Clinicians treating patients with craniofacial anomalies often use a localization scan, called a CT scanogram, to produce a survey scan of the region of interest.2,3 The CT scanogram generates an x-ray projection radiograph using cylindrical projection geometry. The patient is moved through the beam while the detectors (which are arranged in the arc of a circle) and a stationary x-ray source are in a fixed angular position. This is a projected two-dimensional image of a three-dimensional object and is similar in appearance to a conventional projected radiographic image.

Cephalometry is the analysis of complex craniofacial morphology by linear measurements, angles, and ratios. This technique has been of great value to aid the clinical assessment of the orthodontic patient. However, its use has several limitations, including considerable error due to projection errors,4–7 errors within the measuring system,8,9 and particularly errors in landmark identification.8,10–15

Patients with complex craniofacial anomalies often have had CT scanograms as part of their radiographic assessment; it is important to avoid duplication of exposure to x-rays wherever possible.

This study aims to assess whether lateral CT scanograms can be used to obtain information of the same quality for clinical use as is provided by lateral cephalograms. Gigon et al.13 found a statistically significant, but perhaps not clinically significant, higher level of interobserver error in identifying cephalometric landmarks on scanogram tracings (0.45 ± 0.28 mm) compared with cephalograms (0.33 ± 0.30 mm). Chidiac et al.16 concluded that conventional cephalograms and CT scanograms are close in depicting angular measurements but not linear measurements. These results, however, should be interpreted with caution because the linear measurements were not corrected for radiographic magnification. There is also disagreement regarding the accuracy of landmark identification errors when conventional and storage phosphor lateral cephalograms are compared.12,17,18

In this study, with the use of linear and angular measurements, the role of the CT scanogram for the accurate assessment of craniofacial morphology was assessed. Comparisons were made using direct skull (SK) measurements and those obtained from a digitized cephalometric analysis of lateral scanograms (CTs), conventional cephalograms (CCs), and storage phosphor digital cephalograms (SPs) of dry skulls fixed in a reproducible position.

MATERIALS AND METHODS

Twenty normal, dry, adult human skulls were obtained from the Department of Anatomy, University College, London. Each skull had a stable and reproducible occlusion with most permanent teeth present (including the central incisors) and was in good condition. A 3-mm-thick clear acrylic box (R. Denny, London, United Kingdom) was made to hold the skull. A spirit level (Ebisu Diamond 312132, Dieter Schmid-Fine Tools, Berlin, Germany) was used to verify that the plastic box was level and parallel during the radiographic procedures.

The occlusion, in maximal intercuspation, and outer surfaces of the condyle and glenoid fossa were secured using yellow sticky wax. Plastic adhesive tapes, 2 mm wide, were secured to the skull along the midsagittal and anatomic Frankfort planes. The prepared skull was secured in the plastic box using wax strips and sticky wax, with the Frankfort plane perpendicular to the horizontal, the midsagittal plane in the center, and the occiput at the base of the box.

The skull was placed in a conventional cephalostat (Taylor Engineering Ltd, Birmingham, West Midlands, United Kingdom). An aluminum wedge was placed at the anterior portion of the skull, covering nasion to menton. The film-to-midsagittal plane of the skull and the film-to-focal spot distance were fixed at 13 cm and 180 cm, respectively. Radiographs were taken of each of the 20 skulls using a floor-mounted x-ray tube at 62 kVp and 5.2 mAs. The films (Agfa Curix, Agfa-Gevaert, Mortsel, Belgium), measuring 24 × 30 cm, were then processed in an automated chemical processor (Figure 1).

View Figure

• Download as Powerpoint
• Download Figure

Each skull was placed in a storage phosphor digital cephalostat (Figure 2). The film-to-midsagittal plane of the skull and the film-to-focal spot distance were fixed at 18 cm and 162 cm, respectively. Radiographs were taken of all skulls using a storage phosphor cassette (FCR Imaging Plate, Fuji Photo Film, Valhalla, NY) in a conventional radiographic machine (Siemens Orthophos CD, Munich, Germany) at 62 kVp and 5.12 mAs. The cassettes were processed in a laser read-out unit (FCR 5000R, Fuji Photo Film). Postprocessing algorithms were then applied on the monitor-displayed digital image to obtain optimum image quality. The image was printed out on a laser imaging film (DI-AL, Fuji Medical Dry Imaging, Berlin, Germany), measuring 26 × 36 cm, from a dry laser printer (Fujifilm FM-DPL, Tokyo, Japan).

View Figure

• Download as Powerpoint
• Download Figure

Finally, each skull was placed on the patient couch in the center of a computed tomographic machine (SOMATOM Sensation 16, Siemens, Surrey, United Kingdom), as shown in Figure 3. The vertical light beam was used to orient the midsagittal plane of the skull. The horizontal beam was aligned with a line passing through the right and left porion. Scanograms were taken of all skulls at 100 kVp, 80 mA, 2.8 s scan time, 1.0 mm thickness, 256 mm length, and 512 matrix. The monitor-displayed image was then printed out on a laser imaging film measuring 35 × 43 cm, from a dry laser printer. The radiographic magnification was indicated on the film.

View Figure

• Download as Powerpoint
• Download Figure

Landmarks and measurements used in this study are illustrated in Figure 4 and described in Table 1.

View Figure

• Download as Powerpoint
• Download Figure

Table 1 Cephalometric Landmarks and Measurements

View Fullscreen

Direct linear cranial base measurements of all skulls were made by one investigator using an 8-inch curved spring caliper (FAI CALOUT8, Faithfull Tools, Dartford, Kent, United Kingdom). The caliper was measured against a steel ruler to obtain the linear measurement. Five skulls were measured each day. Each distance was measured twice, 2 weeks apart.

Sixty radiographs were directly digitized using a digitizer (Numonics A43BL, Montgomeryville, Pa) with a linear resolution of 1000 lines per inch and an accuracy of ±0.010 inch (0.254 mm). After digitization of each radiograph, 18 measurements were generated by a customized GELA computer program written by Mr Adrian Hart (Consultant Orthodontist, Raigmore Hospital, Inverness IV2 3UJ, United Kingdom).

Each cephalometric landmark was digitized in a predetermined sequence in a darkroom using a black surround around the radiograph. The midpoint was chosen where bilateral landmarks produced a double image. The radiographs were randomly chosen, and a maximum of five radiographs per day were digitized. Each radiograph was digitized twice, 2 weeks apart. The digitizer was checked daily against an accurate template of varying distances and angles. Conventional lateral cephalograms had a magnification of 7.8%, and storage phosphor lateral cephalograms had a magnification of 12.5%. All linear measurements from the radiographs were corrected for the magnification factor.

Repeatability of the Method

“A priori” ranges for all parameters were decided before any data were collected (Table 2). Data were analyzed using Excel (Microsoft Excel 2003; Microsoft Corporation, Redmond, Wash). The standard deviation (SD) of the differences between the first and second readings was calculated (second reading subtracted from first reading). The coefficient of repeatability (CR) is equal to ±2 SD. For each parameter, if CR fell within the a priori range, the difference between the pair of readings was deemed clinically acceptable, and the measuring method was considered repeatable. If CR fell outside the a priori range, then that parameter was believed to exhibit low repeatability.

Table 2 “A priori” Ranges

View Fullscreen

RESULTS

Results for the repeatability of the method are summarized in Table 3. The numeric value of the mean difference for the method error was near zero (range, −0.74 to 0.63) for nearly all parameters (56 of 57) except UIMx (CT, 1.03°). All direct skull measurements were repeatable. CC had the greatest number of repeatable parameters (12 of 18), followed by SP (10 of 18), and CT had the least repeatable parameters (9 of 18).

Table 3 Assessment of Repeatability for Direct Skull Measurements and Cephalometric Analyses of Conventional and Storage Phosphor Cephalograms and Computed Tomographic Scanogram^a

View Fullscreen

Results for the assessment of bias and agreement are summarized in Tables 4 and 5. Most histograms were found to be normally distributed. Where a statistically significant difference in measures was noted between the different measuring techniques, the numeric value of the mean difference was near zero (range, −0.88 to 1.04) for nearly all affected parameters (16 of 19), except BaN (SK vs SP, 1.23 mm) and UIMx (range, −1.95° to −1.69°).

Table 4 Assessment of Bias and Agreement between Direct Skull Measurements vs Cephalometric Analyses of Conventional Cephalogram, Storage Phosphor Cephalogram, and Computed Tomographic Scanogram^a

View Fullscreen

Table 5 Assessment of Bias and Agreement between Cephalometric Analyses of Conventional and Storage Phosphor Cephalograms vs Computed Tomographic Scanogram^a

View Fullscreen

DISCUSSION

Any new method used to assess craniofacial morphology has to be evaluated by comparison with an established technique rather than with the true (but unknown) value. Repeatability and degree of agreement have to be assessed when two methods (neither of which provides the true measurement) are compared.19 The a priori ranges used in this study were chosen on the basis of expected ease or difficulty of identifying the landmark and the relative magnitude of the craniofacial parameter. The a priori ranges were set subjectively at clinically acceptable ranges and can be changed depending on the clinical situation in which they are to be used. To our knowledge, there are no universally acceptable ranges for all these 18 craniofacial parameters among clinicians or researchers.

Despite a small sample size, the results of this study can be considered valid because of the careful method used. The custom-designed skull holder and spirit level ensured accurate positioning and placement in both the cephalostat and the CT scanner. Direct digitization and cephalometric analyses of radiographs were carried out using the same machine and software program, which reduced the potential for manual error and miscalculation. Any differences between and among direct skull measurements and cephalometric analyses should be due to a combination of differences in radiographic technique and projection and landmark identification errors. Landmarks were chosen on the basis of commonly used cephalometric analyses and their ease of identification. These landmarks were visually determined by inspection and were not derived from any construction of additional points or lines.

Correlation coefficients were not used to assess repeatability and degree of agreement between different methods of measurement. A correlation coefficient graph showing separate and parallel straight lines demonstrates perfect correlation but no agreement between lines. Bister et al.20 showed that the statistical method described by Bland and Altman19 provided a 95% range and allowed for sufficient clinical assessment of a method to be undertaken. However, this statistical technique must be used within the context of the a priori ranges.

Although agreement for SN between skull and storage phosphor digital cephalograms was slightly outside the a priori ranges (by ±0.2 mm), generally, cephalograms and CT scanograms were a viable alternative to direct skull measurements for SN. Therefore, clinical or research studies that monitor or quantify anterior cranial base growth using SN (via cephalograms and CT scanograms) can be considered a viable alternative to direct skull measurement within the context of this laboratory-based study.

Radiographs were not a viable alternative to direct skull measurements for SBa. This was due to the combination of estimating sella and the difficulty in identifying the cephalometric landmark basion. Even when measuring directly on the skull, the coefficient of repeatability for SBa (±1.46 mm) was twice that for SN and BaN. In this study, SBa and BaN were not repeatable on cephalograms; these results were not surprising because basion has been known to be poorly located on cephalograms.10,15 However, SBa and BaN were repeatable on CT scanograms but were not in agreement with direct skull measurements; this finding was unexpected. It appears that the low spatial resolution of the CT scanogram1,2 compromised the validity of landmark identification for basion but not for sella and nasion. However, no evidence was found to support this assumption because the validity of landmark identification on CT scanograms was not investigated in this study. Therefore, quantifying cranial base growth using SBa and BaN (via cephalograms and CT scanograms) is inaccurate and cannot be considered valid.

All anterior face height measurements were highly repeatable. Although agreement for AUFH and TAFH measurements between the scanogram and both cephalograms was slightly outside the a priori ranges (by ±0.3 mm), generally the scanogram was a viable alternative to conventional and storage phosphor digital cephalograms for all anterior face height measurements. All posterior face height measurements from the scanogram were not repeatable and were not in agreement with both cephalograms. Because SN has been established to be accurate on radiographs, this suggests that ANS and menton were reliably located on the radiographs, but not PNS and gonion (determined by inspection). Cephalometric studies have shown gonion (determined by inspection) to have a large envelope of error in both horizontal and vertical directions, while menton and anterior and posterior nasal spines have been shown to have a larger horizontal envelope of error.10,15 Furthermore, the difference in projection geometry could have influenced the degree of agreement between the CT scanogram and both cephalograms. However, no evidence has been found to support these assumptions because the reliability and validity of landmark identification were not investigated on the scanogram.

The scanogram had poor repeatability and low agreement with both cephalograms for SnMx and MxMn. These angles can be considered repeatable and in agreement only if the a priori ranges had been set at ±3.0° (SnMx) and ±3.5° (MxMn). Although repeatability for ANB was slightly outside the a priori ranges (by ±0.31 mm), generally all sagittal angular measurements between the cranial base, the maxilla, and the mandible (SNA, SNB, and ANB) were highly repeatable for the CT scanogram. The scanogram and both cephalograms were in agreement for these three angles, provided the a priori ranges for ANB had been set at ±2.0° (instead of ±1.0°). Therefore, cephalometric assessment and monitoring of skeletal 2 and 3 growth patterns and treatment outcomes (via SNA, SNB, and ANB), assessment of the lower vertical facial proportion (lower anterior face height and maxillary-mandibular plane angle), and the correction factor for ANB using the Eastman analysis21 can be based on the CT scanogram within the context of this laboratory-based study.

The axial inclination of the incisors is of interest in diagnosis, treatment planning, and evaluation of stability after treatment. The CT scanogram had poor repeatability and low agreement with both cephalograms for the upper and lower incisor inclinations to the respective maxillary and mandibular bases. Even if the a priori ranges (mean ± 2 SD) had been set at ±6.0°, these incisor inclinations were not in agreement between the scanogram and both cephalograms. The incisor apices were subjectively found to be the most difficult points to identify reliably in this study. However, no evidence was found to support these assumptions because the reliability and validity of landmark identification were not investigated on the radiographs.

CONCLUSIONS

• The CT scanogram is a viable alternative to lateral cephalograms (conventional and storage phosphor digital) for assessment of these craniofacial parameters: SN, all anterior face height measurements, SnMx, MxMn, SNA, SNB, and ANB.

• The CT scanogram is not a viable alternative to lateral cephalograms (conventional and storage phosphor digital) for assessment of the craniofacial parameters SBa, BaN, all posterior face height measurements, and incisor inclinations.