INTRODUCTION

Standard radiographic studies of the temporomandibular joint (TMJ) such as the plain film radiography and panoramic radiography, have little capacity to reveal anything more than gross osseous changes1 within the joint. Therefore, in some cases a more comprehensive radiographic study is indicated.

Three-dimensional (3D) evaluations, such as computed tomography (CT) have been utilized to evaluate the TMJ. However, historically, high cost,23 large radiation dosage,45 large space requirements,23 and the high level of skill required for interpretation have kept its use to a minimum. With the introduction of limited cone beam technology, such deterrents of CT imaging have been greatly diminished. With several cone beam computed tomography (CBCT) scanners now available, lower radiation dosages,6–9 and lower costs,2 3D radiography is becoming more commonplace in the dental profession as it proves to be an valuable diagnostic tool.

Within the condyle, there is variation in bone density and composition. Cortical bone, trabeculae, and intertrabecular tissue have varying densities and mechanical properties.10–14 These differences present a challenge when examining the bony subarticular surfaces of the condyle with 3D CT imaging. For computed tomography, density is often expressed in the form of CT numbers or Hounsfield units (HU).

Hounsfield15 originally described the HU as an absorption value, and he constructed a scale to demonstrate the accuracy to which the absorption values could be ascertained on a visual image. For the machine he described,15 the scale ranged from air (−1000) at the bottom of the scale, to bone (1000) at the top of the scale. The range of tones between black and white seen in an image can be limited to a large or small window within the scale. This window can then be raised or lowered depending upon the absorption value of the material of interest.15 The examiner must be able to decide what window level and width will most accurately represent the anatomic truth of a tissue under examination.

The purpose of the present study was to determine the ideal window width and level for the examination of the condyle, and, if these were identified, could one reliably evaluate the mandibular condyle using the CBCT 3D reconstruction?

MATERIALS AND METHODS

Selecting the Sample

The 25 dry human skulls in this study were used with the permission of The Department of Anatomical Science, Southern Illinois University at Edwardsville, School of Dental Medicine.

For each condyle, six anatomic landmarks were identified, marked, and photographed (Figure 1). Three linear measurements were made on the condyle including: height, width, and length (Table 1). All direct measurements were made by one operator using an electronic digital caliper (P.N. 50001, Chicago Brand, Fremont, Calif). Reproduction of the original landmark locations on the 3D renderings was assisted with photographs and markings on the condyles themselves.

View Figure

• Download as Powerpoint
• Download Figure

Table 1.  Definitions of Condylar Linear Measurements

View Fullscreen

Isolating and Measuring the Condyle

Each of the 50 condyles was isolated prior to making 3D and volumetric measurements. Frankfort horizontal (FH) plane was constructed by creating a plane from the inferior orbital rim to the superior border of the external auditory meatus. An initial cut was made parallel to the FH plane just above the superior aspect of the condyle. The remaining surrounding structures were progressively removed using various sculpting tools (Figure 2).

View Figure

• Download as Powerpoint
• Download Figure

Three-dimensional multiplanar reconstructions were produced for each of the eight window widths defined in Table 2. Condylar width (CW), condylar length (CL), and condylar height (CH) were measured. These landmarks and the measurements made using them are defined in Tables 1 and 3.

Table 2.  Hounsfield Unit Window Widths (W) Used to Create 3D Renderings of the Condyles. 3D Linear Measurements Were Ac complished on Each of These Renderings

View Fullscreen

Table 3.  Definitions of Anatomic Landmarks

View Fullscreen

In the event of disappearance of portions of condylar anatomy due to variation in window widths, virtual planes are constructed. These planes are constructed perpendicular to the line of measurement.

Volume Measurements at Varying Window Widths

Prior to making volumetric measurements, a complete isolation of the condyle was completed. The final cut was made parallel to the FH plane at the level of the inferior point in the sigmoid notch. The isolation process for volumetric measurements is shown in Figure 3. After the isolation, volumetric measurements were made for each of the 23 window widths in order to define percentages of the condylar volume within each window. Each window represents a range of bone densities defined in the software as HU.

View Figure

• Download as Powerpoint
• Download Figure

The first volumetric measurement was made at WW1 (176 to 275 HU), and each of the remaining 22 volumetric measurements were made in 99 HU width increments extending up to 2475 HU. One final volumetric measurement was made with the total range of 176 to 2475 HU (Table 4). WW24 (176 to 2475 HU) was used to find the total volume in the recommended bone density range from V-works.

Table 4.  Hounsfield Unit Window Widths (WW) Used to Create 3D Condyle Renderings for Volumetric Evaluation (Window 24 is Equal to the Sum of Window Widths 1–23)

View Fullscreen

Data Analysis

Differences between the 3D linear measurements and the gold standard measurements were calculated and analyzed using SPSS 14.0 (SPSS Inc, Rainbow Technologies, Chicago, Ill). Significance testing for 3D linear measurement differences was accomplished using independent t-tests with a 95% confidence interval. Linear measurement percent differences were calculated in the 176 to 2476 HU window for each of the 50 condyles. CW, CL, and CH were individually analyzed. The average percent linear measurement change was calculated for the 50 condyles in each measurement group. All percent changes were plotted on a distribution curve above and below the calculated means for CW and CL. The mean percent change was 18.38 for CL and 15.93 for CW. The distribution of numbers was segmented into the following groups: (1) numbers <10% into groups CW0 and CL0, (2) the numbers between 10% and 24% into the mean group and (3) numbers >24% into groups CW1 and CL1. The mean group range of condyles was excluded from the remainder of data analysis. The distributions of condyles in their respective ranges are shown in Figure 4. The percent volumes of the condyles in the outlying groups were compared using independent t-tests at each of the 23 volumetric window widths (WW) (Table 5). The CL0 and CL1 groups were plotted on a distribution curve to compare the distribution of bone volume in the two groups (Figure 5). The same was done for the CW0 and CW1 groups on a separate graph (Figure 6). Linear measurement error was calculated by using the interclass correlation coefficient on the 12 repeat measurements for CL, CW, and CH.

View Figure

• Download as Powerpoint
• Download Figure

Table 5.  The Percent Volumes of the Condyles in the Outlying Groups

View Fullscreen

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

RESULTS

For 3D linear measurement groups, significant differences were found between the direct measurements and CBCT measurements for the CL group at W7 and W8, and for the CW group at W6, W7, and W8. No significant measurement differences were found for the CH group. The average linear measurements for CL, CW, and CH along with the direct measurement (DM) are shown in Table 6.

Table 6.  Hounsfield Unit Window Widths (W) Used to Create 3D Condyle Renderings for Linear Measurements (Table 2 for W ranges). Average Condylar Length (CL), Condylar Width (CW), and Condylar Height (CH) Are Compared With the Direct Measurement (DM)

View Fullscreen

With respect to the CL0 and CL1 groups, a significant difference was found in percent volume at windows 9, 10, 17, 18, 19, 21, and 22 (Figure 5). For the CW0 and CW1 groups significant differences were found in WW 1–9 and 15–21 (Figure 6).

Table 7 lists independent t-test results and significance for CL0 compared with CL1, and CW0 compared with CW1. CL0 and CW0 groups demonstrated greater percent volumes in the higher density windows than the corresponding CL1 and CW1 groups.

Table 7.  Comparison (Independent t-Tests) of Volumetric Mea surements: CL0 vs CL1, and CW0 vs CW1 at 23 Window Levels^a

View Fullscreen

The inverse relationship was evident in the lower density windows (Figures 5 and 6). So generally, the condyles that showed the greatest linear measurement change in W8 (176 to 2476 HU) were those with a higher percentage of low bone mineral density (BMD), while those with little change in measurement from the gold standard appear to have an increased BMD content.

Linear measurement reliability was tested using the intraclass correlation coefficient (ICC). Repeat measurements were accomplished on 12 condyles for all three linear measurements. A Cronbach alpha of 0.917 was found for the ICC test (ICC > .80 is acceptable).

DISCUSSION

Studies have shown that CT images can be remarkably accurate for linear,316–18 geometric,19 and volumetric20 measurements within the maxillofacial complex.

The purpose of the present study was not to test the accuracy of linear measurements made on the 3D reconstruction, but instead, to utilize its proven accuracy for the purposes of measuring the changes that occur in the condyle because of variation in reconstruction HU window level and window width. Window level and width variation for the 3D linear measurements did have a significant effect on the condylar width and length; however, changes in height were statistically insignificant. For this reason, CH was excluded from the volumetric comparison groups. CW was most profoundly affected by window level and width. Medial and lateral poles of the 3D condylar reconstructions were often the first to exhibit areas of erosion, thereby producing reduced CW measurements.

Significance was found in the following ranges: W6, W7, and W8. W8 (176 to 2476 HU) is the recommended window width for viewing bone with V-works 4.0. CL proved to be a challenging dimension to measure. ICC reliability testing showed measurement reproducibility to be acceptable; however, the extent of erosion was difficult to measure with point to point, and point to plane measurements. This likely resulted in fewer windows with significant measurement differences. Therefore, the CL was possibly more profoundly affected by changes in window level than the results indicate.

The Hounsfield unit has been used to describe physical density and achieve reasonable volume estimates of anatomic structures.10–122021 A number of studies have been performed in efforts to quantify bone density based on Hounsfield values.10–12 Many of these studies were accomplished in order to help classify bone types best suitable to support dental implants.101114 For the present study, HU window widths and levels are manipulated in order to create 3D reconstructions most representative of the anatomic truth. For CW and CL the most accurate windows were below the recommended window for bone, and extended into the soft tissue range22 as defined in Table 8. In a dry skull, extending into the soft tissue range will enhance visualization. However, in vivo the soft tissue will begin to appear and reduce one's capacity to view the bony topography. This would suggest that the CBCT 3D reconstructed image by itself may not be a reliable way to diagnose condylar pathology and changes in condylar morphology.

Table 8.  Window Widths for Density Values and Hounsfield Num bers in the Body²²

View Fullscreen

Though there was significant measurement change in W8 for CL and CW groups, there was a range of variation within each group for the 50 condyles. Some condyles exhibited large change while others showed little or no change at all. It is important to define why these condyles are different from each other. Is an increased difference in linear measurements from the anatomic truth directly related to the BMD composition of the condyle? The purpose of defining the volumetric distribution of BMD within the condyles was intended to answer this very question. The two groups compared graphically in Figures 5 and 6 show the volumetric distribution over 23 density ranges.

Significant differences in bone density distribution were found for both the CL and CW groups. Windows of significant difference were in the high- and low-density ranges. No significant differences were found in the midrange windows. The number of significant windows was greater for the CW0/CW1 group than the CL0/CL1 group. One would expect the CW and CL density distributions to be alike. They do follow the same pattern; however, the CL demonstrated fewer windows of significance between the CL0 and CL1 groups. This could relate to the difficulty presented in measuring morphologic defects in this dimension. Incapacity to record accurately the anatomic changes occurring across the range of HU windows could produce error in condyle selection for the CL0 and CL1 groups, thereby producing a volumetric comparison between two groups of condyles with very little variation between them.

The goal of this study was not to define the range of bone mineral densities that constitute the condyle. A study of this nature would require a sample of untreated cadaver condyles. Even then, research has shown that, HU are only reliable as BMD predictors in full trabecular bone, or in the presence of minimal cortical bone. HU readings from CT scans with thicker cortical bone become less reliable.1213 Also, CT numbers cannot be accepted as an absolute for characterization of a tissue type or lesion.23 CT numbers may vary significantly from one scanner to another, or even between two scanners of the same make and model.2324

However, with some degree of success, standardized calibration methods have been employed in efforts to minimize interscan discrepancies.25 With attention to detail, strict standardization in all parameters, continual manufacturer support, and application of proper calibration methods, reproducibility can be optimized.26 Studies focused on CT numbers as an accurate representation of tissue density have, for the most part, been confined to conventional CT scanners. Recently, Aranyarachkul et al10 examined variations in bone density in designated implant recipient sites using both CBCT and conventional CT. They found both modalities to be consistent in their measurements of bone density value, but the values were generally higher for CBCT. Whether CBCT or conventional CT values are closer to corresponding histologic bone densities has yet to be investigated.

CONCLUSIONS

• Assessment of the mandibular condyle, using the 3D reconstruction, is most accurate when accomplished at density levels below that recommended for osseous examination.