Cone Beam Computed Tomography (CBCT) imaging technology was first introduced in the 21st century. Imagine using this technology has become popular and is often used to help dentists and other health care professionals in diagnostic tasks to improve the care towards orthodontic patients. In Indonesia, Cone Beam Computed Tomography (CBCT) is a new tool with a very limited tool population. This tool is currently available in seven hospitals within the Central Java and Special Region of Yogyakarta areas. The use of CBCT endodontic can detect periapical lesions, root fractures, and anatomical exploration of the root canal system and their abnormalities. CBCT examination is justified if the required diagnostic information cannot be obtained using a conventional dental tool. The results of previous CBCT examinations significantly provided more accurate results for detecting periapical lesions than conventional and digital radiographs. [1] This study showed that 26% of periapical lesions detected on CBCT examination were not detected on periapical radiographs. Even if two consecutive periapical X-rays from the same tooth were made at different angles, the accuracy of the detection of periapical radiographs could not be similar to the results of the CBCT scan. [2]. The previous study concluded that CBCT could be used to determine endodontic WL in combination with EAL [3], while other studies stated that limited CBCT could be used for WL measurement. [4, 5] In onother study it was explained that electronic measurements were more reliable than CBCT scans for WL. [6] The difference in the results of these studies arose due to different voxel sizes because each using different CBCT parameters will affect the image quality.

CBCT is a very useful diagnostic tool for conventional endodontic treatment to improve treatment planning. Spatial resolution and noise from CBCT image results will greatly affect the image quality associated with diagnostic accuracy. Unclear anatomical images are likely to occur because of poor image quality. Factors affecting unclear anatomical images are related to voxel size and there is a need to improve image quality. Advances in CBCT technology and 3D pre-endodontic evaluation can be used to detect apical periodontitis, diagnose resorption lesions, and plan endodontic treatment of root canal anatomical complexes. [7, 8] Based on the kids of literature, CBCT examination cannot to record subtle changes in attenuation in various tissue radiodensity ranges. Contrast resolution in endodontic is important to distinguish soft tissue, periapical or sinus contents. Optimal image resolution in endodontics involves the smallest structure that is affected by voxel size. From these backgrounds, the role of the variations in voxel size will be examined whether there is an effect on the quality of CBCT endodontic images. This study aims to prove the role of voxel size on physical/photographic image quality, and prove the role of voxel size on anatomical/pathological image quality. This was an empirical study on the role of the application of variations in voxel size on noise and spatial resolution of CBCT endodontic images and was a factual study of the phenomena in the field.

This was a quasi-experimental study with a pre-post test only group design. The selection of this study design aims to determine changes before (pre-test) and after treatment (post-test) in the treatment group. The study subjects were CBCT images.

This study was carried out sequentially, by conducting a preliminary study first. The study began with the setting of variations in voxel size on the object of ACR CT phantom, to produce noise and spatial resolution values ​​physically. Then study on the test object (human skull) was conducted. The use of test object aimed as a substitute for a human head to avoid the risk of exposure to radiation doses. The study sample were taken using a purposive sampling technique which is a technique in determining samples with certain considerations by following the desired study objectives. [9]

The study samples were ACR CT phantom images and the images of the test object (human skull). CBCT scanning used 2 (two) voxel size variations. Each voxel size was scanned with 6 (six) parameter variations. 3 samples of ACR CT phantom images were taken for each parameter.

Instruments used in this study included CBCT Carestream Tool, Laptop, Image-J Program for filtering median filter and sharpening, test object of ACR CT Phantom with the GAMMEX brand and human skull.

Descriptive statistical analysis was used to assess the spatial resolution and noise of ACR CT Phantom. Different test analysis of with Wilcoxon was used to assess the effect of the difference in endodontic anatomical information with CBCT before and after filtering the median filter and sharpening with Image-J. Determination of the optimal value in producing CBCT endodontic image quality was obtained from the best values compilation.

Noise measurement was done by selecting 3 images from each parameter. Then five ROIs were chosen, namely in the middle of the phantom image, in the directions of 3, 6, 9, 12 o'clock as shown in Figure 1. Furthermore, the mean value of each ROI was measured using Image-J software in the Dicom file with the plug-in Bio Format. The noise was measured from the standard deviation of the pixel value in ROI on the Phantom ACR CT module 3 scans. The results of the measurement of the mean noise value are shown in Table 1.

Table 1: Results of noise value measurement on phantom ACR CT using CBCT

Voxel size of 180µm at high energy of KV 90 obtained the lowest mean value of noise of 4.6670 at a 10 cm FOV10x setting. Voxel size of 300µm at high energy of KV 90 obtained the lowest mean value of noise of 5.3650 in the 10 x 10 cm FOV setting. Voxel size of 180µm at low energy of KV 60 obtained the lowest mean value of noise of 8.0702 at a 10 x 10 cm FOV setting. Voxel size of 300µm at low energy of KV 60 obtained the lowest mean value of noise of 8.6029 at a 10 x 10 cm FOV setting. The lowest noise values that produced good image quality based on this study was shown in table 2.

Figure 1. Image Noise of CBCT on ACR CT phantom noise with the axial slice on the phantom diameter of 200mm. a 90 KV, 180 µm b. ROI Measurement with Image-J

Table 2: Results of mean value of noise measurement that produced 6 good image quality

Figure 2. ACR Phantom module 4 scanning to measure high contrast (spatial) resolution as shown in Figure 2.

The results of high contrast (spatial) resolution measurement used the exposure factors of 60 KV, 90 KV, 40 mAs with voxel size variation of 180.300 µm and FOV size variations of 8X8 cm, 10x5cm, 10x10 cm. Spatial resolution was measured by visual analysis of bar patterns on the axial cut. Visual analysis of the bar pattern is determined only by grouping 9 and 10 lp/cm. Only two groups of images could be seen as pairs of lines that are separated on the axial due to the limited size of FOV as seen in figure 3. Example of axial cut images of voxel size variations.

Figure 3. Pair of axial cut lines of ACR CT phantom module 4 (a) Voxel size of 180 µm, FOV 8x8, 90KV, 5 mA (b) Voxel size of 300 µm, FOV 10x5, 60KV, 5 mA

Spatial resolution was clear at high KV and not clear at low KV parameters. The results of visual measurements are shown in table 3. The mean value of spatial resolution at high energy level of KV 90 was better than at low energy level of KV 60.

Table 3: Results of Visual Spatial Resolution Measurement on ACR CT Phantom

The further test was a scan using a CBCT instrument on a human skull object as shown in figure 4. The exposure factor setting was the same as the test stage on ACR CT phantom. Exposure factors are 60 KV, 90 KV, 40 mAs with voxel size variations of 180 µm and 300 µm FOV 8 X 8cm, 10 x 5cm, 10 x 10 cm. The results of endodontic anatomical images were evaluated for spatial resolution and noise values and were assessed for anatomic information of lower premolar teeth by observers namely two Radiologists with more than 5 years experience. Scanning results of skull object with CBCT instrument and anatomy assessed can be seen in Figure 5.

Figure 4. Human Skull

Figure 5. Anatomy of CBCT images assessed by observers 1. Enamel, 2. Dentin 3. Pulp chamber 4. Root canal, 5. Root length.

As many as 12 parameter variations of the samples of image results measurement were obtained. In Figure 6. The image of the skull that has not been filtered on the median filter and without sharpening with Image-J. And an example of image 7. Skull image that has been filtered on the median filter and uses sharpening with Image-J. In Figure 5. Filtering of the median filter was performed to reduce noise in the image, and then sharpening was arranged to add image detail.

Figure 6. CB CT images of skull scan before Image-J filtering

Figure 7. CBCT images of skull scan after Image-J filtering

Conformity analysis between observers was carried out on the results of the anatomic information and image quality assessments on CBCT scans. Evaluation of all images was carried out by the observers. The results of the image assessment were analyzed by the kappa test and shown in table 4. Observation of the image assessment was carried out twice to measure the consistency of the results conducted by the two observers.

Table 4: Kappa test for anatomic assessment of lower premolar teeth image using CBCT

The results of the kappa test for the assessment of anatomical information on endodontic images of lower premolars using CBCT showed that observers 1 and 2 before processing using image-J obtained 0.756, and after processing using Image-J obtained 0.866. This indicated that the conformity of the answers of the two observers had a value of > 0.75, with a perfect conformity level.

Table 6. Different test on voxel size variations of anatomic information

There were differences in anatomical information on lower premolar teeth using CBCT after filtering the median filter and sharpening with Image-J. Bivariate analysis to test anatomical difference used the Wilcoxon Signed rank Test due to non-parametric scale data. Differences in anatomical information after difference test using the Wilcoxon Signed rank Test are shown in table 6. The p values of <0.05 in the pre and post difference test meant that there was a difference in anatomical information after processing with image-J. Based on table 6, in the variation of high energy paired samples at voxel sizes of 180 and 300 µm, it was obtained p values of <0.05. In the variation of low energy paired samples at 180 and 300 µm voxel sizes obtained p values of > 0.05. In the voxel size of 180µm at high energy of KV 90, the best anatomical information was obtained after filtering with image-J with a mean value of 4.214 in the 8x8 cm FOV setting. An inter-voxel difference test on 300 µm at high energy of KV 90 obtained the best anatomical information after processing with image-J with a mean value of 4.214 in an 8 x 8 cm FOV setting. Anatomical information difference test between the voxel size of 180µm and a voxel size of 300µm at KV 90 obtained a mean value of 4.214 in the 8x8 cm FOV setting.

The inter voxel difference test of 180µm at low energy of KV 60 obtained the best anatomical information after being processed with image-J with a mean value of 3.357 in the 10 x 5 cm FOV setting.

The mean ranks of anatomical information are shown in table 7.

Table 7. Application of filtering median filter and sharpen ing with Image-J towards the optimal anatomical information on CBCT endodontic images

The best anatomical information values were in the parameter settings of voxel sizes of 180 µm and 300 µm 8x8cm, KV 90 after filtering the median filter and sharpening with Image-J. Wilcoxon test to observe differences in anatomical information based on voxel size variations can be seen in table 8.

Table 8. Wilcoxon test to observe differences in anatomical information based on voxel size variations

Information N : Number of Test Parameter Variations (FOV 8x8cm KV90, FOV 8x8cm KV60, FOV 10x5cm KV90, FOV 10x5cm KV60, FOV 10x10cm KV90, FOV 10x10cm KV90

The way to determine the best voxel size values can be seen in table 9 as the compilation of the results of spatial resolution visual analysis, the best noise and the best anatomy. Six parameters were obtained which resulted in optimization of endodontic anatomical image information with low noise value and spatial resolution value using visual analysis which had a good mean.

Based on table 9, it can be seen the best physical test results and selection of voxel size was at the voxel size of 180µm KV 90 FOV 10x10cm. Whereas regarding the best anatomical information on the KV 90 FOV 8x8cm setting to assess the CBCT endodontic, the 180µm and 300 µm voxel settings did not show a difference.

Table 9: Compilation of Study Results

Optimal endodontic images were those that have the good spatial resolution, anatomic values, the lowest noise in the 180µm KV 90 FOV 10x10cm voxel parameter setting as shown in Figure 8.

Figure 8. Endodontic image of lower premolar teeth in a 180µm KV 90 FOV 10x10cm voxel parameter setting