VALIDITY OF CONE BEAM COMPUTED TOMOGRAPHY VOXEL DENSITY VALUES: A REVIEW.

Kareem S. El-Tabarany, Mona m. Aboulfotouh and Mostafa s. Ashmawy. 1. B.D.s, faculty of dentistry, ain shams university 2010, cairo, egypt. 2. Professor of oral radiology and diagnostic science faculty of dentistry, ain shams university, cairo, egypt. 3. Lecturer of oral radiology and diagnostic science, faculty of dentistry, ain shams university, cairo, egypt. ...................................................................................................................... Manuscript Info Abstract ......................... ........................................................................ Manuscript History

Our search resulted in a total of 92264 studies, 84135 from PubMed and 8129 from Cochrane. Summary of search results are in table 1.
In order to analyze the search results, we included all articles studying the accuracy and/or reliability of BDM using CBCT. Only researches in which information regarding (1) machine type, (2) exposure parameters and (3) presence of gold standard were included. For studies using quantitative data, agreement between measurements or accuracy of measurements should be present. We excluded (1) studies using CBCT but evaluating the accuracy of another imaging modality (2) non-quantitative studies, (3) studies using CBCT machines not available in the market. Results:-All articles were screened by one reviewer on two levels according to the inclusion and exclusion criteria; the initial screening of the articles was conducted using the titles and abstracts, but when these were unclear, the full text was used; this yielded a total of 155 studies. Then the second level was performed by thorough reading and data extraction from the full text; this yielded a total number of 27 article.
The included 27 studies are summarized in Table 2. Studies were gathered according to the searched factor in Table  3    All their results showed change between voxel values more than hundred and twenty-five shades. However, they concluded that machines model was an insignificant factor affecting voxel density values.
All of the studies reviewed regarding machines model used improper scientific methodology. The only research which approached the right methodology was spin-Neto 2015 26 as they had identical parameters in two different machines except for the exposure time that differed by 0.5 sec. That factor was also reviewed and revealed inconclusive outcome.
From our point of view, this factor mostly cannot be tested as a single factor as the studies usually did not fix all the parameters between the tested machines as it is mostly impossible to do so. Machines model can be considered as an inconclusive factor as it was not tested using properly scientific methodology by any study.
Concerning the FOV, total of nine papers discussed the effect of FOV size on voxel grey values. All of them reached the same conclusion that it was a significant factor affecting the voxel density values. 28 Testing the effect of FOV should be done on the same machine so all the exposure parameters can be fixed. From the previous studies, this factor is undoubtedly a significant factor on voxel density values.
As regards the Objects position inside the FOV, total of six researches searched its effect. There was only one research concluded that it has insignificant effect. 61 All remaining papers found it has significant effect. 36,47,54,55,58 Studies done by Nackaerts et al 2011 36  The results of those studies showed that the effect of object position inside the FOV might be machines and/or exposure parameter dependent as the variations found in densities were not the same.
They did not agree upon the position of least effect on voxel values inside the FOV. One study concluded that it is the peripheral position like the dental arch 58 . Other study concluded exactly the opposite 36 , as the later concluded that the peripheral position effect is the greatest not the least.
On the other hand Lagrave`re et al 2008 61 used single machine and fixed all the variables. Although, difference in voxel values obtained from this study reached hundred and forty-five shades, authors concluded that changes in object locations effect is minimal and insignificant.
In our opinion this factor can easily be tested with proper methodology as all other variables in the machine can be fixed.
Objects exo-mass means the mass of the radiographed object which lies outside the FOV. In vitro this part can affect the voxel value by its presence or absence, by its direction, or by the increase or decrease of the mass. However, clinically its presence or absence, and its increase or decrease are based upon the FOV size. And its direction is based upon the ROI position. This means that small and medium sized FOV CBCT machines will always have exo-mass.
Total six studies searched this factor 37,38,57,[62][63][64] . All of them concluded it is significant. All their studies concluded that objects exo-mass affect grey value significantly. And They also specified that the effect is directly proportion to the mass itself.
The study done by Bryant et al 2008 62 concluded that the effect is dependent on its direction and unequal along the FOV. Voxel values was noticed to be high near the exo-mass and decrease gradually to reach the middle which is the most accurate number then continue its decrease till the other side of the FOV with a comb like artifact.
This study also came out with a solution to this issue using algorithms to compensate for the exo-mass effect but only on the machine used there. This study concluded that mA had insignificant effect on voxel values. However, the changes done here was subtle as it was changed from 0.83mA to 0.95mA, and from 0.59mA to 0.60. those minimal changes should produce minimal if any effect on voxel grey values.
mA could be seen as debatable conclusion as the final results are tie. But one paper 53 of the insignificant side tested mA as a combined factor not a single one. So, it was eliminated. Now, mA as a factor affecting voxel value can be deduced to be a significant factor.
Regarding the exposure parameters: it is inherently multiple factors and testing it as a single entity might not help to detect which of them actually affecting the voxel values.
mA, kVp, FOV, number of basis all are exposure parameters and all were found to be significant. Only one study tested exposure parameters as a single entity 67 .
Nomura et al 2010 67 used four different exposure protocols on one CBCT machine. They interchanged two kVp and two mA settings. Although they concluded that voxel values of CBCT is not reliable nor repeatable, they also concluded that this factor affected voxel values insignificantly.
Voxel size has limitations in most machines to be changed alone. It is usually attached to FOV, and with lesser extent the rest of the exposure parameters.
Total four researches tested this factor. three of them found it is significant, and one found it otherwise. 37 Although three out of four researches results showed it was a significant factor, a final deduction of this factors influence was inconclusive. As, it was not tested as a single factor by any of the studies.
Objects mass is not machine dependent, so it can be easily tested as a single factor. Total four researches tested this factor. three of them found it was significant. 51,[62][63][64] Bryant et al 2008 62 searched in this factor as a combined factor with objects position. Which was deduced to be a significant factor. They concluded that it was a significant factor. Two researches which found this factor significant were excluded as they tested it in combination with another factor.
This resulted in a tie. as it only has one paper for each conclusion.
Both studies did not accurately represent clinical situation. Each paper only used one machine, and nothing of the exposure parameters of each machine is remotely related to one another. This factor might be affected by machines model and/or exposure parameter. It needs further researches with proper methodology and on multiple machines to reach a definite conclusion. As its outcome is inconclusive.
kVp is mostly not attached to other parameters. In few machine models it is fixed, others are attached to mA. It can be considered in our opinion an easy factor to test for. Total of two papers searched this factor. One of those studies also investigated the relation between kVp and exo-mass. Both concluded that it is a significant factor. 46 Regarding Time between exposures there were two papers searched this factor. Both researches were conducted with the same team on two successive years and reached that it was significant. 26,27 Testing for this factor is easy. As it requires absolute fixation of every variable.
Spin-Neto et al 2014 and 2015 26,27 used six CBCT machines in their studies. They fixed all parameters for each machine. They tested the machines successively with; zero intervals, 15min intervals, and 30min intervals. Repetition of those protocols were done on another day The results were different voxel values between every exposure at the first session and its corresponding one at the second, and between every exposure and its previous one in every group and in every session.

117
Each of the tested machine showed different pattern of variation in response to each time intervals.
Time between exposure was found to be a significant factor and its effect was massive, differed from one machine to the other, and was unpredictable even in the same machine.
Concerning the Software used to measure voxel values. This factor is the only one that can be tested for retrospectively on any DICOM file and compare between different programmes on the same file.
Two researches tested that factor. Both concluded its insignificance. 56,61 Lagrave`re et al 2008 61 and Azeredo et al 2013 both tested this using two and five programmes respectively. Results showed insignificant difference on voxel values in both studies.
Number of basis is mostly a combined factor as it is mostly attached to voxel size and/or exposure time and/or rotation angle.
Two papers searched this factor. However, each one concluded exactly the opposite of the other. 53 So, number of basis can be considered a significant factor to voxel values.
Only one paper tested receptor type 38 . This factor is very hard to be tested as it is always in different machine and mostly different machines possess different sets of exposure parameters.
Katsumata et al 2007 38 used two CBCT machines with different receptors. One was FPD the other was IIT/CCD. Here it is a combined factor as it is in two different machines with two different exposure parameter sets. A significant effect on voxel values was noticed. But it is an inconclusive factor.
Presence of adjacent air to the ROI is an area specific in the radiographed object. Only one research tested this factor. 68 Yamashina et al 2008 68 tested this factor by using phantom with drilled holes to read the density on the periphery of the holes and away from them.
They found its effect is significant on voxel value.
Exposure time factor is usually a combined factor with voxel size and/or number of basis. Two papers searched this factor one is opposing the other. 26,53 Parsa et al 2013 53 used two CBCT machines with multiple combinations of exposure parameters sets. Those multiple sets did not totally clarify whether this factor is combined or not.
They concluded that it bestowed an insignificant effect on voxel values.
Spin-Neto 26 unintentionally tested exposure time between two different machine and results showed significant changes in voxel values. They here tested it surely as combined factor.
It is an inconclusive factor as it was tested as combined factor by both of the testing studies.
118 Exposure dose is a factor which is based upon other factors. It is affected by the exposure parameters which in turn have some affect voxel value significantly and others do not. Two studies tested this factor. 53,67 parsa et al 2013 53 and Nomura et al 2010 67 both tested this factor as a single factor and concluded that its effect is insignificant.
Most of the machines in the market does not provide a direct dose manipulation protocol.
Presence of teeth had one study searched it. It is considered a clinical form of objects mass and/or objects exo-mass according to its position. The only study of it was done by Nomura et al 2013 69 concluded it is an insignificant factor. 69 Presence of metallic post in or out of FOV is considered the clinical form of objects mass and/or objects exomass. Also, one research tested it and found it is not significant at both stages. Insignificantly affecting factors exposure parameters, software, exposure dose, presence of teeth, presence of metallic post in or out the FOV.
All insignificantly affecting factors were tested by single research paper except for software which was tested by two papers. This necessitates further investigations on those factors.

Inconclusive factors
Machines model, voxel size, objects mass, receptor type, exposure time.
All of the inconclusive factors need further investigations to reach a proper conclusion. Finally, voxel values of CBCT are not reliable nor repeatable.