J Oral Maxillotac Surg 56:743-748, 1998

In Vitro Analysis of the Accuracy of Subtraction Radiography

and Computed Tomography Scanning for Determination of Bone Graft Volume

John Jensen, DDS, * Jens Kragskov, DDS, PbD, f Ann Wenzel, DDS, PbD, Dr Odont, f and Steen Sindet-Pedersen, DDS, Dr Med SciJ

Purpose: This study evaluated the accuracy of digital subtraction radiography (DSR) and three- dimensional computed tomography (3D CT) for determination of bone graft volume in the maxillofacial region.

Materials and Methods: Standardized bone defects were made on the top of the alveolar ridge in 10 dry pig mandibles. To resemble the clinical situation, a bone block was harvested from the symphyseal region of the mandible and fixed in the defect. True bone graft volume was determined by the water displacement technique (VOL I) and correlated to direct measurements by calipers (VOL II). The mean gray value of the bone graft as imaged by DSR was correlated to the directly measured thickness. Furthermore, VOL I was correlated to the 3D CT of the bone graft (VOL III) and to the 3D CT with the bone graft fixed in the defect (VOL TV).

Results: There was a strong correlation between VOL I and VOL II 0 = .95), whereas there was a poorer correlation between mean gray level in DSR and measured bone thickness (r = .63). A strong correlation was also registered between VOL I and VOL III (r = .97) and VT)L I and VOL IV (r = .97).

In implant surgery, augmentation of local defects on top of the alveolar ridge with bone grafts may facilitate implantation in patients with insufficient bone vol- ume. Several surgical procedures have been devel- oped to establish a sufficient bone volume, such as autogenous onlay bone grafts fixed with the im- plants,‘-’ autogenous bone grafting followed by im- plant installation several months later,*-’ and guided bone regeneration.8,9 However, experimental studies

*Staff, Department of Oral and Maxillofacial Surgery, Aarhus

University and Aarhus University Hospital, Aarhus, Denmark.

tStaff, Department of Oral and Maxillofacial Surgery, harhus

University and Aarhus University Hospital, Aarhus, Denmark.

*Professor and Chairman, Department of Oral Radiology, Royal

Dental College, Aarhus University, Aarhus, Denmark.

§Profcssor and Chairman, Department of Oral and Maxillofacial

Surgery, Aarhus University and Aarhus University Hospital, Aarhus,

Denmark.

Address correspondence and reprint requests to Dr Jensen:

Department of Oral and Maxillofacial Surgery, Aarhus University

Hospital, Norrebrogade, DK-8000 Aarhus, Denmark.

0 1998 American Association of Oral and Maxillofacial Surgeons

0278.2391/98/.5606-0008$3 00/O

have shown that there is often a significant and unpredictable loss of bone graft volume caused by resorptionlo-‘3 To evaluate new bone grafting proce- dures in reconstructive surgery, it would therefore be helpful to have reliable diagnostic methods to deter- mine longitudinally the survival of the bone graft.

Recently, digital subtraction radiography has been used to quantify bone thickness, relating it to density changes in the subtraction image.14J5 In reconstruc- tive surgery, the use of three-dimensional computed tomography (3D CT) imaging has increased the quali- tative information for preoperative planning. How- ever, very little has been done to assess quantitatively the accuracy of the software used for 3D reconstruc- tion.16J7 Furthermore, most studies have focused on the quantification of larger anatomic structures, such as the liver and large bony structures.lsJ9

The aim of this in vitro study was to evaluate the accuracy of these two noninvasive radiographic diag- nostic methods for determination of bone graft vol- ume in the maxillofacial region.

Materials and Methods

In 10 freeze-dried pig mandibles, slaughtered at 1 year of age, standardized artificial bony defects (ap-

743

744 CLEFT, IMPLANTS, BONE GRAFTS

proximate length: 10 mm, width: 8 mm, and height: 4 mm) were made on the top of the alveolar ridge in the incisal region, and 3-mm titanium microscrews (0s~. Leibinger, Freiburg, Germany) were inserted in the buccal aspect of the bone as reference markers (Fig 1). To simulate a typical clinical situation for reconstruc- tion of alveolar ridge defects or atrophy, a corticocan- cellous “bone transplant” was harvested from the symphyseal region of the same mandible and fixed to the residual ridge by 6-mm microscrews (0s~. Leibin- ger), simulating a reconstruction of the bone defect (Fig 2). All bone transplants were placed with the cortical side upwards and the cancellous side facing the defect.

Before fixation, the true volume of the transplant (VOL I) was determined by the water displacement technique. In addition, the volume of the transplant (length X width X thickness) (VOL II) was measured to the nearest tenth of a millimeter using an Iwansson caliper, and the average of five measurements on each side (in each corner and in the middle) was recorded. All measurements were made twice by two of the authors (J.J. and J.K.), and the means were used for further calculation.

Individual lilmholders were manufactured for serial identical radiography, and a radiograph was taken of the alveolar defect (Kodak Ultra-speed dental film, DF 54 [Rochester, NY]; and Philips Oralix 65 dental x-ray unit [Eidhoven, the Netherlands], 65 kVp, 7.5 mA, focus-film distance, 31 cm; exposure time, 0.4 set). A wedge step was not included in the exposures. After placement of the graft, the radiographic examination was repeated in a similar manner.

The film processing was semi-automatic and per- formed the same day using fresh chemicals for all of the radiographs. The radiographs were digitized by a flatbed scanner with a transparency module (Agfa Arcus, Agfa Gaevert, NV Bis Division, Berchem, Bel-

FIGURE 1. An artificial bone defect made on the top of the alveolar

ridge. Three titanium microscrews were inserted buccally as reference markers (arrows).

FIGURE 2. The bone transplant (arrow], harvested inferior to the incisal region, fixed to the defect by a titanium microscrew simulating a bone graft reconstruction.

gium). Digital subtraction was performed using a well-established software program specifically devel- oped for dental subtraction radiography.2o The subtrac- tion program was able to correct for differences in contrast and brightness between the films before subtraction was performed.

Nine reference points on each image were placed on easily recognizeable structureszl for superimposi- tion of the two images before and after reconstruction (ie, microscrews, root apices, cementoenamel junc- tions), and the two images were then subtracted. To avoid the image of the micro-osteosynthesis screws used for fixation, two windows for bone density measurements were drawn in the subtraction image as trapezoid regions of interest (ROI 1 and ROI 2) on each side of the screw (Fig 3). The mean gray value (density) was used as the parameter for bone mass in the ROIs.

CT scans of the bone transplant were performed first (Picker PQ 2000, modtied ear/dental 130 kV, 263 mAs; axial scan, l-mm slice thickness with 50%

FIGURE 3. A typical subtraction image. Two trapezoid windows (ROI 1 and ROI 2) are drawn for bone density measurement.

JENSEN ET AL 745

overlap, 1024 matrix, 40-cm image size, and standard kernel, zero-degree gantry tilt) and then repeated for each mandible at the time when the residual ridge was reconstructed by the transplant. Data were transmit- ted via a network to a reconstruction program for 3D image analysis (Camra, ISG, Toronto, Canada), where a semiautomatic segmentation was performed by thresholds and seed procedures. Each CT image was segmented into transplanted and residual bone by each of the two investigators (J.J. and J.K.). Each segmented object was 3D-image reconstructed and the volume calculated using the volume tool (mea- sured in mms) of the 3D program (LSG) (Fig 4).

DATA ANALYSES

Data were transferred to a statistical program (SPSS/ PC + , Chicago, IL). The volume (VOL II) of the bone graft measured by the Iwansson caliper was correlated to the volume determined by using the water displace- ment method (VOL I). Thereafter, the mean gray value of ROI 1 and ROI 2, and the mean between these two windows (MROI) in the subtraction images, were correlated to the measured bone graft thickness. The bone volume determined by the 3D CT scan before (VOL III) and after (VOL IV) the bone graft was fixed in the ridge defect was correlated to VOL I. Regression analyses were performed using the true thickness and the volume, as the independent variables and the mean densities in the subtraction image windows (MROI), as well as VOL III and VOL IV from the CT scans, as the dependent variables.

A paired t-test was performed between VOL I and VOL II, VOL I and VOL III, and VOL I and VOL IV to evaluate any differences between the methods.

Results

The bone volume as determined by water displace- ment (VOL I) and by direct measurement with the

FIGURE 4. The bone graft fixed to the alveolar defect reconstructed by 3-D CT-scanning.

Graft No VOL I VOL II VOL III VOL N

1

2

i

2

7 8 9

10

149 150 161 141 188 161 192 184 231 252 217 210 161 142 156 147 192 196 174 167 195 165 193 191 124 121 132 129 260 281 261 257 188 184 186 178 202 190 205 197

Iwansson caliper (VOL II) can be seen in Table 1 for all 10 grafts. A plot of the correlation between VOL I and VOL II is shown in Figure 5. The correlation coeffi- cient (r) was .95, which was statistically significant (P < .002). Furthermore, a t-test (P = ,414) indicated that there was no systematic difference between the volumes obtained by these two methods.

The mean gray shade in the subtraction image windows (ROI 1, ROI 2, and MROI), and the corre- sponding measured thickness (by caliper) for all 10 grafts can be seen in Table 2. Most of the bony structure in one of the transplants, consisted of loose cancellous bone, which resulted in a density measure- ment with a significant discrepancy compared with the other nine (see ’ in Table 2). Therefore, that transplant was excluded from the regression analysis.

28o I 260

1 /

240

220 220 4 i 200 4 - 1 180

160 4 / .

100 I / I / I I I I I i 100 120 140 160 180 200 220 240 260 280 ,,,,,,3

Vol II

FIGURE 5. Correlation of bone volume determined by water displace- ment technique (VOL I) and volume determined by direct measurement (VOL II) (1. = .95, P < ,002).

746 CLEFT, IMPLANTS, BONE GRAFTS

Graft No. ROI 1 ROI 2 MROI

Thickness (mm)

1 2

3 4*

i 7 8

9 10

156.6 157.0 156.8 6.9 155.4 155.0 155.2 6.2 151.6 151.8 151.7 6.5 145.5 147.7 146.6 7.5 155.7 154.1 154.9 7.4 158.6 155.0 156.8 6.5 147.7 148.1 147.9 5.7 158.4 160.0 159.2 8.3 158.1 155.4 156.8 6.7 154.4 153.4 153.9 7.3

The density in ROI 1 was related to bone thickness by a correlation of r = .54 (P = .13), whereas density in ROI 2 had a correlation to thickness of r = .72 (P = ,003). A plot of the correlation between the mean of ROI 1 and ROI 2 (MROI) and thickness of the bone graft is illustrated in Figure 6. There was a correlation of r = .63 (P = .067) between these two variables.

The volume of the bone graft itself (VOL III> and when it was fixed to the mandibular defect (VOL IV) as determined by the 3D CT-scan, can be seen in Table 1. A plot of the correlation between VOL I and VOL III is shown in Figure 7. The correlation coefficient was r = .97, which was statistically significant (P = .002). Similarly, it is seen in Figure 8 that there was a strong correlation (r = .97) between VOL I and VOL IV

148 0

146 I I I I 5 6 7 a m m

Thickness Vol IV

FIGURE 6. Correlation between direct measurement of bone thick- FIGURE 8. Correlation between bone volume determined by water ness and mean gray value in the subtraction image (MROI) (r = .63, displacement technique (VOL I) and by 3D CT scanning of the bone

P= ,067). graft fixed to the alveolar defect (VOL IV] (r = .97, P = ,002).

280 280

260 -

240 -

180 -

120 140 160 180 200 220 240 260

Vol III

FIGURE 7. Correlation between bone volume determined by water displacement technique (VOL I) and by 3D CT scanning of the bone graft before “reconstruction” (VO1 Ill] (r = .97, P = ,002).

(P = .002). A t-test (P = .667) indicated that there was no systematic difference between VOL I and VOL III, whereas there seemed to be an underestimation of the volume determined by the CT scan with the bone graft fixed to the recipient bone (VOL IV) (P = .012).

Discussion

It can be difficult to measure the exact bone volume over time when bone grafts are used for maxillofacial

260 - 260 e

/ 240 240

220

180 180 -

160

120 140 160 180 200 220 240 260 280 mm3

JENSEN ET AL 747

reconstructive procedures. In several experimental studies,11-13J2J3 the volume was calculated as a prod- uct of length, width, and thickness. These measure- ments were most often performed with a caliper. Due to the small size and irregular shape of some of the bone grafts used in the current study, which is in accordance with the clinical situation3,* it was more convenient to use an Iwansson caliper to perform the five measurements on each side of the bone graft. It has previously been reported that volume determined by water displacement correlates strongly with the volume calculated by direct linear measurements,** which is in accordance with the results from the current study. Because of an often irregular remodel- ling of the bone graft, it has only been possible to obtain rough estimates of bone volume in prospective experimental studies.10J2,24 Therefore, direct measure- ments at the end of a volumetric study has been replaced by measurement on en face photographs where the graft area has been calculated using a digitizing method. Furthermore, to determine the volume, the mean thickness was measured from cross-sectional histologic slides at known magnifica- tions.13J3

Digital subtraction radiography is a sensitive, nonin- vasive method to assess early density changes in alveolar bone.1j,25 Recently, investigators have intro- duced the concept of quantitative analysis of the information of the gray-shade values contained in the digital subtraction image as a parameter to assess the amount of bone loss or gain.14,15,26.29,31 These tech- niques may be broadly classified into two categories to include absolute and relative measures of bone change. Absolute methods require the use of a reference wedge to estimate bone change in terms of volume or mass. In the current study, a wedge was not used because we designed the experiment to simulate the clinical situation. However, the radiographic proce- dures, were standardized because all exposures and film developments were performed the same day under the same conditions. It is well known that there is “noise” involved in digitization of images. However, such ‘“noise” seems to be distributed equally in all images.30

In vitro experiments using bone slices down to 50 urn thickness have shown an excellent correlation between the actual bone mass change and the density change as calculated from the digital subtraction image,l*,l5,31 and the correlation was shown to be stronger for cortical than for cancellous bone.sl In the current study, the correlation between bone graft thickness and the density in the subtraction image was weaker than in the previous experimental studies. The main reasons could be that the measured area in the current study was larger than in previous studies and that the irregular-sized bone graft, to simulate the

clinical situation, consisted of a mixture of cortical and cancellous bone. There may be a weakness in the current method because we compared the mean gray level in the subtraction image with the mean thickness of the graft. However, this way of treating the data was chosen because it would seem to be inaccurate to compare each point of measured thickness with the very same point in the subtraction image.

The ability of 3D CT to render dimensionally accu- rate images has previously been evaluated. Hemmy and Tessierl” reported that by a visual comparison 3D CT reconstruction of dry skulls appeared to represent the skulls accurately. Matteson et a1s2 also reported on the dimensional accuracy of 3D-rendered images. Linear measurements of anatomic structures were accurate to within 0.28%, and angular measurements were correct within 1.39%. Furthermore, Quirynen et al33 showed that using a standard reconstruction based on axial slides, the most reliable cross-sectional images were obtained with a mean absolute deviation of 0.5 mm. However, studies that have investigated the ability to obtain quantitative measurements from 3D CT have shown that measurement error does exist because of inaccuracy of the software.32 Recently, it has been stated that thinner CT slices seem to increase measuring accuracy, as would be expected from the less volume-averaging in the thinner slices.34,35 This is in accordance with the current study, in which axial slices of 1 mm thickness with 50% overlap were used, resulting in an excellent correlation between the measured and the true bone graft volume. However, there seemed to be an underestimation of bone graft volume when it was determined by 3D CT scans with the graft fixed in the alveolar ridge defect. The main reason for this could be a compression of the cancel- lous part of the bone graft when screw-fixed.

The cumulative doses in CT scanning, which were considerably higher than those in conventional dental radiography, should be mentioned as a possible disad- vantage of this method. However, the five to six CT scan slices of the maxilla that are necessary for sufficient diagnosis and treatment planning for local- ized alveolar atrophy almost equal the dose of a periapical survey with intraoral films.36 Using subtrac- tion radiography, the dosage could be lowered sign& cantly compared with 3D CT. However, today there seems to be a significant difference in the diagnostic yield between the two methods in favor of 3D CT. However, we do agree with authors who state that CT scans should be used only for examination of the anatomically difficult jaw structures, particularly in case of bone defects.3’

In the current in vitro experiment, which involved a static situation, there was an excellent correlation between bone graft volume determined by 3D CT and by water displacement, whereas there was poorer

748

correlation between bone graft thickness and the density in the subtraction image. It is well known that there is a significant change in the bone graft structure both internally and externally during healing (ie, architechtural adaptation to functional loading).ss Changes in the internal structures do not necessarily affect the size of the graft, but they could affect the density in subtraction images and be misinterpreted as a change in external dimensions.

References 1.

2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

15.

Adell R, Lekholm U, Griindahl K, et al: Reconstruction of severely resorbed edentulous maxillae using osseointegrated fixtures in immediate autogenous bone grafts. Int J Oral Maxillofac Implants 5:233, 1990 Nystrom E, Kahnberg K-E, Gunne J: Bone grafts and Branemark implants in the treatment of the severely resorbed maxilla: A 2 year longitudinal study. Int J Oral Maxillofac Implants 8:45, 1993

Jensen J, Sindet-Pedersen S, Oliver AJ: Varying treatment strategies for reconstruction of maxillary atrophy with im- plants: Results in 98 patients. J Oral Maxillofac Surg 52:210, 1994 Misch CM, Misch CE, Resnik RR, et al: Reconstruction of maxillary alveolar defects with mandibular symphysis grafts for dental implants: A preliminary procedural report. Int J Oral Maxillofac Implants 7:360, 1992 Misch CM, Misch CE: The repair of localized severe ridge defects for implant placement using mandibular bone grafts. Implant Dent 4:261, 1995 Lundgren S, Moy P, Johansson C, et al: Augmentation of the maxillary sinus floor with particulated mandible: A histologic and histomorphometric study. Int J Oral Maxillofac Implants 11:760, 1996 Raghoebar GM, Batenburg RHK, Vissink A, et al: Augmentation of localized defects of the anterior maxillary ridge with autog- enous bone before insertion of implants. J Oral Maxillofac Surg 54:1180, 1996 Buser D, Bragger U, Lang NP, et al: Regeneration and enlarg- ment of jaw bone using guided tissue regeneration. Clin Oral Implant Res 1:22, 1990 Rominger JW, Triplett RG: The use of guided tissue regenera- tion to improve implant osseointegration. J Oral Maxillofac Surg 52:106, 1994 Zins JE, Whitaker LA: Membranous versus endochondral bone: Implications for craniofacial reconstruction. Plast Reconstr Surg 72:778, 1983 Hardesty RA, Marsh JL: Craniofacial onlay bone grafting: A prospective evaluation of graft morphology, orientation, and embryonic origin. Plast Reconstr Surg 82:5, 1990 Phillips JH, Rahn BA: Fixation effect on membranous and endochondral onlay bone graft resorption. Plast Reconstr Surg 82~872, 1988 Lin KY, Bartlett SP, Yaremchuk MJ, et al: The effect of rigid fixation on the survival of onlay bone grafts: An experimental study. Plast Reconstr Surg 86:449, 1990 Jeffcoat MK, Reddy MS, van den Berg HR, et al: Quantitative digital subtraction radiography for the assessment of peri- implant bone change. Clin Oral Implant Res 3:22, 1992 Christgau M, Hiller K-A, Schmalz G, et al: Accuracy of digital subtraction radiography for detection of bone changes. Or- lando, FL, IADR, 1997 (abstr)

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

CLEFT, IMPLANTS, BONE GRAFTS

Hemmy DC, Tessier PL: CT of dry skulls with craniofacial deformities: Accuracy of three-dimensional reconstruction, Ra- diology 157:113, 1985 Bradick JP, Smith AS, Ohman JC, et al: Estimation of maxillary alveolar cleft volume by three-dimensional CT. J Comput Assist Tomogr 14:994,1990 Evrard S, Moyses B, Ghnassia JP, et al: Validation of the measurement of hepatic volume by three-dimensional com- puted tomography (In French). Ann Chir 46:601, 1992 Viegas SF, Hillman GR, Elder K, et al: Measurement of carpal bone geometry by computer analysis of three dimensional CT images. J Hand Surg 18:341, 1993 Wenzel A: Effect of manual compared with reference point superimposition on image quality in digital subtraction radiogra- phy. Dentomaxillofac Radio1 18:145, 1989 Aagaard E, Donslund C, Wenzel A, et al: Performance for obtaining maximal gain from a program for digital subtraction radiography. Stand J Dent Res 99:166, 1991 LaTrenta GS, McCarthy JG, Breitbart AS, et al: The role of rigid fixation in bone graft augmentation of the craniofacial skeleton. Plast Reconstr Surg 84:578, 1989 Chen NT, Glowacki J, Bucky LP, et al: The roles of revasculariza- tion and resorption on endurance of craniofacial onlay bone grafts in the rabbit. Plast Reconstr Surg 93:714, 1994 Smith JE, Abramson M: Membranous versus endochondral bone autografts. Arch Otolaryngol99:203, 1974 Griindahl HG, Grondahl K: Subtraction radiography for the diagnosis of periodontal bone lesions. Oral Surg Oral Med Oral Path01 55:208, 1983 Vos MH, Janssen PTM, Van Aken J, et al: Quantitative measure- ment of periodontal bone changes by digital subtraction. J Periodont Res 21:583, 1986 Ruttimann UE, Webber R: Volumetry of localized bone lesions by subtraction radiography. J Periodont Res 22:215, 1987 Bragger U, Pasquali L, Rylander H, et al: Computer assisted densiometric image analysis in periodontal radiography: A methodological study. J Clin Periodontol 15:29, 1988 Bragger U, Burgin W, Lang NP, Buser D: Digital subtraction radiography for the assessment of changes in peri-implant bone density. Int J Oral Maxlllofac Implants 6:160, 1991 Wenzel A, Sewerin I: Sources of noise in digital subtraction radiography. Oral Surg Oral Med Oral Path01 71:503, 1991 Jeffcoat MK, Reddy MS: Digital subtraction radiography for longitudinal assessments of per&implant bone change: Method and validation. Adv Dent Res 7: 196, 1993 Matteson SR, Bechtold W, Phillips C, et al: A method for three-dimensional image reformation for quantitative cephalo- metric analysis. J Oral Maxillofac Surg 47:1053, 1989 Quirynen M, Lamoral Y, Dekeyser C, et al: The CT scan standard reconstntction technique for reliable jaw bone volume determi- nation. Int J Oral Maxillofac Implants 5:384, 1990 Curry TS, Dowdey JE, Murry RD: Christensen’s Physics of Diagnostic Radiology (ed 4). Philadelphia, PA, Lea & Febinger, 1990, pp 289-322 Covino SW, Mitnick RJ, Shprintzen RJ, et al The accuracy of measurements of three-dimensional computed tomography reconstructions. J Oral Maxillofac Surg 54:982, 1996 Frederiksen NL, Benson BW, Sokolowski TW: Effective dose and risk assessment from computed tomography of the maxiUo- facial complex. Dentomaxillofac Radio1 24:55, 1995 Dula K, Mini R, van der Stelt PF, et al: Hypothetical mortality risk associated with spiral computed tomography of the maxilla and mandible. Eur J Oral Sci 104:503, 1996 Enneking WF, Burchardt H, Puhl JJ, et al: Physical and biological aspects of repair in dog cortical bone transplants. J Bone Joint SurgAm 57:232, 1975