Article
Reference
Comparison of photodensitometric with high-resolution digital analysis of bone density from serial dental radiographs
DUBREZ, Bertrand, et al.
Abstract
Photodensitometry is known to provide high spatial resolution and continuous measurement of optical density for the analysis of dental radiographs, whereas digitization allows powerful image manipulations but, when using conventional video cameras, gives less spatial resolution and fewer grey levels. The aim of this study was therefore to develop a technique of high-resolution digital analysis for the measurement of bone density following the same principles as those of photodensitometry and based upon the use of a CCD Scanner Camera which provides up to 4096 grey levels and a spatial resolution of 4096 x 4096 pixels.
Twenty-four zones were analysed with both techniques in five serial dental radiographs taken before and after periodontal therapy in eight patients. Statistical comparison of the results obtained by digital analysis and photodensitometry shows that the two techniques have the same accuracy.
DUBREZ, Bertrand, et al . Comparison of photodensitometric with high-resolution digital analysis of bone density from serial dental radiographs. Journal of Dentomaxillofacial Radiology, Pathology and Surgery , 1992, vol. 21, no. 1, p. 40-44
PMID : 1397451
DOI : 10.1259/dmfr.21.1.1397451
Available at:
http://archive-ouverte.unige.ch/unige:47442
Disclaimer: layout of this document may differ from the published version.
Comparison of photodensitometric
with high-resolution digital analysis of bone density from serial dental
radiographs
B. Dubrez,A.Jacot-Descombes*, T. Pun* and G. Cimasoni
Division of Physiopathology and Periodontology, School of Dentistry and *Centre Universitaire d'Informatique, Faculty of Sciences, University of Geneva, Switzerland
Received 3 June 1991, and in final form 30 September 1991
Photodensitometry is known to provide high spatial resolution and continuous measurement of optical density for the analysis of dental radiographs, whereas digitization allows powerful image manipulations but, when using conventional video cameras, gives less spatial resolution and fewer grey levels. The aim of this study was therefore to develop a technique of high- resolution digital analysis for the measurement of bone density following the same principles as those of photodensitometry and based upon the use of a CCD Scanner Camera which provides up to 4096 grey levels and a spatial resolution of 4096X4096 pixels. Twenty-four zones were analysed with both techniques in five serial dental radiographs taken before and after periodontal therapy in eight patients. Statistical comparison of the results obtained by digital analysis and photodensitometry shows that the two techniques have the same accuracy.
Keywords: Periodontitis; radiography, dental; radiographic image interpretation, computer assisted;
bone resorption
Dentomaxillofac. Radiol., 1992, Vol. 21,40-44, February
Introduction
Major changes take place in alveolar bone during periodontitis and after periodontal treatment. The response of bone during the healing of periodontal lesions has recently been monitored by several investi- gators using both invasive and non-invasive methods.
Examples of invasive measurements are the techniques of transgingival probingI and re-entry':", These methods however are limited to the assessment of changes only in bone profile and not in bone density.
Moreover, the re-entry technique can only be applied to observations of bone healing after flap procedures.
Among the non-invasive techniques, 1251 absorptiom- etry and quantitative radiography have been used in periodontal research". Absorptiometry allows a precise evaluation of bone thickness or mass but is only suitable for anterior teeth", To be able to detect small changes of bone density, standardized radiation geometry and a sensitive analytical technique are necessary", Photodensitometry and, more recently, digitization have been used for the latter. Photoden- sitometry (with 250 channels) of dental radiographs has a spatial resolution of about 30,um and provides a continuous measurement of optical density. However the method is complex and the radiograph itself cannot be visualized. In contrast, digitization provides good visualization of the radiographs on a computer screen and has often been combined with image subtraction
techniques to detect minute intrabony lesions/-". Other
in~estig~to~s have also attemp~ed to ~uanti~y bone mineralization from subtracted images'" I. WIth con- ventional video cameras, however, the spatial resolu- tion is only of about 70f.Lm and the optical density is read as 256 grey levels (8 bits acquisition).
The aim of the present investigation was to develop digital analysis with a spatial resolution comparable to that of a photodensitometer and including far more grey levels. The accuracy of such a technique has been compared with that of the previously tested method12 of photodensitometry for the assessment of bone heal- ing after periodontal therapy'F.
Materials and methods
Standardized serial radiographs were obtained from a previous clinical study on bone healing after subgingi- val instrumentation'F in ten patients with an intrabony periodontal lesion. A total of five radiographs were obtained for each patient before and immediately after treatment and then 2, 6 and 12 months later. Reprodu- cible radiographs were achieved by using a standard- ized paralleling instrument':' which contained an aluminium alloy penetrameter.
Preparation of the films
Eight series of five radiographs were selected in which
Analysis of bone density from dental radiographs: B Dubrezet al.
9
there was no overlap between the penetrameter and the superficial part of the crowns of the teeth.
Three horizontal scans termed 'superficial', 'deep' and 'control' were chosen across the lesion in each series of radiographs (Figure 1). To indicate the ends of these scans, tiny reference holes were punched with a thin needle under a binocular microscope on the first film. The second film of the series was superimposed on the first under a stereomicroscope, using the outlines of enamel, fillings and roots as a reference; the same reference points as on the first radiograph were then punched out. The same procedure was applied to the three other films of the series.
To achieve the best possible placement during the digital analysis, three more points were added: two in the lower corners and the third in the middle upper portion of the radiographs.
>
::J 0- Cll
;;(
E E
Quantitative analysis
The standardized radiographs could easily be super- imposed by means of the reference holes. To compen- sate for variations in optical density between individual radiographs, readings were standardized by reference to the penetrameter and the densities transformed with the aid of a computer into mm Al eqs. as described previously'f-!". The readings obtained from the pene- trameter are not linear, due to the fact that the X-ray beam is polychromatic and does not conform to the Lambert-Beer law. This phenomenon, also called beam hardening, has been minimized in the present study by using an AA 7075T6 (Epunal) wedge (the Al alloy which is the closest to bone when exposed to an X-ray beam)!" and fitting a third degree polynomial curve.
Figure I (a) Interproximal area of a periapical radiograph showing the position of the three scans, superficial (S), deep (D) and control (C); the white dots represent the pinholes used for the registration.
(b) The photodensitometric tracings of the three scans, in mm Al eqs.
The ordinate represents the mm Al eqs. for each scan and the abscissa the channel sequence of the photodensitometer. (Repro- duced with permission from Dubrezet ai. 199012)
Channel no.
b
50 100 150 200 250 Photo densitometric analysis
The photodensitometric readings were obtained with a 3C double-beam microdensitometer (Joyce-Loebel, Gateshead on Tyne, UK) driven by a computer (Digital Corp PDP 11/23). The software was programmed in Fortran by the Department of Medical Engineering, Ecole Polytechnique Federale de Lausanne. The scan consisted of reading 250 channels along a predefined
10 . - - - . . . . , . - - - ,
8
Distal root Mesial root
I
Bone4000
3000
OL---'
2000
Scanning length in pixels 1000
Scanning length in channels
b
4 6
oL..---~
a
<Ii
C'Q)
«
E E
Figure 2 The series of curves obtained from scans of Series 2, superficial site, by photodensitometry (a) and by digitization (b). The horizontal straight line allows the integration of the area bounded by each curve. (Adapted with permission from Dubrezet al. 199012)
S. superficial; D. deep: C. control.
140 r - - - ,
Table I Statistical comparison of the differences between the values for changes in bone density by photodensitometry and digitization
line was calculated for these differences, available for the pre-operative radiographs and those obtained at 2 and 6 months and at 1 year (Figure 3b). The probability that the slope of the regression line differed from zero was estimated by means of at-test.
1yr 1yr
-
6 mth Time
6 mth Time
Site Slope P
S 0.0 0.83
D 0.83 0.10
C 0.47 0.35
S -0.25 0.62
D 0.06 0.83
C 0.38 0.63
S 0.21 0.40
D -0.71 0.82
C -2.6 0.26
S -0.22 0.53
D 0.18 0.83
C 0.61 0.79
S 0.88 0.44
D -2.04 0.79
C 3.15 0.11
S 3.23 0.24
D 3.60 0.25
C 6.32 0.44
S -0.14 0.91
D 3.21 0.46
C 4.18 0.28
S 1.74 0.36
D 7.83 0.32
C 7.83 0.47
--
-<
/.
/ / / / /
--.
/o
10 20 r- 30 f-
40 . . . - - - , 90 L -_ _..I..-_ _- - ' -_ _- - ' -_ _- ' -_ _----'-_ _----'
Pre-op. 2mth
-10L..-_ _-'---_ _....L-_ _- - ' -_ _- - ' -_ _- ' -_ _- '
Pre-op. 2mth 130
120 110 100
7 2
6 4
5
8
a
3
b
Series
Figure 3 (a) Comparison of the readings obtained by photodensito- metry (A, ) and by digital analysis (B, --e--) for the scans shown in Figure 2. Values are calculated on a percentage of that obtained immediately post-operatively for the pre-operative radio- graph and at 2, 6 and 12 months post-operatively. (b) Shows the absolute differences of percentages between (A) and (B) at each time and the calculated regression line (slope -0.25,P=0.62)
axis. The scans were performed for each film (Figure 1): three corresponding to the three horizontal scans defined by the holes over the lesions and the fourth along the penetrameter'? and the optical densities expressed as mm Al eqs. Since the same segment of the penetrameter was always used for the reference curve, all the readings of a series had to be completed before any calculation could begin. Readings were expressed graphically with a personal computer (IBM ATPC). As shown in Figure 2, a horizontal line was calculated to fit two points arbitrarily chosen on the left and right constant parts of the curves (these correspond to the root surface at the edge of the lesion). Changes in cumulative bone density values between successive curves could thus be quantified by integrating the areas between each curve and the horizontal line, and expressed as a percentage of that found for the second radiograph (immediately after treatment).
Data analysis
As shown in Figure 3, the results of both photodensito- metric and digital analysis were expressed as percenta- ges of the areas obtained immediately after treatment.
For each series of scans, the percentages calculated by photodensitometry were then subtracted from those obtained by digital analysis (Figure 3b). A regression High-resolution digital analysis
The radiographs were placed on a viewing box and digitized with a high-resolution video camera (Kodak Eikonix Corp., Bedford, Mass., USA) which has a maximum resolution of 4096x4096 pixels and up to 4096 grey levels (12 bits acquisition). To obtain a resolution similar to that of the photodensitometer, the radiographs were digitized at about 1200x1600 pixels and all the available gr~ levels utilized. The camera was driven by a SUNT 3/160 (Sun Microsystem Inc., Mountain View, California, USA) workstation and by using Labolmage software (Computer Science Center, Geneva University)!", it was possible to correct varia- tions in light through the viewing box. Digitized radiographs were then stored on a hard disk, each image requiring approximately 4 MBytes of storage space. The images were analysed using LaboImage:
(1) The penetrameter image was scanned on the five radiographs of each series and a calibration curve derived for each radiograph by a third degree polynomial fit from the same area of the wedge.
The grey levels of the radiographs were converted into mm Al eqs. from the calibration curve.
(2) Successive images were superimposed in identical positions by means of LaboImage using the three pinholes. The full, standardized images were stored and direct pixel by pixel comparison between them was thus possible.
(3) Using LaboImage in the scanning mode, bone density in the interdental lesions was quantified in the same way as with the photodensitometer. The ends of the scans were manually clicked with the mouse between the dots previously punched on the radiographs.
(4) Values obtained with scanning mode were pro- cessed on a personal computer in the same way as described for the photodensitometric analysis.
Analysis of bone density from dental radiographs: B Dubrez et al.
Results
The values for the slopes of the difference between the changes in bone density obtained by photodensitometry and digitization, and the probabilities that such slopes are different from zero, are shown in TableI. The data are given for all three sites (superficial, deep and control) for each of the eight sets of radiographs. In all cases P was greater than 0.05, showing that both techniques can be used to measure change in bone density with similar accuracy.
Discussion
Both photodensitometry and digitization have been used in periodontology to quantify bone density from standardized radiographs, but, to our knowledge, the accuracy of the two methods has never been compared.
Double-beam photodensitometry is highly accurate in quantifying optical density. Moreover, its spatial resolution (30ILm) is particularly convenient for analysing small lesions such as those found in periodon- tal disease. However, there are certain practicallimita- tions. First, the equipment must be calibrated before each session and when comparing several serial radio- graphs, they must all be read at the same session.
Second, the analysis can only be performed along a given axis and therefore corresponds to a rectangular shape. Third, coordinates of the scans have to be checked separately for each radiograph.
In contrast, the conventional video cameras that have been used previously to digitize dental radiographs19 have neither the spatial resolution nor the sensitivity of a photodensitorneter. This is the reason why a more powerful CCD camera was used in the present investigation. Itwas also decided to adapt the entire system (spatial resolution of the camera and analytical procedure) so as to be as close as possible to that used with the photodensitometer. In particular, the software had to be able to quantify the density of any region of the film and present the values in the form of a graph. Our technique, therefore, is different from the classical, semi-quantitative subtraction but in concept is closer to the 'computer-assisted densito- metric analysis' described by others10,11.
One of the main advantages in digitizing radiographs is that geometrically standardized images converted in Al equivalents can be stored for further analysis.
Any error during the procedure of standardization (geometry or conversion) affects the whole image and is easier to detect than a localized error during the photodensitometric analysis. Moreover, placement and superimposability of the radiographs can be visually checked by subtracting them on the screen. Finally, with digitization it is possible to quantify zones other than rectangular in shape. A disadvantage of high- resolution image processing is the high cost of the equipment. The analysis is time-consuming with both techniques: a series of five radiographs required about 6 hours.
With regard to the mathematical manipulations, a slight vertical alignment had to be done before the calculations of area. Interestingly, the amount was exactly the same for both photodensitometry and digitization: it probably corresponds to a shift in the grey levels that cannot be corrected by the aluminium
wedge and is explained by the 10% variation in mains voltage compensation.
Direct comparison between photodensitometry and high-resolution digitization has proved difficult. The principles of the two techniques are fundamentally divergent, and it was therefore not appropriate to compare the absolute values obtained at each observa- tion. Indeed, for both techniques, the values at succes- sive intervals are not independent. The regression line (Figure 3b) takes this fact into account and allows the results, as they evolved over the whole observation period, to be compared. The statistical analysis (Table I) shows that both techniques are equivalent.
Although in the context of the present investigation, neither technique is superior to the other, it should be remembered that digitization could still offer better spatial resolution. Indeed, the maximal resolution of our camera, when used on a zone of 30x30 mm, is of 7ILm which is not far short of the 1ILm size of an individual grain in a Kodak UItraspeed film (personal communication, Eastman-Kodak, September, 1991).
Acknowledgements
The authors express their gratitude to Dr M. Vuagnat, Department of Mathematics, University of Geneva, for the statistical analysis. The help of Mr K. Todorov, Computing Science Center, University of Geneva, is also gratefully acknowledged.
References
1. Durwin A, Chamberlain H, Garrett S, Renvert S, Egelberg J.
Healing after treatment of periodontal intraosseous defects. IV Effect of a non-resective versus a partially resective approach.J Clin Periodontol 1985; 12: 525-39.
2. Polson AM, Heijl LC. Osseous repair in infrabony periodontal defects.JClin Periodontol1978; 5: 13-23.
3. Froum SJ, Coran M, Thaller B. Kushner L. Scopp TW. Stahl SS.
Periodontal healing following open debridement flap proce- dures. I Clinical assessment of soft tissue and osseous repair.J Periodontol1982; 53: 8-14.
4. Hausmann E. A contemporary perspective on techniques for the clinical assessment of alveolar bone. J Periodontal 1990; 61:
149-56.
5. Henrikson CO, Bergstrom J. Quantitative long-term determina- tions of the alveolar bone mineral mass in man by l~sT
absorptiometry. I Accuracy and precision of the method. Acta Radio11974; 13: 377-92.
6. Jeffcoat MK, Reddy MS, Webber RL, Williams RC. Riittimann VE. Extraoral control of geometry for digital subtraction radio- graphy.J Periodont Res 1987;22: 396-402.
7. Webber RL, Riittimann VE, Grondahl HG. X-ray image subtraction as a basis for assessment of periodontal changes. J Periodont Res 1982; 17: 509-11.
8. Grondahl HG, Grondahl K, Webber RL. A digital subtraction technique for dental radiography. Oral Surg Oral Med Oral Patho11983; 55: 96-102.
9. Vos MH, Janssen PTM, van Aken J, Heethar RM. Quantitative measurement of periodontal bone changes by digital subtraction.
J Periodont Res 1986; 21: 583-91.
10. Ortman LF, Dunford R, McHenry K, Hausmann E. Subtraction radiography and computer assisted densitometric analyses of standardized radiographs. A comparison study with l~sIabsorp- tiometry.J Periodont Res 1985; 20: 644-51.
II. Bragger D, Pasquali L, Rylander H, Carnes D, Kornman KS.
Computer-assisted densitometric image analysis in periodontal
radiography. A methodological study. J Clin Periodontol 1988; IS: 27-37.
12. Dubrez B, Graf JM, Vuagnat P, Cimasoni G. Increase of interproximal bone density after subgingival instrumentation: a quantitative radiographical study.J.Periodontol1990; 61: 725-31.
13. Graf JM, Mounir A, Payot P, Cimasoni G. A simple paralleling instrument for superimposing radiographs of the molar regions.
Oral Surg Oral Med Oral Patho11988; 66: 502-6.
14. Payot P, Haroutunian B, Pochon Y, Herr P, Bickel M, Cimasoni G. Densitometric analysis of lower molar interradicular areas in superimposable radiographs.JClin Periodontol1987; 14: 1-7.
15. Payot P, Bickel M, Cimasoni G. Longitudinal quantitative radiodensitometric study of treated and untreated lower molar furcation involvements.JClin Periodontol1987; 14: 8-18.
16. Cohn SH. Non-invasive Measurements of Bone Mass and their Clinical Application. Boca Raton: CRC Press, 1981: 62-4.
17. Steen WHA, Trouerbach WT, Zwamborn AW, Schouten HJA.
An improved microdensitometric method to investigate inter- dental bone pathology.Dentomaxillofac Radio11985; 14: 53-58.
18. Jacot-Descombes A, Todorov K, Hochstrasser DF, Pellegrini C, Pun T. LaboImage: a workstation environment for research in image processing and analysis.Comput Appl Biosci 1991; 7: 225-
32. .
19. Bragger U. Digital imaging in periodontal radiography. A review.JClin Periodontol1988; IS: 551-7.
Address: Professor G. Cimasoni, Dental School, University of Geneva, 19, rue Barthelemy Menn, CH-1211 Geneva 4, Switzerland.
