d e n t a l m a t e r i a l s 3 1 ( 2 0 1 5 ) 453–461
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.intl.elsevierhealth.com/journals/dema
3D volumetric displacement and strain analysis of composite polymerization Annelies Van Ende a , Elke Van de Casteele a,b , Maarten Depypere c , Jan De Munck a , Xin Li a , Frederik Maes c , Martine Wevers b , Bart Van Meerbeek a,∗ a
KU Leuven – BIOMAT, Department of Oral Health Sciences, KU Leuven (University of Leuven) & Dentistry, University Hospitals Leuven, Leuven, Belgium b Department of Materials Engineering, KU Leuven (University of Leuven), Leuven, Belgium c ESAT – PSI, iMINDS – MIC Medical Image Computing, KU Leuven (University of Leuven), Leuven, Belgium
a r t i c l e
i n f o
a b s t r a c t
Article history:
Objective. The present study aimed at a better understanding of the internal shrinkage
Received 28 November 2014
patterns within different cavity sizes.
Received in revised form
Methods. Ten cylindrical cavities in two sizes were filled with a flowable composite and
23 January 2015
scanned using X-ray micro-computed tomography (-CT) before filling, before and after
Accepted 28 January 2015
polymerization. Three-dimensional (3D) non-rigid image registration was applied to sets of two subsequent -CT images, before and after polymerization in order to calculate the displacements and strains caused by polymerization shrinkage.
Keywords:
Results. 3D volumetric displacement analysis disclosed a main vertical component for both
Composite materials
the small and large cavities, however in the latter the downward direction reversed to an
Micro-computed tomography
upward direction from a depth of approximately 2 mm due to debonding at the bottom. Air
Image registration
bubbles and voids in the restorations increased upon polymerization, causing a reverse in
Shrinkage
strain in the surrounding areas. Significance. Polymerization-induced shrinkage stress in composite restorations cannot be measured directly. This exploratory study revealed more information on cavity-size dependent shrinkage patterns and opens the way to more extensive studies using different composite materials and varying geometric cavity configurations. © 2015 Academy of Dental Materials. Published by Elsevier Ltd. All rights reserved.
1.
Introduction
Polymerization shrinkage is an undesired, yet inherent property of dental composites that causes a discrepancy in dimensions between the restoration as it was placed right before light-curing and its final solidified state after
polymerization. This results in a disparity at the interface, which results in either residual strains at the interface or in gap formation in case the bond strength is exceeded, both affecting the longevity of the restored tooth [1–3]. Shrinkage strains within a resin composite are nonuniformly distributed. First, they are influenced by the complex geometry of a restoration [4] and the quality of the
∗ Corresponding author at: KU Leuven – BIOMAT, Kapucijnenvoer 7 blok a bus 7001, 3000 Leuven, Belgium. Tel.: +32 016 33 75 87; fax: +32 016 33 27 52. E-mail address: [email protected] (B. Van Meerbeek). http://dx.doi.org/10.1016/j.dental.2015.01.018 0109-5641/© 2015 Academy of Dental Materials. Published by Elsevier Ltd. All rights reserved.
454
d e n t a l m a t e r i a l s 3 1 ( 2 0 1 5 ) 453–461
bonding [5]. Second, the curing-light intensity throughout the restoration changes due to attenuation and scattering of the light that enters from the irradiated surface [6]. Consequently, the polymerization rate and degree of conversion vary within the bulk of the restoration and thus composite properties cannot be expected to be homogeneous throughout the restoration bulk. While many studies investigated the overall shrinkage parameters [7–9], only a few studies focused on localized shrinkage and strain. Among the latter studies, for example, digital image correlation (DIC) of CCD-captured images has been used to visualize the regional strains at the surface of the material [10]. Typically, a random speckle pattern is spray-painted on the surface, which provides the surface contrast needed for DIC. When two cameras were used, the out-of-plane strains could also be measured [11]. Recently, a more advanced form of DIC, making use of a fluorescent particle suspension and laser illumination of the specimen, was developed. This enabled particle tracking in real time, thus providing additional information about the shrinkage kinetics. However, this method is limited to registrations at the surface, and has so far only been carried out in 2D on sliced specimens [12]. Lately, high-resolution X-ray computed tomography (-CT) has been introduced to investigate polymerization shrinkage. In addition to surface measurements, this technique can also visualize internal deformations. To visualize the displacement field, Chiang et al. and Cho et al. added spherical particles to the composite and traced their movement before and after polymerization by means of rigid image registration [13,14]. In a more recent study, Takemura et al. intentionally included air bubbles, which were used as tracers [15]. However, the intentional inclusion of air bubbles by mixing seems impractical and can considerably change the properties of the composite [16]. Virtually all restorations contain a certain amount of porosities without intentional mixing, but these porosities are not distributed evenly within the composite and thus may be less suitable as traceable markers. In order to calculate the displacement field, image registration has been used. This is a well-known technique in the biomedical field, for instance to relate information in different images for diagnosis, treatment or basic science [17]. Applications of image registration involve combining images of the same patient from different modalities [18], aligning temporal sequences of images to compensate for motion between scans, and image guidance during interventions. Unlike rigid image registration, non-rigid registration allows some localized scaling in addition to simple transformations such as rotation or translation, this in order to obtain maximal correspondence between images. Hence, distortion of a shape can also be traced. Experimental measurement of the actual orientation of polymerization shrinkage within a composite restoration is indispensable to estimate the deformation that may occur in a realistic clinical setting. Therefore, the purpose of this study was to visualize the local displacements and to quantify and spatially resolve the accompanying volumetric strains that occur in a composite restoration upon its polymerization by means of non-rigid image registration. We tested the hypothesis that cavity size/depth does not influence volumetric
displacement and strain within the composite upon polymerization.
2.
Materials and methods
Ten composite cylinders with a diameter of 4.7 mm and a height of 6.0 mm were prepared from a microhybrid composite (Gradia Direct Anterior, GC, Tokyo, Japan) using a polypropylene mold. This composite was selected because of its radio-translucency, so to not interfere with the X-ray computed tomography (-CT) of the composite test material (see below). The specimens were divided into two groups and standardized cylindrical occlusal cavities with different dimensions were prepared. The small cavities had a diameter of 2.4 mm and 2.0 mm depth, while the large cavities had a diameter of 3.4 mm and 4.0 mm depth. The C-factors of the small and large cavities were 2.2 and 5.8, and the volumes were 9.0 mm3 and 36.3 mm3 , respectively. The cavities were ultrasonically cleaned in distilled water to remove cutting debris. A 10-MDP-containing adhesive (Clearfil SE Bond, Kuraray Noritake, Tokyo, Japan) was applied, without the prior use of a primer and cured for 10 s with a LED curing device (Bluephase 20i, Ivoclar-Vivadent, Schaan, Liechtenstein). The curing device was always used in ‘high’ mode with an output of 1200 mW/cm2 , which was verified prior to each use with the accompanying radiometer (Bluephase meter, Ivoclar-Vivadent). Micro-CT scans were acquired using a nanofocus X-ray computed tomography system (Phoenix nanotom 180, GE Sensing & Inspection Technologies, Wunstorf, Germany), combined with Phoenix Datos|x CT acquisition software. Three scans were taken from each sample. First, the empty cavity was scanned, prior to composite application (‘empty’). Next, the cavity was filled with a flowable composite (G-ænial Flo, GC) and a second scan was taken prior to polymerization (‘uncured’). After polymerization for 40 s with the LED curing device (Bluephase 20i), the third and last scan was taken (‘cured’). The sample was fixed on a rotating stage where the position between source and detector determines the geometrical magnification. In this way the obtained voxel sizes for the small and large cavities were 2.50 m3 and 3.13 m3 , respectively. The X-ray tube has a diamond-tungsten target which generates a broad energy spectrum with a peak voltage set at 70 kVp. The system has a fast scanning mode (frame averaging of 1 and image skip of 0), taking 1200 images with an exposure time of 500 ms leading to a total scan time of 10 min. The projection images were reconstructed in tomographic slices using a modified Feldkamp cone-beam algorithm (NRecon 1.6.5.8, Bruker MicroCT, Kontich, Belgium), creating a 3D dataset. Thereafter, a rigid registration (Euler transformation) was performed to align the three subsequent scans. Although the sample was not removed from the scanner between scans, small movements might have occurred during the filling procedure, as the tip of the composite syringe needs to be in contact with the cavity walls to avoid the inclusion of air bubbles. After the alignment of the scans, the empty cavity was selected and used to accurately define the region of interest (ROI) of the subsequent ‘uncured’ and ‘cured’ scans, so that
455
d e n t a l m a t e r i a l s 3 1 ( 2 0 1 5 ) 453–461
only the cavity content remained, meaning the flowable composite filling. Three-dimensional, non-rigid, intensity-based image registration [18] was then performed between the ROIs in the ‘uncured’ and ‘cured’ scans. Specialized 2D/3D analysis software (CTAn, Bruker MicroCT) and a custom-made software package based on Elastix [19] were used for the ROI determination and registration, respectively. The registration was computed using a B-spline transformation model, with a grid spacing of 250 m, 4 resolution levels, 250 iterations and 2048 spatial samples. Advanced Mattes mutual information [20] was used as similarity metric; for the optimization, an adaptive stochastic gradient descent algorithm was applied. After registration, the regional displacements and principal strains were derived analytically from the B-spline deformation field. For the interpretation, the displacement of the particles within the composite was visualized by means of a 3D vector field. The maximum principal strains were visualized by means of a color spectrum. Descriptive statistics were used to summarize the displacement and strain results. In order to evaluate the integrity of the interface between the cavity and restoration in more detail, samples from both groups were sectioned through their greatest diameter and gold-sputtered for examination by means of scanning electron microscopy (SEM; JEOL JSM-6610/6610LV, Tokyo, Japan) in secondary electron (SE) and backscattered electron (BSE) mode at an accelerating voltage of 15 kV and a spot size of 30. In order to estimate the influence of the cavity depth on light attenuation, two Teflon molds with a diameter of 4 mm and a depth of 2 mm and 4 mm, respectively, were placed on top of a 0.15 mm microscope cover glass (Marienfeld, LaudaKönigshofen, Germany) and filled with the flowable composite (G-ænial Flo) for spectrophotometric analysis of the transmission of the curing light through the composite. The irradiance and the concomitant spectra were measured at the bottom of the specimen with a calibrated, NiST-referenced spectrometer
(USB4000, Ocean Optics, Florida, USA) attached to a sensor with a diameter of 3.9 mm; the resulting delivered energy at the bottom of the cavity was calculated. The specimen diameter of 4 mm was wider than that of both the small and large -CT cavities/specimens (2.4 and 3.4 mm, respectively) in order to fully cover the spectrometer sensor, while being able to measure light attenuation with depth, the latter being the same as that of the -CT cavities (2 and 4 mm, respectively). Because the degree of conversion (DC) cannot be assumed identical at both cavity depths, it might have influenced regional shrinkage. In order to check for possible differences in DC in the small and large cavities, both a small (diameter: 2.4 mm, height: 2 mm) and large (diameter: 3.4 mm, height: 4 mm) sample were prepared and sectioned vertically through their greatest diameter (Isomet, Buehler, Lake Bluff, Illinois) and stored in the dark at room temperature for 24 h. Next, they were analyzed with micro-Raman spectroscopy (Senterra, Bruker Optik, Ettlingen, Germany) using the following parameters: 785 nm wavelength laser with 100 mW power, 50 × 1000 m aperture, 9–15 cm−1 resolution, 50–3500 cm−1 spectral range, integration time of 10 s with 2 co-additions. The laser beam was focused through an ×100 objective lens (BX51, Olympus, London, United Kingdom). The spectrum of the uncured composite was taken as a reference. To calculate DC, the absorbance peak intensities of vinyl C C located at 1640 cm−1 and phenyl C C located at 1610 cm−1 were measured to compare spectra before and after polymerization. The double bond conversion was determined using the following equation: DC(%) =
1−
Rcured Runcured
× 100
where R is the ratio of peak heights at 1640 cm−1 and 1610 cm−1 for the cured and uncured material.
Fig. 1 – Representative displacement vector fields for a small cavity in (a) and a large cavity in (b). Top: vector field as it was generated by the image registration. All vectors were magnified by 3 for visualization purposes. Bottom: a simplified, schematic representation of both respective groups for illustrative purposes.
456
d e n t a l m a t e r i a l s 3 1 ( 2 0 1 5 ) 453–461
Table 1 – Descriptive statistics of displacement (m).
Small cavities Large cavities
Mean Length ± SD
Mean X ± SD
21.65 ± 19.10 54.96 ± 30.32
−1.31 ± 5.50 −8.2 ± 21.80
Mean Y ± SD
Mean Z ± SD
1.71 ± 4.92 −0.0023 ± 16.73
−15.97 ± 22.09 29.07 ± 39.76
SD = standard deviation; X, Y and Z refer to displacement in the X-, Y- and Z-direction, respectively.
3.
Results
All vector fields showed a non-uniform displacement pattern, irrespective of the cavity size (Fig. 1). A clear occlusal intrusion was observed for all samples. The Z-axial displacement component was clearly dominating (Table 1), while only a slight transversal component was observed, generally oriented inwards to the cavity center (Fig. 1). In a few samples, a higher transversal component (X- or Y-direction) indicated asymmetry in lateral detachment, such as a growing void in the cavity bottom corner. However, these displacements were still smaller than the vertical Z-component. Displacements were substantially larger in the large than the small cavities (Table 1). In the small cavities, the displacement was generally oriented toward the bottom (Fig. 1a), and resulted in a mean Z-axis displacement of −15.97 m (Table 1). Close to the bottom, the displacement values were close to zero. In the large cavities, however, the displacement was oriented downwards in the upper 2 mm, but below this depth, the direction flipped to upward (Fig. 1b). Overall, the upward direction in the larger cavities was larger than the downward direction, leading to a mean Z-displacement of 29.07 m. The strain histogram showed a peak strain of around −3.9% for both the small and large cavities (Fig. 2 and Table 2: mode). Although the mean strain in the large cavities was only slightly higher than in the small cavities (Table 2), more extreme values were recorded, not only at the negative, but also at the positive tail of the graph (Fig. 2). This resulted in a significantly higher skewness of the strain histogram for the large cavities (Table 2).
(%) 0.25
0.2
Small cavities Large cavities
0.15
0.01
0.05
-20
-15
-10
-5
0
5
10
15
4.
Discussion
In order to study regional volumetric displacement and strain upon polymerization of composite in a cavity in 3D, -CT was applied to artificial cavities that differed in size and depth. The employed nanofocus X-ray computed tomography system enabled to study volumetric shrinkage effects in situ with a high resolution in terms of -CT; a voxel size of 3.13 m3 or better was achieved. An empty cavity was used to facilitate and objectify the ROI determination. A high resolution and accurate ROI determination improves the accuracy of the measurements, because it enables to detect discontinuities in displacements patterns, such as at the interface and in locations where discontinuities occur, such as air bubbles. For interpretation of the displacement and strain data, the transmission of light energy through composite and the resultant DC were measured as well. Finally, bond integrity was examined using SEM, since gaps of a smaller extent than the voxel size cannot
20
Strain (%)
Fig. 2 – Strain histogram. Both cavities showed a peak strain of around −3.9%, representing a pre-dominant regional shrinkage, while both cavities also revealed a small percentage of positive strains, representing regional expansion.
Regional differences in strain were disclosed for both the small (Fig. 3) and large cavities (Fig. 4). Negative strain expresses the decrease in distance between two reference points, and thus denotes regional shrinkage, as it was caused by polymerization. Positive strain, consequently, refers to expansion or zones of stress relaxation, such as shown around voids (Figs. 3 and 4) and interface de-bonding (Fig. 4). No interfacial gaps were imaged by SEM in the small composite-filled cavities (Fig. 5a), while the half-sectioned large cavities disclosed bond detachment at the cavity bottom (Fig. 5b). Spectrophotometric analysis of light transmission through composite revealed that the adopted curing protocol delivered a significantly higher irradiation energy of 12 J/cm2 at the bottom of the small 2 mm deep filling (or 23.5% of the irradiation energy at the surface, being 51 J/cm2 ), versus only 1.4 J/cm2 at the bottom of the 4 mm deep filling (or 2.75% of the surface irradiation energy). Finally, mean DC, being 62.08 ± 1.96% and 61.67 ± 2.93% for the small and the large composite specimen, respectively, did not differ substantially (Fig. 5: bottom). The bottom-to-top conversion ratio (DCbottom /DCtop ) was only slightly higher for the small cavity (0.98) than for the large cavity (0.91).
Table 2 – Descriptive statistics of strain.
Small cavities Large cavities
Mean ± SD
Skewness ± SD
Mode ± SD
−2.97 ± 0.48 −3.11 ± 0.37
0.67 ± 0.72 3.38 ± 1.38
−3.90 ± 0.42 −3.93 ± 0.30
SD = standard deviation.
d e n t a l m a t e r i a l s 3 1 ( 2 0 1 5 ) 453–461
457
Fig. 3 – Collage of -CT images and reconstructions representing a small cavity. (a) Reconstructed slices of an uncured (left) and cured sample (right). Upon curing, the entrapped air bubble (white arrow) had slightly moved downwards and subtly increased in size. The occlusal surface noticeably intruded (open white arrows). (b) Color-coded strain map of the same section (left) and the 3D rendered model (right). Absence of strain, representing a zone of relaxation, can be detected immediately around the air bubble with a gradually increasing negative strain (shrinkage) further away. Small areas revealing regionally reduced strains can be detected (orange-colored globules), this in particular also along the right cavity margin, most likely representing other areas of void entrapment and/or de-bonding. Very striking is the bottom area, revealing (green-colored) positive strain, most likely representing a compensating zone between the bonding and the negative strain due to the inherent shrinkage. (c) Strain map of the surface of the restoration bottom (left) and restoration top (right). Except for the center part and the outer margin, both revealing negative strain, the main part of the bottom surface revealed (yellow-colored) absence of strain to (green-colored) positive strain. At the top surface, negative strain was detected for the major part, with a clear zone of reduced strain evolving into a relaxation zone of (green-colored) positive strain toward the right cavity side. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
be detected by -CT. Overall, a clear difference in the orientation of the displacement vectors was detected between the two cavity sizes, by which the hypothesis was rejected. Micro-CT makes use of the differences in X-ray attenuation to create a visible image of the specimen represented by different shades of gray or brightness levels. The image registration
software takes the differences in attenuation into account in the similarity measure between the uncured and cured situation. As a consequence, a certain level of heterogeneity with sufficiently contrasting elements inside the composite furthers an accurate registration. Although composites are constitutes from different materials, the radiopaque and
458
d e n t a l m a t e r i a l s 3 1 ( 2 0 1 5 ) 453–461
Fig. 4 – Collage of -CT images and reconstructions representing a large cavity. (a) Reconstructed slices of an uncured sample. No air bubbles can be detected in this section. (b) Reconstructed slices after curing, revealing an entrapped air bubble within the composite bulk, but also near the cavity corner (white arrows). Bonding detachment can be seen at the bottom (open white arrows). (c) Color-coded strain map of the same section. Negative strains were detected within the main restoration part, with even some (royal blue) areas (globules) exhibiting very high negative strains. The voids near the cavity border resulted in a large zone of (turquoise) very high positive strain due to void expansion, gradually decreasing more remote from the void. Likewise, the interface de-bonding resulted in a relaxation zone of positive strain in the surrounding area. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
radiolucent elements should be of sufficient size and/or distance from each other in relation to the spatial resolution to meet this condition. The flowable composite that was used in this study had sufficient contrast without any addition. However, it is possible that other composites might need the addition of radiopacifiers to obtain the same accuracy, such as glass fillers [13] or iodine-doped resins [21]. Unfortunately, most adhesives are rather radiolucent, which hampers the interpretation of gap occurrence and precise gap location with regard to the adhesive layer, especially if gaps were already present before the curing of the composite due to air
entrapment rather than shrinkage. In this study, no modifications were made to the existing materials, but zirconia-doping adhesives could facilitate the interpretation of gaps occurring at the tooth–adhesive–resin interface, as was shown in a recent study [22]. We opted in this exploratory study for a simplified yet more uniform design with cavities prepared in small composite cylinders rather than natural teeth. Cylindrical shapes of a more homogeneous and less radiopaque substrate are less subjected to scanning artifacts and their relatively small size made it feasible to scan at high resolution in accordance with
d e n t a l m a t e r i a l s 3 1 ( 2 0 1 5 ) 453–461
459
Fig. 5 – BSE images (top) and color-coded micro-Raman conversion mapping (bottom) of a representative small (a, left) and a large composite-filled cavity (b, right). A tight bond without de-bonding was observed in the small cavity contrary to clear de-bonding and gap formation in the large cavity. The degree of conversion was rather homogeneous for both cavities throughout the whole depth.
a voxel size of respectively 2.5 and 3.125 mm3 for the small and large cavities. This resolution is much higher than in most -CT studies. Moreover, using the ‘fast-scan’ mode of the scanner, it was possible to reduce the scan time to 10 min, limiting the risk on movement artifacts and still rendering images of adequate quality with an acceptable minimal amount of noise. In addition, it is sensible to keep the scan time as short as possible to avoid effects of spontaneous polymerization after brief exposure to ambient light. Another reason to opt for a composite substrate was to establish a more uniform bonding condition. As bonding to enamel is far better established than bonding to dentin [23], the dentin bond is more likely to fail due to the stress development and the shrinkage may thus be directed by the stronger enamel bond. This would have complicated the interpretation of the shrinkage directions. In this study, hence explanation cannot be sought in the weaker bonding to deep dentin. So far, only three studies have evaluated the internal, volumetric deformations of light-cured composite restorations in 3D, two by adding spherical filler particles as markers [13,14]. Takemura et al. intentionally created air bubbles within the composite, which were traced in a same manner [15]. In our study, we made use of the filler particles present in the flowable composite, by which any effect of adding other particles on the material properties was excluded. The displacement pattern found in the large cavities in this study is very similar to the pattern in the ‘bonded’ specimens
of the study by Cho et al., when cavities were also prepared in composite [14]. In that study, a rigid registration method and a clustering algorithm to trace the filler particles and to perform a pairing operation before and after polymerization were used. In the second study of Chiang et al., a similar set-up was used; however, more and different patterns were found, this very likely due to the fact that they used natural teeth [13]. Shrinkage was directed toward the bottom of the small cavity. The pattern disclosed in this study using -CT closely resembles the pattern described by Versluis et al., who applied finite element analysis to perfectly bonded composite restorations [24]. Shrinkage direction was concluded to be predominantly the result of boundary conditions, while it was only slightly affected by the relation toward the light. In the large cavities, a twofold displacement pattern was recorded with a downward and upward shrinkage direction for respectively the upper and bottom 2 mm composite. The latter upward displacement caused the composite to de-bond from the cavity bottom, as was confirmed by SEM. In previous studies, de-bonding within Class-I cavities often occurred at the bottom of the cavity [25,26]. This has often been attributed to the weaker bond to dentin than to enamel, which makes it more likely for the composite to detach in the deeper dentin regions [24]. However, this assumption is not valid in our study, since we intentionally used a uniform composite substrate to rule out the inherent difference in bonding receptiveness of enamel versus dentin. Even more, as the cavities in our
460
d e n t a l m a t e r i a l s 3 1 ( 2 0 1 5 ) 453–461
study were cut in freshly polymerized composite cylinders, the bonding effectiveness of the flowable composite was not only equal to the whole cavity wall and bottom, but most likely even higher than it would have been to enamel and dentin as the bond was formed by chemical co-polymerization via the adhesive in addition to micro-mechanical interlocking in the surface irregularities created by bur. Because the volume of the cylindric substrate was kept constant, the composite walls in the larger cavities were thinner and are thus more compliant; however, due to the low e-modulus of the composite substrate, the strain distribution is not expected to be very different in terms of its pattern, especially since the dominant movement occurred in the vertical direction. Moreover, strains perpendicular to the vertical cavity walls are more evenly distributed in a circular structure, as compared with more traditionally used rectangular cavity shapes, where one would expect stress peaks to occur in the corners [27]. Hence, in this study, the highest shrinkage stress upon polymerization was subjected to the cavity bottom of the large cavities, as can be deducted from the strain concentration in this area and the de-bonding. Despite only 2.75% irradiation energy delivered by the curing light could reach the bottom of the cavity of the 4 mm deep composite specimens, the bottom-to-top conversion ratio was, conspicuously, satisfactory even for the 4 mm deep larger specimens (0.91). This must be attributed to the combined effect of the high light output (1200 mW/cm2 ) of the curing device employed and the according to current clinical standards relatively extended curing time of 40 s. Compared to the 2 mm deep composite specimens, DC was only slightly lower for the 4 mm deep composite specimens, so that the somewhat reduced shrinkage appeared insufficient to prevent interface de-bonding in the deeper cavity region. However, a delay in conversion rate and differences in cross-linking [28] cannot be ruled out. It should be noted that the manufacturer (GC) does not recommend curing the flowable composite in such a thick 4 mm increment. Although not measured, it is known that the gel point will be delayed in resin composite at greater distances from the irradiated surface due to the decreased local intensity of the attenuated light with increasing depth [29]. The resin matrix will be sooner immobilized in the upper region than in the slower polymerizing bottom region. As a consequence, the later occurring displacements were oriented opposite to that of the upper 2 mm composite part and toward this already immobilized top part, because the stiffer part could now exert a stronger pull. This explains the twofold displacement pattern recorded for the 4 mm deep composite specimens. This is in accordance with a recent study [30], in which was demonstrated that the polymerization process in conventional composites was significantly delayed from a depth of 2 mm. In the current study, we could not only visualize the displacement vectors but also studied the displacement gradients, or in other words, the strain. A clear peak strain was found around −3.9% for both cavities. However, it must be taken into account that only a small part of the total shrinkage will contribute to shrinkage stress development, as the free shrinkage was also monitored. This value corresponds relatively well to the composite’s volumetric shrinkage of 4.38%,
as was provided by the manufacturer (GC), and thus confirms the adequate accuracy of our registration. The mean strain, however, was slightly lower (Table 2: around 3% for both cavities), because of local variety in strain due to the restriction of composite flow at regions where it adhered to the cavity wall. The strain showed a non-uniform pattern in all cavities. A composite is per definition a heterogenous material and on the micro-scale of its constituents, it will behave as such. Local heterogeneities, such as prepolymerized or inorganic particles within the resin matrix may all have contributed to the ‘blotted’ strain pattern throughout the whole restoration (Figs. 3 and 4). Furthermore, local flaws like entrapped air bubbles and interface de-bonding influenced the regional strain distribution (Figs. 3b, c and 4c). Voids increased in volume upon polymerization, which attributes to the free surface allowing the composite to shrink without building up shrinkage stress. The yellow-colored (thus close to zero) strain at the edge of the entrapped air bubble (Fig. 3b) illustrates the strain relief; further away strain gradually became more negative, indicating higher shrinkage. In the zone of interfacial debonding, the green-turquoise colors indicate positive strain, caused by the expansion of the void. These areas of expansion may compensate for the surrounding shrinkage, and thus regionally relieve shrinkage stress, as previously suggested by Alster et al. [16]. On the other hand, such small local deficiencies might be accountable for an increased susceptibility to mechanical fatigue degradation and a loss of cohesive strength in the long term. In the recent study of Takemura et al., where air bubbles were used as traceable markers, no mention has been made of deformation of the bubbles upon curing [15]. It is possible that due to the lower resolution achieved in that study, these regional effects got unnoticed. In conclusion, the proposed digital volume correlation using a non-rigid image registration enabled to measure and visualize the regional shrinkage strain vectors within the entire volume of the composite restoration. Experimental determination of directional flow and shrinkage is important to study the significance of polymerization contraction patterns in relation to marginal gap formation and stress distribution. This setup can also be used with natural teeth and for any shape, which makes it interesting to study more complex geometrical cavity designs. Moreover, internal regional flaws, such as air bubbles or cracking within the bulk, are often overlooked in other methods such as mathematical models or surface-based calculations. This may give important insight in how such flaws could relate to the initiation of de-bonding and provide a more accurate and realistic insight in the effect of shrinkage in real clinical situations.
Acknowledgements We would like to thank GC and Kuraray for providing the materials used in this study, and the Hercules Foundation for funding the project AKUL/09/001 “Micro- and nano-CT for the hierarchical analysis of materials”. A. Van Ende has been granted a PhD fellowship from the Research Foundation – Flanders (FWO). We declare no potential conflicts of interest with respect to the authorship and/or publication of this article.
d e n t a l m a t e r i a l s 3 1 ( 2 0 1 5 ) 453–461
references
[1] Davidson CL, De Gee A, Feilzer JA. The competition between the composite-dentin bond strength and the polymerization contraction stress. J Dent Res 1984;63:1396–9, http://dx.doi.org/10.1177/00220345840630121101. [2] Bergenholtz G. Evidence for bacterial causation of adverse pulpal responses in resin-based dental restorations. Crit Rev Oral Biol Med 2000;11:467–80, http://dx.doi.org/10.1177/10454411000110040501. [3] Braga RR, Ballester RY, Ferracane JL. Factors involved in the development of polymerization shrinkage stress in resin-composites: a systematic review. Dent Mater 2005;21:962–70, http://dx.doi.org/10.1016/j.dental.2005.04.018. [4] Boaro LCC, Brandt WC, Meira JBC, Rodrigues FP, Palin WM, Braga RR. Experimental and FE displacement and polymerization stress of bonded restorations as a function of the C-factor, volume and substrate stiffness. J Dent 2014;42:140–8, http://dx.doi.org/10.1016/j.jdent.2013.11.016. [5] Versluis A, Tantbirojn D, Pintado MR, DeLong R, Douglas WH. Residual shrinkage stress distributions in molars after composite restoration. Dent Mater 2004;20:554–64, http://dx.doi.org/10.1016/j.dental.2003.05.007. [6] Shortall AC, Palin WM, Burtscher P. Refractive index mismatch and monomer reactivity influence composite curing depth. J Dent Res 2008;87:84–8, http://dx.doi.org/10.1177/154405910808700115. [7] Fleming GJP, Khan S, Afzal O, Palin WM, Burke FJT. Investigation of polymerisation shrinkage strain, associated cuspal movement and microleakage of MOD cavities restored incrementally with resin-based composite using an LED light curing unit. J Dent 2007;35:97–103, http://dx.doi.org/10.1016/j.jdent.2006.05.003. [8] Satterthwaite JD, Maisuria A, Vogel K, Watts DC. Effect of resin-composite filler particle size and shape on shrinkage-stress. Dent Mater 2012;28:609–14, http://dx.doi.org/10.1016/j.dental.2012.01.007. [9] Schneider LFJ, Cavalcante LM, Silikas N. Shrinkage stresses generated during resin-composite applications: a review. J Dent Biomech 2010:2010, http://dx.doi.org/10.4061/2010/131630. [10] Li J, Fok AS-L, Satterthwaite J, Watts DC. Measurement of the full-field polymerization shrinkage and depth of cure of dental composites using digital image correlation. Dent Mater 2009;25:582–8, http://dx.doi.org/10.1016/j.dental.2008.11.001. [11] Miletic V, Manojlovic D, Milosevic M, Mitrovic N, Stankovic TS, Maneski T. Analysis of local shrinkage patterns of self-adhering and flowable composites using 3D digital image correlation. Quintessence Int (Berl) 2011;42: 797–804. [12] Kweon H-J, Ferracane JL, Kang K, Dhont J, Lee I-B. Spatio-temporal analysis of shrinkage vectors during photo-polymerization of composite. Dent Mater 2013;29:1236–43, http://dx.doi.org/10.1016/j.dental.2013.09.011. [13] Chiang Y-C, Rösch P, Dabanoglu A, Lin C-P, Hickel R, Kunzelmann K-H. Polymerization composite shrinkage evaluation with 3D deformation analysis from microCT images. Dent Mater 2010;26:223–31, http://dx.doi.org/10.1016/j.dental.2009.09.013. [14] Cho E, Sadr A, Inai N, Tagami J. Evaluation of resin composite polymerization by three dimensional micro-CT imaging and nanoindentation. Dent Mater 2011;27:1070–8, http://dx.doi.org/10.1016/j.dental.2011.07.008.
461
[15] Takemura Y, Hanaoka K, Kawamata R, Sakurai T, Teranaka T. Three-dimensional X-ray micro-computed tomography analysis of polymerization shrinkage vectors in flowable composite. Dent Mater J 2014;33:476–83, http://dx.doi.org/10.4012/dmj.2013-288. [16] Alster D, Feilzer A, De Gee JA, Moli J, Davidson ACL. The dependence of shrinkage stress reduction on porosity concentration in thin resin layers. J Dent Res 1992;71:1619–22, http://dx.doi.org/10.1177/00220345920710091401. [17] Oliveira FPM, Tavares JMRS. Medical image registration: a review. Comput Methods Biomech Biomed Eng 2014;17:73–93, http://dx.doi.org/10.1080/10255842.2012.670855. [18] Maes F, Collignon A, Vandermeulen D, Marchal G, Suetens P. Multimodality image registration by maximization of mutual information. IEEE Trans Med Imaging 1997;16:187–98, http://dx.doi.org/10.1109/42.563664. [19] Klein S, Staring M, Murphy K, Viergever MA, Pluim JPW. Elastix: a toolbox for intensity-based medical image registration. IEEE Trans Med Imaging 2010;29:196–205, http://dx.doi.org/10.1109/TMI.2009.2035616. [20] Mattes D, Haynor DR, Vesselle H, Lewellen TK, Eubank W. PET-CT image registration in the chest using free-form deformations. IEEE Trans Med Imaging 2003;22:120–8, http://dx.doi.org/10.1109/TMI.2003.809072. [21] Zhao Y, Lan J, Wang X, Deng X, Cai Q, Yang X. Synthesis of iodine-containing cyclophosphazenes for using as radiopacifiers in dental composite resin. Mater Sci Eng C Mater Biol Appl 2014;43:432–8, http://dx.doi.org/10.1016/j.msec.2014.07.050. [22] Rominu M, Manescu A, Sinescu C, Negrutiu ML, Topala F, Rominu RO, et al. Zirconia enriched dental adhesive: a solution for OCT contrast enhancement. Demonstrative study by synchrotron radiation microtomography. Dent Mater 2014;30:417–23, http://dx.doi.org/10.1016/j.dental.2014.01.004. [23] Van Meerbeek B, Peumans M, Poitevin A, Mine A, Van Ende A, Neves A, et al. Relationship between bond-strength tests and clinical outcomes. Dent Mater 2010;26:e100–21, http://dx.doi.org/10.1016/j.dental.2009.11.148. [24] Versluis A, Tantbirojn D, Douglas WH. Do dental composites always shrink toward the light? J Dent Res 1998;77:1435–45, http://dx.doi.org/10.1177/00220345980770060801. [25] Furness A, Tadros MY, Looney SW, Rueggeberg FA. Effect of bulk/incremental fill on internal gap formation of bulk-fill composites. J Dent 2014;42:439–49, http://dx.doi.org/10.1016/j.jdent.2014.01.005. [26] Kim HJ, Park SH. Measurement of the internal adaptation of resin composites using micro-CT and its correlation with polymerization shrinkage. Oper Dent 2014;39:E57–70, http://dx.doi.org/10.2341/12-378-L. [27] Rodrigues FP, Silikas N, Watts DC, Ballester RY. Finite element analysis of bonded model Class I “restorations” after shrinkage. Dent Mater 2012;28:123–32, http://dx.doi.org/10.1016/j.dental.2011.10.001. [28] Asmussen E, Peutzfeldt A. Short communication: polymer structure of a light-cured resin composite in relation to distance from the surface. Eur J Oral Sci 2003;111:277–9. [29] Martinsen M, El-Hajjar RF, Berzins DW. 3D full field strain analysis of polymerization shrinkage in a dental composite. Dent Mater 2013;29:e161–7, http://dx.doi.org/10.1016/j.dental.2013.04.019. [30] Steinhaus J, Moeginger B, Grossgarten M, Rosentritt M, Hausnerova B. Dielectric analysis of depth dependent curing behavior of dental resin composites. Dent Mater 2014;30:679–87, http://dx.doi.org/10.1016/j.dental.2014.03. 002.
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.intl.elsevierhealth.com/journals/dema
3D volumetric displacement and strain analysis of composite polymerization Annelies Van Ende a , Elke Van de Casteele a,b , Maarten Depypere c , Jan De Munck a , Xin Li a , Frederik Maes c , Martine Wevers b , Bart Van Meerbeek a,∗ a
KU Leuven – BIOMAT, Department of Oral Health Sciences, KU Leuven (University of Leuven) & Dentistry, University Hospitals Leuven, Leuven, Belgium b Department of Materials Engineering, KU Leuven (University of Leuven), Leuven, Belgium c ESAT – PSI, iMINDS – MIC Medical Image Computing, KU Leuven (University of Leuven), Leuven, Belgium
a r t i c l e
i n f o
a b s t r a c t
Article history:
Objective. The present study aimed at a better understanding of the internal shrinkage
Received 28 November 2014
patterns within different cavity sizes.
Received in revised form
Methods. Ten cylindrical cavities in two sizes were filled with a flowable composite and
23 January 2015
scanned using X-ray micro-computed tomography (-CT) before filling, before and after
Accepted 28 January 2015
polymerization. Three-dimensional (3D) non-rigid image registration was applied to sets of two subsequent -CT images, before and after polymerization in order to calculate the displacements and strains caused by polymerization shrinkage.
Keywords:
Results. 3D volumetric displacement analysis disclosed a main vertical component for both
Composite materials
the small and large cavities, however in the latter the downward direction reversed to an
Micro-computed tomography
upward direction from a depth of approximately 2 mm due to debonding at the bottom. Air
Image registration
bubbles and voids in the restorations increased upon polymerization, causing a reverse in
Shrinkage
strain in the surrounding areas. Significance. Polymerization-induced shrinkage stress in composite restorations cannot be measured directly. This exploratory study revealed more information on cavity-size dependent shrinkage patterns and opens the way to more extensive studies using different composite materials and varying geometric cavity configurations. © 2015 Academy of Dental Materials. Published by Elsevier Ltd. All rights reserved.
1.
Introduction
Polymerization shrinkage is an undesired, yet inherent property of dental composites that causes a discrepancy in dimensions between the restoration as it was placed right before light-curing and its final solidified state after
polymerization. This results in a disparity at the interface, which results in either residual strains at the interface or in gap formation in case the bond strength is exceeded, both affecting the longevity of the restored tooth [1–3]. Shrinkage strains within a resin composite are nonuniformly distributed. First, they are influenced by the complex geometry of a restoration [4] and the quality of the
∗ Corresponding author at: KU Leuven – BIOMAT, Kapucijnenvoer 7 blok a bus 7001, 3000 Leuven, Belgium. Tel.: +32 016 33 75 87; fax: +32 016 33 27 52. E-mail address: [email protected] (B. Van Meerbeek). http://dx.doi.org/10.1016/j.dental.2015.01.018 0109-5641/© 2015 Academy of Dental Materials. Published by Elsevier Ltd. All rights reserved.
454
d e n t a l m a t e r i a l s 3 1 ( 2 0 1 5 ) 453–461
bonding [5]. Second, the curing-light intensity throughout the restoration changes due to attenuation and scattering of the light that enters from the irradiated surface [6]. Consequently, the polymerization rate and degree of conversion vary within the bulk of the restoration and thus composite properties cannot be expected to be homogeneous throughout the restoration bulk. While many studies investigated the overall shrinkage parameters [7–9], only a few studies focused on localized shrinkage and strain. Among the latter studies, for example, digital image correlation (DIC) of CCD-captured images has been used to visualize the regional strains at the surface of the material [10]. Typically, a random speckle pattern is spray-painted on the surface, which provides the surface contrast needed for DIC. When two cameras were used, the out-of-plane strains could also be measured [11]. Recently, a more advanced form of DIC, making use of a fluorescent particle suspension and laser illumination of the specimen, was developed. This enabled particle tracking in real time, thus providing additional information about the shrinkage kinetics. However, this method is limited to registrations at the surface, and has so far only been carried out in 2D on sliced specimens [12]. Lately, high-resolution X-ray computed tomography (-CT) has been introduced to investigate polymerization shrinkage. In addition to surface measurements, this technique can also visualize internal deformations. To visualize the displacement field, Chiang et al. and Cho et al. added spherical particles to the composite and traced their movement before and after polymerization by means of rigid image registration [13,14]. In a more recent study, Takemura et al. intentionally included air bubbles, which were used as tracers [15]. However, the intentional inclusion of air bubbles by mixing seems impractical and can considerably change the properties of the composite [16]. Virtually all restorations contain a certain amount of porosities without intentional mixing, but these porosities are not distributed evenly within the composite and thus may be less suitable as traceable markers. In order to calculate the displacement field, image registration has been used. This is a well-known technique in the biomedical field, for instance to relate information in different images for diagnosis, treatment or basic science [17]. Applications of image registration involve combining images of the same patient from different modalities [18], aligning temporal sequences of images to compensate for motion between scans, and image guidance during interventions. Unlike rigid image registration, non-rigid registration allows some localized scaling in addition to simple transformations such as rotation or translation, this in order to obtain maximal correspondence between images. Hence, distortion of a shape can also be traced. Experimental measurement of the actual orientation of polymerization shrinkage within a composite restoration is indispensable to estimate the deformation that may occur in a realistic clinical setting. Therefore, the purpose of this study was to visualize the local displacements and to quantify and spatially resolve the accompanying volumetric strains that occur in a composite restoration upon its polymerization by means of non-rigid image registration. We tested the hypothesis that cavity size/depth does not influence volumetric
displacement and strain within the composite upon polymerization.
2.
Materials and methods
Ten composite cylinders with a diameter of 4.7 mm and a height of 6.0 mm were prepared from a microhybrid composite (Gradia Direct Anterior, GC, Tokyo, Japan) using a polypropylene mold. This composite was selected because of its radio-translucency, so to not interfere with the X-ray computed tomography (-CT) of the composite test material (see below). The specimens were divided into two groups and standardized cylindrical occlusal cavities with different dimensions were prepared. The small cavities had a diameter of 2.4 mm and 2.0 mm depth, while the large cavities had a diameter of 3.4 mm and 4.0 mm depth. The C-factors of the small and large cavities were 2.2 and 5.8, and the volumes were 9.0 mm3 and 36.3 mm3 , respectively. The cavities were ultrasonically cleaned in distilled water to remove cutting debris. A 10-MDP-containing adhesive (Clearfil SE Bond, Kuraray Noritake, Tokyo, Japan) was applied, without the prior use of a primer and cured for 10 s with a LED curing device (Bluephase 20i, Ivoclar-Vivadent, Schaan, Liechtenstein). The curing device was always used in ‘high’ mode with an output of 1200 mW/cm2 , which was verified prior to each use with the accompanying radiometer (Bluephase meter, Ivoclar-Vivadent). Micro-CT scans were acquired using a nanofocus X-ray computed tomography system (Phoenix nanotom 180, GE Sensing & Inspection Technologies, Wunstorf, Germany), combined with Phoenix Datos|x CT acquisition software. Three scans were taken from each sample. First, the empty cavity was scanned, prior to composite application (‘empty’). Next, the cavity was filled with a flowable composite (G-ænial Flo, GC) and a second scan was taken prior to polymerization (‘uncured’). After polymerization for 40 s with the LED curing device (Bluephase 20i), the third and last scan was taken (‘cured’). The sample was fixed on a rotating stage where the position between source and detector determines the geometrical magnification. In this way the obtained voxel sizes for the small and large cavities were 2.50 m3 and 3.13 m3 , respectively. The X-ray tube has a diamond-tungsten target which generates a broad energy spectrum with a peak voltage set at 70 kVp. The system has a fast scanning mode (frame averaging of 1 and image skip of 0), taking 1200 images with an exposure time of 500 ms leading to a total scan time of 10 min. The projection images were reconstructed in tomographic slices using a modified Feldkamp cone-beam algorithm (NRecon 1.6.5.8, Bruker MicroCT, Kontich, Belgium), creating a 3D dataset. Thereafter, a rigid registration (Euler transformation) was performed to align the three subsequent scans. Although the sample was not removed from the scanner between scans, small movements might have occurred during the filling procedure, as the tip of the composite syringe needs to be in contact with the cavity walls to avoid the inclusion of air bubbles. After the alignment of the scans, the empty cavity was selected and used to accurately define the region of interest (ROI) of the subsequent ‘uncured’ and ‘cured’ scans, so that
455
d e n t a l m a t e r i a l s 3 1 ( 2 0 1 5 ) 453–461
only the cavity content remained, meaning the flowable composite filling. Three-dimensional, non-rigid, intensity-based image registration [18] was then performed between the ROIs in the ‘uncured’ and ‘cured’ scans. Specialized 2D/3D analysis software (CTAn, Bruker MicroCT) and a custom-made software package based on Elastix [19] were used for the ROI determination and registration, respectively. The registration was computed using a B-spline transformation model, with a grid spacing of 250 m, 4 resolution levels, 250 iterations and 2048 spatial samples. Advanced Mattes mutual information [20] was used as similarity metric; for the optimization, an adaptive stochastic gradient descent algorithm was applied. After registration, the regional displacements and principal strains were derived analytically from the B-spline deformation field. For the interpretation, the displacement of the particles within the composite was visualized by means of a 3D vector field. The maximum principal strains were visualized by means of a color spectrum. Descriptive statistics were used to summarize the displacement and strain results. In order to evaluate the integrity of the interface between the cavity and restoration in more detail, samples from both groups were sectioned through their greatest diameter and gold-sputtered for examination by means of scanning electron microscopy (SEM; JEOL JSM-6610/6610LV, Tokyo, Japan) in secondary electron (SE) and backscattered electron (BSE) mode at an accelerating voltage of 15 kV and a spot size of 30. In order to estimate the influence of the cavity depth on light attenuation, two Teflon molds with a diameter of 4 mm and a depth of 2 mm and 4 mm, respectively, were placed on top of a 0.15 mm microscope cover glass (Marienfeld, LaudaKönigshofen, Germany) and filled with the flowable composite (G-ænial Flo) for spectrophotometric analysis of the transmission of the curing light through the composite. The irradiance and the concomitant spectra were measured at the bottom of the specimen with a calibrated, NiST-referenced spectrometer
(USB4000, Ocean Optics, Florida, USA) attached to a sensor with a diameter of 3.9 mm; the resulting delivered energy at the bottom of the cavity was calculated. The specimen diameter of 4 mm was wider than that of both the small and large -CT cavities/specimens (2.4 and 3.4 mm, respectively) in order to fully cover the spectrometer sensor, while being able to measure light attenuation with depth, the latter being the same as that of the -CT cavities (2 and 4 mm, respectively). Because the degree of conversion (DC) cannot be assumed identical at both cavity depths, it might have influenced regional shrinkage. In order to check for possible differences in DC in the small and large cavities, both a small (diameter: 2.4 mm, height: 2 mm) and large (diameter: 3.4 mm, height: 4 mm) sample were prepared and sectioned vertically through their greatest diameter (Isomet, Buehler, Lake Bluff, Illinois) and stored in the dark at room temperature for 24 h. Next, they were analyzed with micro-Raman spectroscopy (Senterra, Bruker Optik, Ettlingen, Germany) using the following parameters: 785 nm wavelength laser with 100 mW power, 50 × 1000 m aperture, 9–15 cm−1 resolution, 50–3500 cm−1 spectral range, integration time of 10 s with 2 co-additions. The laser beam was focused through an ×100 objective lens (BX51, Olympus, London, United Kingdom). The spectrum of the uncured composite was taken as a reference. To calculate DC, the absorbance peak intensities of vinyl C C located at 1640 cm−1 and phenyl C C located at 1610 cm−1 were measured to compare spectra before and after polymerization. The double bond conversion was determined using the following equation: DC(%) =
1−
Rcured Runcured
× 100
where R is the ratio of peak heights at 1640 cm−1 and 1610 cm−1 for the cured and uncured material.
Fig. 1 – Representative displacement vector fields for a small cavity in (a) and a large cavity in (b). Top: vector field as it was generated by the image registration. All vectors were magnified by 3 for visualization purposes. Bottom: a simplified, schematic representation of both respective groups for illustrative purposes.
456
d e n t a l m a t e r i a l s 3 1 ( 2 0 1 5 ) 453–461
Table 1 – Descriptive statistics of displacement (m).
Small cavities Large cavities
Mean Length ± SD
Mean X ± SD
21.65 ± 19.10 54.96 ± 30.32
−1.31 ± 5.50 −8.2 ± 21.80
Mean Y ± SD
Mean Z ± SD
1.71 ± 4.92 −0.0023 ± 16.73
−15.97 ± 22.09 29.07 ± 39.76
SD = standard deviation; X, Y and Z refer to displacement in the X-, Y- and Z-direction, respectively.
3.
Results
All vector fields showed a non-uniform displacement pattern, irrespective of the cavity size (Fig. 1). A clear occlusal intrusion was observed for all samples. The Z-axial displacement component was clearly dominating (Table 1), while only a slight transversal component was observed, generally oriented inwards to the cavity center (Fig. 1). In a few samples, a higher transversal component (X- or Y-direction) indicated asymmetry in lateral detachment, such as a growing void in the cavity bottom corner. However, these displacements were still smaller than the vertical Z-component. Displacements were substantially larger in the large than the small cavities (Table 1). In the small cavities, the displacement was generally oriented toward the bottom (Fig. 1a), and resulted in a mean Z-axis displacement of −15.97 m (Table 1). Close to the bottom, the displacement values were close to zero. In the large cavities, however, the displacement was oriented downwards in the upper 2 mm, but below this depth, the direction flipped to upward (Fig. 1b). Overall, the upward direction in the larger cavities was larger than the downward direction, leading to a mean Z-displacement of 29.07 m. The strain histogram showed a peak strain of around −3.9% for both the small and large cavities (Fig. 2 and Table 2: mode). Although the mean strain in the large cavities was only slightly higher than in the small cavities (Table 2), more extreme values were recorded, not only at the negative, but also at the positive tail of the graph (Fig. 2). This resulted in a significantly higher skewness of the strain histogram for the large cavities (Table 2).
(%) 0.25
0.2
Small cavities Large cavities
0.15
0.01
0.05
-20
-15
-10
-5
0
5
10
15
4.
Discussion
In order to study regional volumetric displacement and strain upon polymerization of composite in a cavity in 3D, -CT was applied to artificial cavities that differed in size and depth. The employed nanofocus X-ray computed tomography system enabled to study volumetric shrinkage effects in situ with a high resolution in terms of -CT; a voxel size of 3.13 m3 or better was achieved. An empty cavity was used to facilitate and objectify the ROI determination. A high resolution and accurate ROI determination improves the accuracy of the measurements, because it enables to detect discontinuities in displacements patterns, such as at the interface and in locations where discontinuities occur, such as air bubbles. For interpretation of the displacement and strain data, the transmission of light energy through composite and the resultant DC were measured as well. Finally, bond integrity was examined using SEM, since gaps of a smaller extent than the voxel size cannot
20
Strain (%)
Fig. 2 – Strain histogram. Both cavities showed a peak strain of around −3.9%, representing a pre-dominant regional shrinkage, while both cavities also revealed a small percentage of positive strains, representing regional expansion.
Regional differences in strain were disclosed for both the small (Fig. 3) and large cavities (Fig. 4). Negative strain expresses the decrease in distance between two reference points, and thus denotes regional shrinkage, as it was caused by polymerization. Positive strain, consequently, refers to expansion or zones of stress relaxation, such as shown around voids (Figs. 3 and 4) and interface de-bonding (Fig. 4). No interfacial gaps were imaged by SEM in the small composite-filled cavities (Fig. 5a), while the half-sectioned large cavities disclosed bond detachment at the cavity bottom (Fig. 5b). Spectrophotometric analysis of light transmission through composite revealed that the adopted curing protocol delivered a significantly higher irradiation energy of 12 J/cm2 at the bottom of the small 2 mm deep filling (or 23.5% of the irradiation energy at the surface, being 51 J/cm2 ), versus only 1.4 J/cm2 at the bottom of the 4 mm deep filling (or 2.75% of the surface irradiation energy). Finally, mean DC, being 62.08 ± 1.96% and 61.67 ± 2.93% for the small and the large composite specimen, respectively, did not differ substantially (Fig. 5: bottom). The bottom-to-top conversion ratio (DCbottom /DCtop ) was only slightly higher for the small cavity (0.98) than for the large cavity (0.91).
Table 2 – Descriptive statistics of strain.
Small cavities Large cavities
Mean ± SD
Skewness ± SD
Mode ± SD
−2.97 ± 0.48 −3.11 ± 0.37
0.67 ± 0.72 3.38 ± 1.38
−3.90 ± 0.42 −3.93 ± 0.30
SD = standard deviation.
d e n t a l m a t e r i a l s 3 1 ( 2 0 1 5 ) 453–461
457
Fig. 3 – Collage of -CT images and reconstructions representing a small cavity. (a) Reconstructed slices of an uncured (left) and cured sample (right). Upon curing, the entrapped air bubble (white arrow) had slightly moved downwards and subtly increased in size. The occlusal surface noticeably intruded (open white arrows). (b) Color-coded strain map of the same section (left) and the 3D rendered model (right). Absence of strain, representing a zone of relaxation, can be detected immediately around the air bubble with a gradually increasing negative strain (shrinkage) further away. Small areas revealing regionally reduced strains can be detected (orange-colored globules), this in particular also along the right cavity margin, most likely representing other areas of void entrapment and/or de-bonding. Very striking is the bottom area, revealing (green-colored) positive strain, most likely representing a compensating zone between the bonding and the negative strain due to the inherent shrinkage. (c) Strain map of the surface of the restoration bottom (left) and restoration top (right). Except for the center part and the outer margin, both revealing negative strain, the main part of the bottom surface revealed (yellow-colored) absence of strain to (green-colored) positive strain. At the top surface, negative strain was detected for the major part, with a clear zone of reduced strain evolving into a relaxation zone of (green-colored) positive strain toward the right cavity side. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
be detected by -CT. Overall, a clear difference in the orientation of the displacement vectors was detected between the two cavity sizes, by which the hypothesis was rejected. Micro-CT makes use of the differences in X-ray attenuation to create a visible image of the specimen represented by different shades of gray or brightness levels. The image registration
software takes the differences in attenuation into account in the similarity measure between the uncured and cured situation. As a consequence, a certain level of heterogeneity with sufficiently contrasting elements inside the composite furthers an accurate registration. Although composites are constitutes from different materials, the radiopaque and
458
d e n t a l m a t e r i a l s 3 1 ( 2 0 1 5 ) 453–461
Fig. 4 – Collage of -CT images and reconstructions representing a large cavity. (a) Reconstructed slices of an uncured sample. No air bubbles can be detected in this section. (b) Reconstructed slices after curing, revealing an entrapped air bubble within the composite bulk, but also near the cavity corner (white arrows). Bonding detachment can be seen at the bottom (open white arrows). (c) Color-coded strain map of the same section. Negative strains were detected within the main restoration part, with even some (royal blue) areas (globules) exhibiting very high negative strains. The voids near the cavity border resulted in a large zone of (turquoise) very high positive strain due to void expansion, gradually decreasing more remote from the void. Likewise, the interface de-bonding resulted in a relaxation zone of positive strain in the surrounding area. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
radiolucent elements should be of sufficient size and/or distance from each other in relation to the spatial resolution to meet this condition. The flowable composite that was used in this study had sufficient contrast without any addition. However, it is possible that other composites might need the addition of radiopacifiers to obtain the same accuracy, such as glass fillers [13] or iodine-doped resins [21]. Unfortunately, most adhesives are rather radiolucent, which hampers the interpretation of gap occurrence and precise gap location with regard to the adhesive layer, especially if gaps were already present before the curing of the composite due to air
entrapment rather than shrinkage. In this study, no modifications were made to the existing materials, but zirconia-doping adhesives could facilitate the interpretation of gaps occurring at the tooth–adhesive–resin interface, as was shown in a recent study [22]. We opted in this exploratory study for a simplified yet more uniform design with cavities prepared in small composite cylinders rather than natural teeth. Cylindrical shapes of a more homogeneous and less radiopaque substrate are less subjected to scanning artifacts and their relatively small size made it feasible to scan at high resolution in accordance with
d e n t a l m a t e r i a l s 3 1 ( 2 0 1 5 ) 453–461
459
Fig. 5 – BSE images (top) and color-coded micro-Raman conversion mapping (bottom) of a representative small (a, left) and a large composite-filled cavity (b, right). A tight bond without de-bonding was observed in the small cavity contrary to clear de-bonding and gap formation in the large cavity. The degree of conversion was rather homogeneous for both cavities throughout the whole depth.
a voxel size of respectively 2.5 and 3.125 mm3 for the small and large cavities. This resolution is much higher than in most -CT studies. Moreover, using the ‘fast-scan’ mode of the scanner, it was possible to reduce the scan time to 10 min, limiting the risk on movement artifacts and still rendering images of adequate quality with an acceptable minimal amount of noise. In addition, it is sensible to keep the scan time as short as possible to avoid effects of spontaneous polymerization after brief exposure to ambient light. Another reason to opt for a composite substrate was to establish a more uniform bonding condition. As bonding to enamel is far better established than bonding to dentin [23], the dentin bond is more likely to fail due to the stress development and the shrinkage may thus be directed by the stronger enamel bond. This would have complicated the interpretation of the shrinkage directions. In this study, hence explanation cannot be sought in the weaker bonding to deep dentin. So far, only three studies have evaluated the internal, volumetric deformations of light-cured composite restorations in 3D, two by adding spherical filler particles as markers [13,14]. Takemura et al. intentionally created air bubbles within the composite, which were traced in a same manner [15]. In our study, we made use of the filler particles present in the flowable composite, by which any effect of adding other particles on the material properties was excluded. The displacement pattern found in the large cavities in this study is very similar to the pattern in the ‘bonded’ specimens
of the study by Cho et al., when cavities were also prepared in composite [14]. In that study, a rigid registration method and a clustering algorithm to trace the filler particles and to perform a pairing operation before and after polymerization were used. In the second study of Chiang et al., a similar set-up was used; however, more and different patterns were found, this very likely due to the fact that they used natural teeth [13]. Shrinkage was directed toward the bottom of the small cavity. The pattern disclosed in this study using -CT closely resembles the pattern described by Versluis et al., who applied finite element analysis to perfectly bonded composite restorations [24]. Shrinkage direction was concluded to be predominantly the result of boundary conditions, while it was only slightly affected by the relation toward the light. In the large cavities, a twofold displacement pattern was recorded with a downward and upward shrinkage direction for respectively the upper and bottom 2 mm composite. The latter upward displacement caused the composite to de-bond from the cavity bottom, as was confirmed by SEM. In previous studies, de-bonding within Class-I cavities often occurred at the bottom of the cavity [25,26]. This has often been attributed to the weaker bond to dentin than to enamel, which makes it more likely for the composite to detach in the deeper dentin regions [24]. However, this assumption is not valid in our study, since we intentionally used a uniform composite substrate to rule out the inherent difference in bonding receptiveness of enamel versus dentin. Even more, as the cavities in our
460
d e n t a l m a t e r i a l s 3 1 ( 2 0 1 5 ) 453–461
study were cut in freshly polymerized composite cylinders, the bonding effectiveness of the flowable composite was not only equal to the whole cavity wall and bottom, but most likely even higher than it would have been to enamel and dentin as the bond was formed by chemical co-polymerization via the adhesive in addition to micro-mechanical interlocking in the surface irregularities created by bur. Because the volume of the cylindric substrate was kept constant, the composite walls in the larger cavities were thinner and are thus more compliant; however, due to the low e-modulus of the composite substrate, the strain distribution is not expected to be very different in terms of its pattern, especially since the dominant movement occurred in the vertical direction. Moreover, strains perpendicular to the vertical cavity walls are more evenly distributed in a circular structure, as compared with more traditionally used rectangular cavity shapes, where one would expect stress peaks to occur in the corners [27]. Hence, in this study, the highest shrinkage stress upon polymerization was subjected to the cavity bottom of the large cavities, as can be deducted from the strain concentration in this area and the de-bonding. Despite only 2.75% irradiation energy delivered by the curing light could reach the bottom of the cavity of the 4 mm deep composite specimens, the bottom-to-top conversion ratio was, conspicuously, satisfactory even for the 4 mm deep larger specimens (0.91). This must be attributed to the combined effect of the high light output (1200 mW/cm2 ) of the curing device employed and the according to current clinical standards relatively extended curing time of 40 s. Compared to the 2 mm deep composite specimens, DC was only slightly lower for the 4 mm deep composite specimens, so that the somewhat reduced shrinkage appeared insufficient to prevent interface de-bonding in the deeper cavity region. However, a delay in conversion rate and differences in cross-linking [28] cannot be ruled out. It should be noted that the manufacturer (GC) does not recommend curing the flowable composite in such a thick 4 mm increment. Although not measured, it is known that the gel point will be delayed in resin composite at greater distances from the irradiated surface due to the decreased local intensity of the attenuated light with increasing depth [29]. The resin matrix will be sooner immobilized in the upper region than in the slower polymerizing bottom region. As a consequence, the later occurring displacements were oriented opposite to that of the upper 2 mm composite part and toward this already immobilized top part, because the stiffer part could now exert a stronger pull. This explains the twofold displacement pattern recorded for the 4 mm deep composite specimens. This is in accordance with a recent study [30], in which was demonstrated that the polymerization process in conventional composites was significantly delayed from a depth of 2 mm. In the current study, we could not only visualize the displacement vectors but also studied the displacement gradients, or in other words, the strain. A clear peak strain was found around −3.9% for both cavities. However, it must be taken into account that only a small part of the total shrinkage will contribute to shrinkage stress development, as the free shrinkage was also monitored. This value corresponds relatively well to the composite’s volumetric shrinkage of 4.38%,
as was provided by the manufacturer (GC), and thus confirms the adequate accuracy of our registration. The mean strain, however, was slightly lower (Table 2: around 3% for both cavities), because of local variety in strain due to the restriction of composite flow at regions where it adhered to the cavity wall. The strain showed a non-uniform pattern in all cavities. A composite is per definition a heterogenous material and on the micro-scale of its constituents, it will behave as such. Local heterogeneities, such as prepolymerized or inorganic particles within the resin matrix may all have contributed to the ‘blotted’ strain pattern throughout the whole restoration (Figs. 3 and 4). Furthermore, local flaws like entrapped air bubbles and interface de-bonding influenced the regional strain distribution (Figs. 3b, c and 4c). Voids increased in volume upon polymerization, which attributes to the free surface allowing the composite to shrink without building up shrinkage stress. The yellow-colored (thus close to zero) strain at the edge of the entrapped air bubble (Fig. 3b) illustrates the strain relief; further away strain gradually became more negative, indicating higher shrinkage. In the zone of interfacial debonding, the green-turquoise colors indicate positive strain, caused by the expansion of the void. These areas of expansion may compensate for the surrounding shrinkage, and thus regionally relieve shrinkage stress, as previously suggested by Alster et al. [16]. On the other hand, such small local deficiencies might be accountable for an increased susceptibility to mechanical fatigue degradation and a loss of cohesive strength in the long term. In the recent study of Takemura et al., where air bubbles were used as traceable markers, no mention has been made of deformation of the bubbles upon curing [15]. It is possible that due to the lower resolution achieved in that study, these regional effects got unnoticed. In conclusion, the proposed digital volume correlation using a non-rigid image registration enabled to measure and visualize the regional shrinkage strain vectors within the entire volume of the composite restoration. Experimental determination of directional flow and shrinkage is important to study the significance of polymerization contraction patterns in relation to marginal gap formation and stress distribution. This setup can also be used with natural teeth and for any shape, which makes it interesting to study more complex geometrical cavity designs. Moreover, internal regional flaws, such as air bubbles or cracking within the bulk, are often overlooked in other methods such as mathematical models or surface-based calculations. This may give important insight in how such flaws could relate to the initiation of de-bonding and provide a more accurate and realistic insight in the effect of shrinkage in real clinical situations.
Acknowledgements We would like to thank GC and Kuraray for providing the materials used in this study, and the Hercules Foundation for funding the project AKUL/09/001 “Micro- and nano-CT for the hierarchical analysis of materials”. A. Van Ende has been granted a PhD fellowship from the Research Foundation – Flanders (FWO). We declare no potential conflicts of interest with respect to the authorship and/or publication of this article.
d e n t a l m a t e r i a l s 3 1 ( 2 0 1 5 ) 453–461
references
[1] Davidson CL, De Gee A, Feilzer JA. The competition between the composite-dentin bond strength and the polymerization contraction stress. J Dent Res 1984;63:1396–9, http://dx.doi.org/10.1177/00220345840630121101. [2] Bergenholtz G. Evidence for bacterial causation of adverse pulpal responses in resin-based dental restorations. Crit Rev Oral Biol Med 2000;11:467–80, http://dx.doi.org/10.1177/10454411000110040501. [3] Braga RR, Ballester RY, Ferracane JL. Factors involved in the development of polymerization shrinkage stress in resin-composites: a systematic review. Dent Mater 2005;21:962–70, http://dx.doi.org/10.1016/j.dental.2005.04.018. [4] Boaro LCC, Brandt WC, Meira JBC, Rodrigues FP, Palin WM, Braga RR. Experimental and FE displacement and polymerization stress of bonded restorations as a function of the C-factor, volume and substrate stiffness. J Dent 2014;42:140–8, http://dx.doi.org/10.1016/j.jdent.2013.11.016. [5] Versluis A, Tantbirojn D, Pintado MR, DeLong R, Douglas WH. Residual shrinkage stress distributions in molars after composite restoration. Dent Mater 2004;20:554–64, http://dx.doi.org/10.1016/j.dental.2003.05.007. [6] Shortall AC, Palin WM, Burtscher P. Refractive index mismatch and monomer reactivity influence composite curing depth. J Dent Res 2008;87:84–8, http://dx.doi.org/10.1177/154405910808700115. [7] Fleming GJP, Khan S, Afzal O, Palin WM, Burke FJT. Investigation of polymerisation shrinkage strain, associated cuspal movement and microleakage of MOD cavities restored incrementally with resin-based composite using an LED light curing unit. J Dent 2007;35:97–103, http://dx.doi.org/10.1016/j.jdent.2006.05.003. [8] Satterthwaite JD, Maisuria A, Vogel K, Watts DC. Effect of resin-composite filler particle size and shape on shrinkage-stress. Dent Mater 2012;28:609–14, http://dx.doi.org/10.1016/j.dental.2012.01.007. [9] Schneider LFJ, Cavalcante LM, Silikas N. Shrinkage stresses generated during resin-composite applications: a review. J Dent Biomech 2010:2010, http://dx.doi.org/10.4061/2010/131630. [10] Li J, Fok AS-L, Satterthwaite J, Watts DC. Measurement of the full-field polymerization shrinkage and depth of cure of dental composites using digital image correlation. Dent Mater 2009;25:582–8, http://dx.doi.org/10.1016/j.dental.2008.11.001. [11] Miletic V, Manojlovic D, Milosevic M, Mitrovic N, Stankovic TS, Maneski T. Analysis of local shrinkage patterns of self-adhering and flowable composites using 3D digital image correlation. Quintessence Int (Berl) 2011;42: 797–804. [12] Kweon H-J, Ferracane JL, Kang K, Dhont J, Lee I-B. Spatio-temporal analysis of shrinkage vectors during photo-polymerization of composite. Dent Mater 2013;29:1236–43, http://dx.doi.org/10.1016/j.dental.2013.09.011. [13] Chiang Y-C, Rösch P, Dabanoglu A, Lin C-P, Hickel R, Kunzelmann K-H. Polymerization composite shrinkage evaluation with 3D deformation analysis from microCT images. Dent Mater 2010;26:223–31, http://dx.doi.org/10.1016/j.dental.2009.09.013. [14] Cho E, Sadr A, Inai N, Tagami J. Evaluation of resin composite polymerization by three dimensional micro-CT imaging and nanoindentation. Dent Mater 2011;27:1070–8, http://dx.doi.org/10.1016/j.dental.2011.07.008.
461
[15] Takemura Y, Hanaoka K, Kawamata R, Sakurai T, Teranaka T. Three-dimensional X-ray micro-computed tomography analysis of polymerization shrinkage vectors in flowable composite. Dent Mater J 2014;33:476–83, http://dx.doi.org/10.4012/dmj.2013-288. [16] Alster D, Feilzer A, De Gee JA, Moli J, Davidson ACL. The dependence of shrinkage stress reduction on porosity concentration in thin resin layers. J Dent Res 1992;71:1619–22, http://dx.doi.org/10.1177/00220345920710091401. [17] Oliveira FPM, Tavares JMRS. Medical image registration: a review. Comput Methods Biomech Biomed Eng 2014;17:73–93, http://dx.doi.org/10.1080/10255842.2012.670855. [18] Maes F, Collignon A, Vandermeulen D, Marchal G, Suetens P. Multimodality image registration by maximization of mutual information. IEEE Trans Med Imaging 1997;16:187–98, http://dx.doi.org/10.1109/42.563664. [19] Klein S, Staring M, Murphy K, Viergever MA, Pluim JPW. Elastix: a toolbox for intensity-based medical image registration. IEEE Trans Med Imaging 2010;29:196–205, http://dx.doi.org/10.1109/TMI.2009.2035616. [20] Mattes D, Haynor DR, Vesselle H, Lewellen TK, Eubank W. PET-CT image registration in the chest using free-form deformations. IEEE Trans Med Imaging 2003;22:120–8, http://dx.doi.org/10.1109/TMI.2003.809072. [21] Zhao Y, Lan J, Wang X, Deng X, Cai Q, Yang X. Synthesis of iodine-containing cyclophosphazenes for using as radiopacifiers in dental composite resin. Mater Sci Eng C Mater Biol Appl 2014;43:432–8, http://dx.doi.org/10.1016/j.msec.2014.07.050. [22] Rominu M, Manescu A, Sinescu C, Negrutiu ML, Topala F, Rominu RO, et al. Zirconia enriched dental adhesive: a solution for OCT contrast enhancement. Demonstrative study by synchrotron radiation microtomography. Dent Mater 2014;30:417–23, http://dx.doi.org/10.1016/j.dental.2014.01.004. [23] Van Meerbeek B, Peumans M, Poitevin A, Mine A, Van Ende A, Neves A, et al. Relationship between bond-strength tests and clinical outcomes. Dent Mater 2010;26:e100–21, http://dx.doi.org/10.1016/j.dental.2009.11.148. [24] Versluis A, Tantbirojn D, Douglas WH. Do dental composites always shrink toward the light? J Dent Res 1998;77:1435–45, http://dx.doi.org/10.1177/00220345980770060801. [25] Furness A, Tadros MY, Looney SW, Rueggeberg FA. Effect of bulk/incremental fill on internal gap formation of bulk-fill composites. J Dent 2014;42:439–49, http://dx.doi.org/10.1016/j.jdent.2014.01.005. [26] Kim HJ, Park SH. Measurement of the internal adaptation of resin composites using micro-CT and its correlation with polymerization shrinkage. Oper Dent 2014;39:E57–70, http://dx.doi.org/10.2341/12-378-L. [27] Rodrigues FP, Silikas N, Watts DC, Ballester RY. Finite element analysis of bonded model Class I “restorations” after shrinkage. Dent Mater 2012;28:123–32, http://dx.doi.org/10.1016/j.dental.2011.10.001. [28] Asmussen E, Peutzfeldt A. Short communication: polymer structure of a light-cured resin composite in relation to distance from the surface. Eur J Oral Sci 2003;111:277–9. [29] Martinsen M, El-Hajjar RF, Berzins DW. 3D full field strain analysis of polymerization shrinkage in a dental composite. Dent Mater 2013;29:e161–7, http://dx.doi.org/10.1016/j.dental.2013.04.019. [30] Steinhaus J, Moeginger B, Grossgarten M, Rosentritt M, Hausnerova B. Dielectric analysis of depth dependent curing behavior of dental resin composites. Dent Mater 2014;30:679–87, http://dx.doi.org/10.1016/j.dental.2014.03. 002.
