Skip to main content
Dryad

Photoacoustic tomography versus cone-beam computed tomography versus micro-computed tomography: Accuracy of 3D reconstructions of human teeth

Cite this dataset

Schneider, Sonja; Höhne, Christian; Schneider, Martin; Schmitter, Marc (2022). Photoacoustic tomography versus cone-beam computed tomography versus micro-computed tomography: Accuracy of 3D reconstructions of human teeth [Dataset]. Dryad. https://doi.org/10.5061/dryad.cc2fqz699

Abstract

Objectives: In this in-vitro study, teeth were imaged using photoacoustic tomography (PAT), cone-beam computed tomography (CBCT), and micro-computed tomography (µ-CT). The study had the aim: to identify the best wavelength for PAT images to determine the accuracy of the three imaging methods and to determine whether PAT images of teeth can achieve acceptable reconstruction quality.  

Methods: Nineteen human mandibular single-rooted incisors were extracted from patients with trauma or periodontitis. To determine the best wavelength for acquiring photoacoustic images, all 19 teeth were scanned in vitro with PAT, using different laser wavelengths between 680 and 960 nm. The images were analyzed using image analysis software. To assess the accuracy of PAT and compare it with the accuracy of CBCT, each tooth was also scanned in vitro using CBCT and the reference standard technique of µ-CT. Subsequently, three different three-dimensional models, one for each imaging technique, were created for each tooth. Finally, the three different three-dimensional models acquired for the same tooth were matched and analyzed regarding volume and surface.

Results: The highest quality tooth images were achieved using the 680 nm wavelength, which showed the best contrast ratio. The full geometry of the dental root (µ-CT compared with PAT) could be visualized with relative standard deviations of 0.12 mm for the surface and -7.33 mm3 for the volume (n=19). The full geometry of the dental root (µ-CT compared with CBCT) could be visualized with relative standard deviations of 0.06 mm for the surface and -14.56 mm3 for the volume (n=19). The difference between the PAT–µ-CT group and CBCT–µ-CT group regarding the total average of the root surface area was not significant (p > 0.06).

Conclusion: Images, which were acquired using PAT at 680nm showed the best contrast ratio, enabling the identification of dentin, cementum and the dental pulp. No significant differences were found between the PAT–µ-CT group and CBCT–µ-CT group regarding the total average of the RSA and the total volume. Thus, three-dimensional reconstructions based on in-vitro PAT are already of acceptable reconstruction quality.

Methods

1.1 Sample preparation 

This study was conducted in compliance with the Declaration of Helsinki, and the use of extracted human teeth was approved by the Medical Ethics Committee of the University of Wuerzburg (application number 15/15). All patients provided written informed consent for the use of their teeth.

Nineteen human single-rooted incisors were extracted from the mandible of patients with trauma or periodontitis. After extraction, the teeth were cleaned by means of an ultrasonic scaler (SONICflex LUX 2000L, KAVO; Biberach, Germany). After cleaning, the teeth were stored for seven months in a 1% chloramine-T solution at 5ºC to prevent bacterial colonization.

1.2  Experimental Setup: µ-CT

To acquire µ-CT images, the teeth were scanned using a µ-CT device (MetRIC: Micro and Region of Interest CT, Fraunhofer IIS; Wuerzburg, Germany, Fig. 1a). The setup for µ-CT acquisition, which compromises source, sample, and detector (anode material: tungsten), is shown in Fig. 1a.

To acquire µ-CT images, each tooth was held in place with oasis foam (Gravidus GmbH; Bremen, Germany) in a screw-capped centrifuge tube that was filled with 1% chloramine-T solution (Fig. 1b). For image acquisition, the image-acquisition parameters were set to 120 kV, 3 W (=25 muA) and an exposure time of 150 ms. µ-CT acquisition resulted in a stack of two-dimensional images (voxel size: 2 µm) for each tooth, which was saved in tagged image file format (TIFF).

1.3  Experimental Setup: CBCT

To acquire CBCT images, the Orthophos XG 3D (Dentsply Sirona; York, PA, USA, Fig. 2a) was used. Before CBCT imaging, the teeth were fixed in place in a plastic holder. The holder was specifically designed for this study in the software “Cubify Design” (3D Systems; Rock Hill, USA) and printed in polyethylene terephthalate using the “German RepRap X350” three-dimensional printer (ccd systems; Berlin, Germany). The holder was then placed on a tripod to bring the teeth to the right height (Fig. 2b).

For image acquisition, the image-acquisition parameters were set to high definition (voxel size 160 µm; field of view Ø 8 × 8 cm). CBCT acquisition resulted in a stack of two-dimensional images of each tooth, which were saved as TIFF files.

1.4  Experimental Setup: PAT

To create the PAT images, the MSOT inVision 256 (Multispectral Optoacoustic Tomography, iThera Medical GmbH; Munich, Germany, Fig. 3) was used. Each tooth was embedded in a 20 ml plastic syringe, in a solution of 2% agarose with 1% lipofundin (Intralipid). The agarose was used to fix the tooth in a suitable position, and the lipofundin ensured better light scattering. Fig. 3a shows how the tooth was fixed in the agarose and lipofundin mixture, and Fig. 3b shows the MSOT inVision holder. A schematic of the components of the MSOT is shown in Fig. 3c.

For PAT imaging, the MSOT inVision was filled with warm water (25ºC), and the step size (distance between one picture to the next) was set to 0.3 mm (data filter 50 kHz – 5 MHz; resolution 150 µm). An algorithm based on back-projection was used for image reconstruction. Back-projection reconstruction is a very fast and semi-quantitative reconstruction algorithm that allows efficient generation of anatomical images.

After being fixed in place, teeth were scanned at wavelengths of 680–960 nm, in 10 nm increments. Depending on the size of the individual tooth, 9–9.3 minutes were needed to record each tooth. PAT image acquisition resulted in a stack of two-dimensional images of each tooth, which were saved as a TIFF file.

After data acquisition, the images were analyzed to determine the imaging quality of each wavelength.

1.5  Data Analysis

First, the most suited wavelength for PAT when imaging teeth was identified by using image analysis software (ImageJ2, public domain software): the distribution of the grey values was analyzed by choosing “Plot Profile”. The grey values and their distribution stand for the contrast of the images. The grey values were analyzed further using the Kruskal-Wallis-Test and the post-hoc Mann-Whitney U-test.

For data analysis, the volume and surface data for PAT and CBCT were compared with the corresponding data from µ-CT, with special emphasis on the dental root. We chose µ-CT images as the reference standard because the spatial resolution of µ-CT (up to 2 µm) is more accurate than that of CBCT (160 µm) and PAT (150 µm).

The root surface deviation (RSD) and root volume deviation (RVD) between the PAT and µ-CT images, and between the CBCT and µ-CT images were calculated and analyzed. To calculate the RSD, the total average of the RSD (average of the positive and negative RSD values) was determined.

The following standardized protocol was used when creating and matching the three-dimensional models:

First, the acquired two-dimensional images (TIFF files) of the incisors were assembled and compiled in three-dimensional models in version 4.10.1 of “3D Slicer,” a proprietary software program written in Matlab 2010b (MathWorks; Natick, Massachusetts, USA).

Second, GOM Inspect was used to match the PAT images with the µ-CT images.  

The GOM Inspect software is used by the industry in product development, quality control and production with a focus on alignment and deviation measurement. The software is certified by NIST (National Institute of Standards and Technology, Gaithersburg, Maryland, United States) and PTB (Physikalisch-Technische Bundesanstalt, Braunschweig und Berlin). GOM Inspect has been placed in Category 1 with the smallest measurement deviations and was used in many studies[18]. GOM was used with a double-stage automized matching process.

In the first stage, GOM performed a preliminary matching process. In the second stage, the automized precision matching was performed. Subsequently, the software program Fusion 360TM (2020 Autodesk, Inc.) was applied to crop the three-dimensional models of the same mandibular incisor to the same size. Afterwards, the volume of the root was calculated by applying the software program Meshmixer 3.5.474 (2017 Autodesk, Inc.).

Once the three-dimensional models had been matched and cropped, the total average of the deviation, the normalized total average of the deviation, the positive deviation, and the negative deviation of the root surface area (RSA) were calculated, using µ-CT as the reference standard. In addition, the total volume of the root was determined and compared among the three models.

1.6  Statistical Evaluation

Data were analyzed in the software program SPSS Statistics (Statistical Package for the Social Sciences, version 25.0.0; Armonk, USA). Several statistical tests were used, which are described in the following subsections.

1.6.1  Frequency Distribution

As a first step, the normal distribution of the data was analyzed using the Kolmogorov–Smirnov test. This test compares the observed cumulative distribution of scores with the theoretical cumulative distribution for a normally distributed population.

1.6.2   Assessment of Differences between Groups

Because not all groups showed a normal distribution, a non-parametric test was applied to assess the significance level. For this purpose, the Kruskal-Wallis test and Mann–Whitney U tests were applied. In the present study, the level of significance was set at ≤ 0.05 (highly significant: p-value ≤ 0.01).