Search

Convolutional Neural Networks (CNNs) such as U-Net have been widely used for medical image segmentation. Dental restorations are prominent features of dental radiographs. Applying U-Net on the panoramic image is challenging, as the shape, size and frequency of different restoration types vary. We hypothesized that models trained on smaller, equally spaced rectangular image crops (tiles) of the panoramic would outperform models trained on the full image. A total of 1781 panoramic radiographs were annotated pixelwise for fillings, crowns, and root canal fillings by dental experts. We used different numbers of tiles for our experiments. Five-times-repeated three-fold cross-validation was used for model evaluation. Training with more tiles improved model performance and accelerated convergence. The F1-score for the full panoramic image was 0.7, compared to 0.83, 0.92 and 0.95 for 6, 10 and 20 tiles, respectively. For root canals fillings, which are small, cone-shaped features that appear less frequently on the radiographs, the performance improvement was even higher (+294%). Training on tiles and pooling the results thereafter improved pixelwise classification performance and reduced the time to model convergence for segmenting dental restorations. Segmentation of panoramic radiographs is biased towards more frequent and extended classes. Tiling may help to overcome this bias and increase accuracy.

Objectives Although ultra-short pulsed laser (USPL) microstructuring has previously improved zirconia bond-strength, it is yet unclear how different laser-machined surface microstructures and patterns may influence the material’s mechanical properties. Therefore, the aim of this study was to assess the flexural strength of zirconia after different USPL settings creating three different geometrical patterns with structures in micrometer scale. Methods One hundred sixty zirconia bars (3Y-TZP, 21 × 4 × 2.1 mm) were prepared and randomly divided into five groups (n = 32): no surface treatment (negative control-NC); sandblasting with Al2O3 (SB); and three laser groups irradiated with USPL (Nd:YVO4/1064 nm/2-34 J/cm2/12 ps): crossed-lines (LC), random-hatching (LR), and parallel-waves (LW). Bars were subjected to a four-point flexural test (1 mm/min) and crystal phase content changes were identified by X-ray diffraction. Surface roughness and topography were analyzed through 3D-laser-profilometry and SEM. Data were analyzed with parametric tests for roughness and Weibull for flexural strength (α = 5%). Results LR (Mean[95%CI]: 852.0 MPa, [809.2–894.7]) was the only group that did not show a significantly different flexural strength than NC (819.8 MPa, [796.6–842.9]), (p > 0.05). All laser groups exhibited higher Weibull moduli than NC and SB, indicating higher reliability and homogeneity of the strength data. An increase of monoclinic phase peak was only observed for SB. Conclusion In conclusion, USPL created predictable, homogeneous, highly reproducible, and accurate surface microstructures on zirconia ceramic. The laser-settings of random-hatching (12 ps pulses) increased 3Y-TZP average surface roughness similarly to SB, while not causing deleterious crystal phase transformation or loss of flexural strength of the material. Furthermore, it has increased the Weibull modulus and consequently material’s reliability. Clinical significance Picosecond laser microstructuring (LR conditions) of 3Y-TZP ceramic does not decrease its flexural strength, while increasing materials realiability and creating highly reproducible and accurate microstructures. These features may be of interest both for improving clinical survival of zirconia restorations as well as enhancing longevity of zirconia implants.

We aimed to assess the effects of hyperparameter tuning and automatic image augmentation for deep learning-based classification of orthodontic photographs along the Angle classes. Our dataset consisted of 605 images of Angle class I, 1038 images of class II, and 408 images of class III. We trained ResNet architectures for classification of different combinations of learning rate and batch size. For the best combination, we compared the performance of models trained with and without automatic augmentation using 10-fold cross-validation. We used GradCAM to increase explainability, which can provide heat maps containing the salient areas relevant for the classification. The best combination of hyperparameters yielded a model with an accuracy of 0.63-0.64, F1-score 0.61-0.62, sensitivity 0.59-0.65, and specificity 0.80-0.81. For all metrics, it was apparent that there was an ideal corridor of batch size and learning rate combinations; smaller learning rates were associated with higher classification performance. Overall, the performance was highest for learning rates of around 1-3 x 10(-6) and a batch size of eight, respectively. Additional automatic augmentation improved all metrics by 5-10% for all metrics. Misclassifications were most common between Angle classes I and II. GradCAM showed that the models employed features relevant for human classification, too. The choice of hyperparameters drastically affected the performance of deep learning models in orthodontics, and automatic image augmentation resulted in further improvements. Our models managed to classify the dental sagittal occlusion along Angle classes based on digital intraoral photos.

PURPOSE Low-molecular weight residuals eluting from dental materials may contribute to local and systemic adverse effects. Therefore, the aim of the present study was to compare triethylene glycol dimethacrylate (TEGDMA)-based commercial infiltrant with different conventional resin-based materials regarding their release of monomers. MATERIALS AND METHODS Cylindrical blocks (n = 10) of either two sealants (Helioseal, Delton FS+), a composite (EcuSphere), an adhesive (Teco) and an infiltrant (Icon) were prepared. Additionally, 20 artificial lesions (depths ≥100 µm) were created in bovine enamel and after etching with phosphoric acid infiltrated with the infiltrant. Except for 10 infiltrated lesions, all other specimens were polished. Each specimen was stored in 1 ml distilled water (elution medium) for 240 h. The medium was renewed in logarithmical divided time periods (4.5 min-76 h). RESULTS Total concentrations of eluted monomers within 240 h from the cylindrical specimens were 0.04-0.09 mg/ml (p >0.05; Mann-Whitney test). Unpolished infiltrated specimens showed significantly higher monomer concentrations compared to all other groups, whereas polishing of specimens resulted in significantly lower concentrations (p

OBJECTIVES The aim of this study was to compare the caries preventive effect of highly fluoridated dentifrices and gels on sound dentin as well as on artificial dentin caries-like lesions. METHODS Bovine dentin specimens (n = 240), with 2 different surfaces each (1 sound surface [sound treatment (ST)] and one caries lesion [demineralized treatment (DT)]), were prepared and randomly allocated to one highly (6 × 120 min demineralization/day [H]) and one lowly cariogenic (6 × 60 min demineralization/day [L]) pH-cycling model. Treatments during pH-cycling (28 days) were: brushing 2×/day with: 0 ppm F [H0/L0], 1,450 ppm F [H1,450/L1,450], 2,800 ppm F [H2,800/L2,800], 5,000 ppm F [H5,000/L5,000], 5,000 ppm F plus TCP [H5,000+TCP/L5,000+TCP], and 12,500 ppm F [H12,500/L12,500] containing dentifrices/gels. Dentifrice/gel slurries were prepared with deionized water (1:2 wt/wt). Differences in integrated mineral loss (∆∆Z) and ∆ lesion depth were calculated between values before and after pH-cycling using transversal microradiography. RESULTS The correlation between ΔΔZDT and F- was strong for the highly (rH = 0.691; p < 0.001) and moderate (rL = 0.500; p < 0.001) for the lowly cariogenic model, indicating a fluoride dose-response for both. Significant differences for ΔΔZDT and ΔΔZST could be found between H0, H1,450, H5,000, and H12,500 as well as L0, L5,000, and L125,000 (p ≤ 0.046; analysis of covariance [ANCOVA]). Except for 0 ppm F-, no significant difference in ΔΔZST and ΔΔZDT could be found between the highly and lowly cariogenic model (p ≥ 0.056; ANCOVA). CONCLUSION For both pH-cycling conditions a dose-response for fluoride could be revealed. For elderly people with exposed root surfaces, the use of gels containing 12,500 ppm F instead of regularly (1,450 ppm F) or highly (5,000 ppm F) fluoridated dentifrices should be further investigated, as it offered higher caries-preventive effects in vitro.

OBJECTIVES The aim of this double-blinded, randomized, cross-over in situ study was to evaluate the re- and demineralization characteristics of sound dentin as well as highly and lowly demineralized dentin after the application of different fluoride concentrations. METHODS In each of four experimental legs of four weeks 20 participants wore intraoral mandibular appliances containing two (highly demineralized [EH]) bovine enamel and four (lowly and highly demineralized [DL,DH]) bovine dentin specimens (n = 480). Each specimen included one sound (ST) and one demineralized lesion area (DT). The four randomly allocated treatments included the following dentifrices: fluoride-free, zinc-carbonate-nano-hydroxyapatite [nHA0], 0 ppm F- [negative control,NaF0], 1100 ppm F- as NaF [standard therapy,NaF1100] and 5000 ppm F- as NaF [positive control,NaF5000]. Differences in integrated mineral loss (ΔΔZ) and lesion depth (ΔLD) were calculated between values before and after the in situ period using transversal microradiography. RESULTS After the in situ period specimens of nHA0 and NaF0 showed signs of demineralization, indicated by significantly lower ΔZ&LD values for EH and DL (only nHA0)(p ≤ 0.028), whereas specimens of NaF1100 and NaF5000 showed signs of remineralization, indicated by significantly higher ΔZ values for EH (only NaF5000), DL and DH (p ≤ 0.012). The correlation between ΔΔZDT/ΔΔZST and F- was moderate for EH(rDT = 0.497;rST = 0.463) and DL(rDT = 0.575;rST = 0.598) and strong for DH(rDH = 0.700;rST = 0.611)(p < 0.001). No significant differences for ΔΔZDT/ΔΔZST were observed between nHA0 and NaF0(p ≥ 0.333;ANCOVA). CONCLUSION The present in situ model was capable to reveal a fluoride dose-response on sound, lowly and highly demineralized dentin and also enamel specimens. Furthermore, both fluoride-free dentifrices, one containing nanohydroxyapatite, did not hamper demineralization. CLINICAL SIGNIFICANCE The present in situ model was capable to reveal a fluoride dose-response on dentin similar to the anticipated clinical efficacy. Highly demineralized specimens seem to be recommendable for measuring anti-caries effects on dentin in situ. Furthermore both fluoride-free dentifrices, one containing nanohydroxyapatite, did not hamper demineralization. The study was registered in the German Clinical Trials Register (DRKS-ID: DRKS00011653).