Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

< Back to Article

Fig 1.

Schematic view of the overall work process.

More »

Fig 1 Expand

Table 1.

Exposure conditions for image data acquisition.

More »

Table 1 Expand

Fig 2.

The data preparation process.

CBCT images were generated by masking out the background signal noise (a). MDCT images were resampled to match the resolution of the CBCT images and then prepared using region-of-interest (ROI) masking and a two-step registration process utilizing affine and non-rigid registration to match the CBCT images (b). The final set of CBCT and MDCT image pairs was used for training and testing (c). CBCT, cone-beam computed tomography; MDCT, multidetector computed tomography.

More »

Fig 2 Expand

Fig 3.

Schematic of COMPUNet (multi-planar 2.5D U-Net), which comprises three single-planar 2.5D U-Nets.

(a) The architecture of a single-planar 2.5D U-Net with a ResNet-34 encoder and (b) attention modules. xl: input features, α: attention coefficients, g: gating signal collected from a coarser scale.

More »

Fig 3 Expand

Table 2.

Clinical CBCT image evaluation chart.

More »

Table 2 Expand

Fig 4.

Quantitative evaluation through an ablation study.

NRMSE (a), SSIM (b), and MAE (c) for comparisons between the original (oCBCT) and predicted CBCT (pCBCT), showing that the best-performing model was the proposed model compared to Model 1 and Model 2. (* p < 0.005).

More »

Fig 4 Expand

Fig 5.

The trabecular bone pattern (arrow) in the same image slides.

The fine bone details are best preserved in the predicted CBCT of the proposed model, COMPUNet, compared to that of Model 1 and Model 2.

More »

Fig 5 Expand

Fig 6.

Mean scores of individual items in the clinical CBCT evaluation chart.

Items for the artifacts and noise criteria showed improvements in the pCBCT images, while those for the resolution criterion improved less.

More »

Fig 6 Expand

Fig 7.

Original CBCT with an overall image grade of poor and the corresponding predicted CBCT image.

(a) The maxilla in an axial view (b), the anterior teeth region in a sagittal view, and (c) the temporomandibular region in a parasagittal view show reduced artifacts and noise in the COMPUNet CBCT compared to the original CBCT images.

More »

Fig 7 Expand

Fig 8.

Distribution of overall image quality evaluation grades.

The proportion of CBCT with poor grades decreases (sky blue) and those with good grades increases (dark blue) in the predicted CBCT (pCBCT) compared to the original CBCT (oCBCT) for both observers.

More »

Fig 8 Expand

Table 3.

Scores from the qualitative evaluation.

More »

Table 3 Expand