Figure 2
From: Validation of bone mineral density measurement using quantitative CBCT image based on deep learning
The QCBCT-NET architecture combining Cycle-GAN and the multi-channel U-net77. The Cycle-GAN consisted of two generators of GCBCT⟶QCT, and GQCT⟶CBCT, and two discriminators of \({\text{D}}_{\text{CBCT}}\), and \({\text{D}}_{\text{QCT}}\). The multi-channel U-Net had two-channel inputs of CBCT and corresponding CYC_CBCT images, consisting of 3 × 3 convolution layers with batch normalization and ReLU activation, and had skip connections at each layer level. Max-pooling was used for down-sampling and transposed convolution was used for up-sampling. Consequently, the QCBCT-NET generated QCBCT images from CBCT images to quantitatively measure BMD in CBCTs.
