Deep-learning synthesized pseudo-CT for MR high-resolution pediatric cranial bone imaging (MR-HiPCB)

Purpose CT is routinely used to detect cranial abnormalities in pediatric patients with head trauma or craniosynostosis. This study aimed to develop a deep learning method to synthesize pseudo-CT (pCT) images for MR high-resolution pediatric cranial bone imaging to eliminating ionizing radiation from CT. Methods 3D golden-angle stack-of-stars MRI were obtained from 44 pediatric participants. Two patch-based residual UNets were trained using paired MR and CT patches randomly selected from the whole head (NetWH) or in the vicinity of bone, fractures/sutures, or air (NetBA) to synthesize pCT. A third residual UNet was trained to generate a binary brain mask using only MRI. The pCT images from NetWH (pCT(NetWH)) in the brain area and NetBA (pCT(NetBA)) in the nonbrain area were combined to generate pCT(Com). A manual processing method using inverted MR images was also employed for comparison. Results pCT(Com) (68.01 +/- 14.83 HU) had significantly smaller mean absolute errors (MAEs) than pCT(NetWH) (82.58 +/- 16.98 HU, P < 0.0001) and pCT(NetBA) (91.32 +/- 17.2 HU, P < 0.0001) in the whole head. Within cranial bone, the MAE of pCT(Com) (227.92 +/- 46.88 HU) was significantly lower than pCT(NetWH) (287.85 +/- 59.46 HU, P < 0.0001) but similar to pCT(NetBA) (230.20 +/- 46.17 HU). Dice similarity coefficient of the segmented bone was significantly higher in pCT(Com) (0.90 +/- 0.02) than in pCT(NetWH) (0.86 +/- 0.04, P < 0.0001), pCT(NetBA) (0.88 +/- 0.03, P < 0.0001), and inverted MR (0.71 +/- 0.09, P < 0.0001). Dice similarity coefficient from pCT(Com) demonstrated significantly reduced age dependence than inverted MRI. Furthermore, pCT(Com) provided excellent suture and fracture visibility comparable to CT. Conclusion MR high-resolution pediatric cranial bone imaging may facilitate the clinical translation of a radiation-free MR cranial bone imaging method for pediatric patients.