This paper presents a two-stage method for beam hardening artifact correction of dental cone beam computerized tomography (CBCT). The proposed artifact reduction method is designed to improve the quality of maxillofacial imaging, where soft tissue details are not required. Compared to standard CT, the additional difficulty of dental CBCT comes from the problems caused by offset detector, FOV truncation, and low signal-to-noise ratio due to low X-ray irradiation. To address these problems, the proposed method primarily performs a sinogram adjustment in the direction of enhancing data consistency, considering the situation according to the FOV truncation and offset detector. This sinogram correction algorithm significantly reduces beam hardening artifacts caused by high-density materials such as teeth, bones, and metal implants, while tending to amplify special types of noise. To suppress such noise, a deep convolutional neural network is complementarily used, where CT images adjusted by the sinogram correction are used as the input of the neural network. Numerous experiments validate that the proposed method successfully reduces beam hardening artifacts and, in particular, has the advantage of improving the image quality of teeth, associated with maxillofacial CBCT imaging.

中文翻译：

---

在低剂量牙科 CBCT 中减少光束硬化伪影的两阶段方法

本文提出了一种用于牙锥束计算机断层扫描 (CBCT) 光束硬化伪影矫正的两阶段方法。所提出的伪影减少方法旨在提高不需要软组织细节的颌面成像质量。与标准 CT 相比，牙科 CBCT 的额外困难来自于探测器偏移、FOV 截断和低 X 射线照射导致的低信噪比等问题。为了解决这些问题，所提出的方法主要在增强数据一致性的方向上进行正弦图调整，考虑到根据 FOV 截断和偏移检测器的情况。这种正弦图校正算法显着减少了由牙齿、骨骼和金属植入物等高密度材料引起的光束硬化伪影，同时倾向于放大特殊类型的噪音。为了抑制这种噪声，补充使用了深度卷积神经网络，其中通过正弦图校正调整的 CT 图像用作神经网络的输入。大量实验证实，所提出的方法成功地减少了光束硬化伪影，特别是具有提高与颌面部 CBCT 成像相关的牙齿图像质量的优势。