He Rong-Ting, Tu Ming-Gene, Huang Heng-Li, Tsai Ming-Tzu, Wu Jay, Hsu Jui-Ting
School of Dentistry, College of Dentistry, China Medical University, Taichung, 404, Taiwan, Republic of China.
Department of Dentistry, China Medical University and Hospital, 91 Hsueh-Shih Road, Taichung, 404, Taiwan, Republic of China.
BMC Med Imaging. 2019 Jan 23;19(1):10. doi: 10.1186/s12880-019-0313-9.
In this study, we explored how various preprocessing approaches can be employed to enhance the capability of dental CBCT to accurately estimate trabecular bone microarchitectural parameters.
In total, 30 bovine vertebrae cancellous bone specimens were used for in study. Voxel resolution 18-μm micro-computed tomography (micro-CT) and 100-μm dental CBCT were used to scan each specimen. Micro-CT images were used to calculate trabecular bone microarchitectural parameters; the results were set as the gold standard. Subsequently, before the dental CBCT images were converted into binary images to calculate trabecular bone microarchitectural parameters, three preprocessing approaches were used to process the dental CBCT images. For Group 1, no preprocessing approach was applied. For Group 2, images were sharpened and despeckable noises were removed. For Group 3, the function of local thresholding was added to Group 2 to form Group 3. For Group 4, the air pixels was removed from Group 3 to form Group 4. Subsequently, all images were imported into a software package to estimate trabecular bone microarchitectural parameters (bone volume fraction (BV/TV), trabecular thickness (TbTh), trabecular number (TbN), and trabecular separation (TbSp)). Finally, a paired t-test and a Pearson correlation test were performed to compare the capability of micro-CT with the capability of dental CBCT for estimating trabecular bone microarchitectural parameters.
Regardless of whether dental CBCT images underwent image preprocessing (Groups 1 to 4), the four trabecular bone microarchitectural parameters measured using dental CBCT images were significantly different from those measured using micro-CT images. However, after three image preprocessing approaches were applied to the dental CBCT images (Group 4), the BV/TV obtained using dental CBCT was highly positively correlated with that obtained using micro-CT (r = 0.87, p < 0.001); the correlation coefficient was greater than that of Group 1 (r = -0.15, p = 0.412), Group 2 (r = 0.16, p = 0.386), and Group 3 (r = 0.47, p = 0.006). After dental CBCT images underwent image preprocessing, the efficacy of using dental CBCT for estimating TbN and TbSp was enhanced.
Image preprocessing approaches can be used to enhance the efficacy of using dental CBCT for predicting trabecular bone microarchitectural parameters.
在本研究中，我们探讨了如何采用各种预处理方法来提高牙科锥形束计算机断层扫描（CBCT）准确估计小梁骨微结构参数的能力。
本研究共使用了30个牛椎体松质骨标本。采用体素分辨率为18μm的显微计算机断层扫描（micro-CT）和100μm的牙科CBCT对每个标本进行扫描。利用micro-CT图像计算小梁骨微结构参数；结果设为金标准。随后，在将牙科CBCT图像转换为二值图像以计算小梁骨微结构参数之前，使用三种预处理方法对牙科CBCT图像进行处理。第1组未应用任何预处理方法。第2组对图像进行锐化并去除斑点噪声。第3组在第2组的基础上增加局部阈值化功能。第4组在第3组的基础上去除空气像素。随后，将所有图像导入一个软件包中以估计小梁骨微结构参数（骨体积分数（BV/TV）、小梁厚度（TbTh）、小梁数量（TbN）和小梁间距（TbSp））。最后，进行配对t检验和Pearson相关检验，以比较micro-CT与牙科CBCT估计小梁骨微结构参数的能力。
无论牙科CBCT图像是否经过图像预处理（第1组至第4组），使用牙科CBCT图像测量的四个小梁骨微结构参数均与使用micro-CT图像测量的参数有显著差异。然而，在对牙科CBCT图像应用三种图像预处理方法后（第4组）, 使用牙科CBCT获得的BV/TV与使用micro-CT获得的BV/TV高度正相关（r = 0.87，p < 0.001）；相关系数大于第1组（r = -0.15，p = 0.412）、第2组（r = 0.16，p = 0.386）和第3组（r = 0.47，p = 0.006）。牙科CBCT图像经过图像预处理后，使用牙科CBCT估计TbN和TbSp的效能得到提高。
图像预处理方法可用于提高牙科CBCT预测小梁骨微结构参数的效能。
