Department Human Health and Nutritional Science, College of Biological Science, University of Guelph, Guelph, Ontario, Canada.
Department of Biomedical Sciences, Ontario Veterinary College, University of Guelph, Guelph, Ontario, Canada.
PLoS One. 2019 Mar 25;14(3):e0214123. doi: 10.1371/journal.pone.0214123. eCollection 2019.
This study's objective was to determine the accuracy of using current computed tomography (CT) scan and software techniques for rapid prototyping by quantifying the margin of error between CT models and laser scans of canine skull specimens. Twenty canine skulls of varying morphology were selected from an anatomy collection at a veterinary school. CT scans (bone and standard algorithms) were performed for each skull, and data segmented (testing two lower threshold settings of 226HU and -650HU) into 3-D CT models. Laser scans were then performed on each skull. The CT models were compared to the corresponding laser scan to determine the error generated from the different types of CT model parameters. This error was then compared between the different types of CT models to determine the most accurate parameters. The mean errors for the 226HU CT models, both bone and standard algorithms, were not significant from zero error (p = 0.1076 and p = 0.0580, respectively). The mean errors for both -650HU CT models were significant from zero error (p < 0.001). Significant differences were detected between CT models for 3 CT model comparisons: Bone (p < 0.0001); Standard (p < 0.0001); and -650HU (p < 0.0001). For 226HU CT models, a significant difference was not detected between CT models (p = 0.2268). Independent of the parameters tested, the 3-D models derived from CT imaging accurately represent the real skull dimensions, with CT models differing less than 0.42 mm from the real skull dimensions. The 226HU threshold was more accurate than the -650HU threshold. For the 226HU CT models, accuracy was not dependent on the CT algorithm. For the -650 CT models, bone was more accurate than standard algorithms. Knowing the inherent error of this procedure is important for use in 3-D printing for surgical planning and medical education.
本研究旨在通过量化 CT 模型与犬颅骨标本激光扫描之间的误差幅度，确定当前 CT 扫描和软件技术在快速成型中的准确性。从兽医学校解剖收藏中选择了 20 个形态各异的犬颅骨。对每个颅骨进行 CT 扫描（骨和标准算法），并将数据分段（测试 226HU 和-650HU 两个较低的阈值设置）成 3-D CT 模型。然后对每个颅骨进行激光扫描。将 CT 模型与相应的激光扫描进行比较，以确定不同类型的 CT 模型参数产生的误差。然后在不同类型的 CT 模型之间比较此误差，以确定最准确的参数。226HU CT 模型（骨和标准算法）的平均误差与零误差无显著差异（p=0.1076 和 p=0.0580）。两种-650HU CT 模型的平均误差均与零误差有显著差异（p<0.001）。在 3 个 CT 模型比较中检测到 CT 模型之间存在显著差异：骨（p<0.0001）；标准（p<0.0001）；和-650HU（p<0.0001）。对于 226HU CT 模型，CT 模型之间未检测到显著差异（p=0.2268）。无论测试参数如何，CT 成像得出的 3-D 模型都准确地代表了真实颅骨的尺寸，CT 模型与真实颅骨尺寸的差异小于 0.42 毫米。226HU 阈值比-650HU 阈值更准确。对于 226HU CT 模型，准确性不依赖于 CT 算法。对于-650 CT 模型，骨比标准算法更准确。了解该过程的固有误差对于在手术规划和医学教育中进行 3-D 打印非常重要。
