Effect of voxel size on the accuracy of 3D reconstructions with cone beam CT

Abstract

Objectives: The various types of cone beam CT (CBCT) differ in several technical characteristics, notably their spatial resolution, which is defined by the acquisition voxel size. However, data are still lacking on the effects of voxel size on the metric accuracy of threedimensional (3D) reconstructions. This study was designed to assess the effect of isotropic voxel size on the 3D reconstruction accuracy and reproducibility of CBCT data. Methods: The study sample comprised 70 teeth (from the Institut d'Anatomie Normale, Strasbourg, France). The teeth were scanned with a KODAK 9500 3DH CBCT (Carestream Health, Inc., Marne-la-Vallée, France), which has two voxel sizes: 200 mm (CBCT 200 mm group) and 300 mm (CBCT 300 mm group). These teeth had also been scanned with the KODAK 9000 3DH CBCT (Carestream Health, Inc.) (CBCT 76 mm group) and the SCANCO Medical micro-CT XtremeCT (SCANCO Medical, Brü ttisellen, Switzerland) (micro-CT 41 mm group) considered as references. After semi-automatic segmentation with AMIRAH software (Visualization Sciences Group, Burlington, MA), tooth volumetric measurements were obtained. Results: The Bland-Altman method showed no difference in tooth volumes despite a slight underestimation for the CBCT 200 mm and 300 mm groups compared with the two reference groups. The underestimation was statistically significant for the volumetric measurements of the CBCT 300 mm group relative to the two reference groups (Passing-Bablok method). Conclusions: CBCT is not only a tool that helps in diagnosis and detection but it has the complementary advantage of being a measuring instrument, the accuracy of which appears connected to the size of the voxels. Future applications of such measurements with CBCT are discussed.

Impact of Voxel Size Variation on CBCT-Based Diagnostic Outcome in Dentistry: a Systematic Review

“…The remaining eight studies were based on numerical outcomes measured in the images. Three studies measured 3D surface reconstructions [17,30,31] (and used micro-CT of the teeth as the gold standard), two measured the height and thickness of alveolar bone [11,32], one assessed linear bone measurements (distance between established anatomical landmarks) [33], one measured the thickness of the soft tissue of the face [34], and one measured tooth and root length [35], and these last five studies used caliper measurements of the skulls as the gold standard. Six studies further compared the CBCT images in different resolutions to images acquired using intraoral radiographic systems [18,21,[25][26][27]35].…”

“…Micro-CT (isotropic 25 μm resolution) was the gold standard CBCT reconstructions were larger than their micro-CT counterparts. For the CBCT-based images, differences were found between the 0.25 and 0.2-mm voxel sizes, but not between the 0.2 and 0.133 mm Maret et al [31] One observer measured the images, in duplicate. [32] Two observers measured the images, in duplicate.…”

“…Of the three studies which measured the accuracy of 3D reconstructions quantitatively, two used micro-CT of the teeth as a gold standard [30,31]. Both studies found that CBCTbased images, independent of the voxel size, underestimated, though not statistically significant, the volumes measured with micro-CT. Micro-CT suffers from artifacts similar to those observed in CBCT, since the scanning technology is comparable.…”

Influência do tamanho do voxel na qualidade de imagem tomográfica: revisão de literatura

“…No entanto, os protocolos atrelam o uso de voxels menores a tempos de exposição e mA maiores, o que, invariavelmente, provoca uma melhor resolução da imagem e o aumento da dose de radiação. Isso ocorre quando os protocolos são totalmente preestabelecidos pelo fabricante de alguns aparelhos de TCFC 11,16,17 . Para comparar a qualidade de imagem de dois diferentes aparelhos de TCFC, o Siremobil Iso-C3D (Siemens Medical Solutions, Erlangen Alemanha) e o NewTom 9000 (NIM srl, Verona, Itália), foram adquiridas imagens de um crânio seco humano.…”

Surfing the spectrum – what is on the horizon?

“…However, with increasing voxel size there can be a tendency to underestimate volumetric measurements which can become significant above 300µm 62 .…”