(G-251) Exploring CT Pixel and Voxel Size Effect on Anatomical Modeling in Mandibular Reconstruction

Friday, October 13, 2023

3:30 PM – 4:30 PM PDT

Location: Exhibit Hall - Row G - Poster # 251

Presenting Author(s)

LM

Luigi Melaragno

Student Researcher
The Ohio State University
Columbus, Ohio, United States

Co-Author(s)

Maariyah Ahmed, BSE (she/her/hers)

Student Researcher
The Ohio State University
Mason, Ohio, United States

Co-Author(s)

MG

Myra A. Garzanich

Student Research Assistant 1
The Ohio State University
Columbus, Ohio, United States
Nv

Natalia von Windheim, PhD

Postdoctoral Scholar
The Ohio State University, United States

Last Author(s)

KV

Kyle VanKoevering

Clinical Director / Otolaryngologist
The Ohio State University, United States

Primary Investigator(s)

MM

Megan Malara

Director
Modeling, Materials, and Manufacturing (M4) Division at The Ohio State University, United States

Introduction:: The role of 3D printing (3DP) in healthcare has continued to grow, becoming a standard of care in some specialties such as craniomaxillofacial reconstruction surgery. Computer-aided modeling and design (CAM/CAD) of patient anatomy from Computed Tomography (CT) imaging has allowed for improved planning of these surgical procedures. 3D printing in conjunction with CAM/CAD allows for tangible models of patient-specific anatomy that surgeons can use both preoperatively and intraoperatively for more effective outcomes in the operating room. The absence of CT imaging quality guidelines risks unsuitable imaging for patient-specific modeling production. CT scans consist of horizontal slices forming a 2D (X-Y) image of individual pixels. CAD allows these images to be layered, adding a Z-dimension termed “slice-thickness” to form voxels [1]. Previous studies assessing the Z-dimension have shown that requesting the highest resolution possible may be unnecessary, depending on clinical indication [2]. However, most clinically relevant scans vary in the X-Y dimension as well as the Z-dimension; therefore it is important to establish a standard for the pixel size to maintain accuracy consistently. This study aims to gain a more comprehensive understanding of the effect of CT quality on patient-specific mandibular modeling by evaluating the effect of both pixel size (X-Y) and slice thickness (Z), categorized by voxel size (X-Y-Z), on the accuracy of resulting mandibular models.

Materials and Methods:: 6 formalin-fixed cadaver heads were CT scanned at varying slice thicknesses and pixel sizes. Scans were labeled A-G (Table 1) based on their voxel dimensions. Scans A, C, E, and G have “Clinically Relevant” pixel sizes commonly associated with head and neck clinical applications at varying slice thicknesses. Scans B, D-F were taken with a “Constant Slice Thickness” of 3.00mm but varying pixel sizes. CT scans were segmented to isolate the mandible and digitally smoothed into resulting CAD models. The cadaveric mandibles underwent a metrology surface scan, resulting in a CAD model used as a gold standard for point-by-point comparison with the CT-developed CAD models (Figure 1). Two metrics were used to evaluate accuracy: The root mean square (RMS) value, a measure of the overall discrepancy between the CAD model and the true cadaveric anatomy, and the percentage of points that deviated from the true cadaveric anatomy by over 2.00mm. The RMS value is a common measure of accuracy for comparisons of an entire 3D model [3]. The percentage of points deviating by more than 2.00mm was used as the standard error for orthognathic surgeries is typically between 2.00-3.00mm [4]. For each measure of accuracy, three one-way ANOVA tests (α=0.05) were used to determine significant differences between CT scan resolutions, and Tukey Test Post-Hoc analysis determined the significance of specific comparison groups. The first test included the data from every scan (A-G), the second assessed the “Clinically Relevant” scans, and the third focused on the “Constant Slice Thickness” scans.

Results, Conclusions, and Discussions::

CAD Models created from CT scans with a pixel size less than or equal to 0.75mm and a slice thickness less than or equal to 3.00mm (voxels A-D) were significantly more accurate (p < 0.05) than scans with a pixel size of 1.32mm and a slice thickness of 5.00mm (voxel G) when comparing all scan data for both measures of accuracy. Additionally, voxels A and C were significantly more accurate than voxel G (p < 0.05) when analyzing the “Clinically Relevant” set for both measures of accuracy. Analysis of the “Constant Slice Thickness” set revealed that a pixel size of 0.32mm (voxel B) is significantly more accurate (p < 0.05) than a pixel size of 1.32mm (voxel F) at a 3.00mm slice thickness. However, this held true only in comparing the RMS values, but not the percentage of deviation over 2.00mm. This result is inconsistent with the ANOVA test comparing RMS values of all scan data. The discrepancy between results could be due to the smaller data set used for the “Constant Slice Thickness” test.

Results also showed that although voxel B is smaller (indicating higher resolution) than voxel C, C-level scans showed better accuracy. This indicates that voxel size and accuracy may not be an inverse-linear relationship. Rather, the Z-dimension may play a greater role in influencing accuracy than the X-Y pixel size. The findings of this study support this idea, indicating that solely scan G, which was the only scan with a slice thickness greater than 3.00mm, showed significant differences when compared to all the data. These results are consistent with previous research that suggests slice thicknesses of 3.00mm or less can produce models of similar accuracy [2]. Overall, results revealed that smaller pixel size in CT imaging shows a general trend towards improved accuracy in mandibular modeling. However, voxel data indicates that the Z-dimension may be stronger factor in determining accuracy. Therefore, further regression analysis will be explored to determine the specific relationship between pixel size and slice thickness in predicting the accuracy of a CAD model.

Acknowledgements (Optional): : CCE Cures Accelerator Award, Wright Center of Innovation, Medical Modeling Materials and Manufacturing Lab

References (Optional): : [1] Shafiq-Ul-Hassan, M., Zhang, G. G., Latifi, K., Ullah, G., Hunt, D. C., Balagurunathan, Y., Abdalah, M. A., Schabath, M. B., Goldgof, D. G., Mackin, D., Court, L. E., Gillies, R. J., & Moros, E. G. (2017). Intrinsic dependencies of CT radiomic features on voxel size and number of gray levels. Medical physics, 44(3), 1050–1062. https://doi.org/10.1002/mp.12123

[2] Ahmed, M., Melaragno, L. E., Nyirjesy, S. C., von Windheim, N., Fenberg, R., Herster, R., Sheldon, A., Binzel, K., Knopp, M. V., Herderick, E. D., & VanKoevering, K. K. (2023). Higher Computed Tomography (CT) Scan Resolution Improves Accuracy of Patient-specific Mandibular Models When Compared to Cadaveric Gold Standard. Journal of oral and maxillofacial surgery : official journal of the American Association of Oral and Maxillofacial Surgeons, S0278-2391(23)00476-7. Advance online publication. https://doi.org/10.1016/j.joms.2023.05.014

[3] Msallem, B., Sharma, N., Cao, S., Halbeisen, F. S., Zeilhofer, H. F., & Thieringer, F. M. (2020). Evaluation of the dimensional accuracy of 3D-printed anatomical mandibular models using FFF, SLA, SLS, MJ, and BJ printing technology. Journal of clinical medicine, 9(3), 817.

[4] Tondin, G. M., Leal, M. D. O. C. D., Costa, S. T., Grillo, R., Jodas, C. R. P., & Teixeira, R. G. (2022). Evaluation of the accuracy of virtual planning in bimaxillary orthognathic surgery: a systematic review. British Journal of Oral and Maxillofacial Surgery, 60(4), 412-421.