Objectives: To evaluate the accuracy of ITK-SNAP software for measuring volumes of a non-regular shape structure, using cone beam computed tomography (CBCT) scans, besides for developing a mathematical model to correct the software measurement error in case it existed.
Methods: A phantom made by moulding a rubber duck's head was filled with total (38,000 mm3) and partial volumes of water (7000 mm3, 14,000 mm3, 21,000 mm3, 28,000 mm3 and 35,000 mm3), which constituted the gold standards. The sound phantom and the phantom filled with different volumes of water were scanned in a Picasso Trio CBCT unit set at 80 kVp, 3.7 mA, 0.2 mm3 voxel and 12 × 8.5 cm field of view. Semi-automatic segmentation was performed with ITK-SNAP 3.0 software by two trained oral radiologists. Linear regression analyzed the relation between ITK-SNAP calculated volumes and the gold standard. Intraclass correlation coefficient was applied to analyze the reproducibility of the method. Significance level was set at 5%.
Results: Linear regression analysis showed a significant relationship between ITK-SNAP volumes and the gold standard (F = 22,537.3, p < 0.0001), with an R2 of 0.9993. The average error found was 4.7 (± 4.3) %. To minimize this error, a mathematical model was developed and provided a reduction of it. ICC revealed excellent intra-examiner agreements for both examiners 1 (ICC = 0.9991, p < 0.0001) and 2 (ICC = 0.9989, p < 0.0001). Likewise, inter-examiner agreement was excellent (ICC = 0.9991, p < 0.0001).
Conclusion: The software showed to be accurate for evaluating non-regular shape structures. The mathematical model developed reduced an already small error on the software's measurements.
Keywords: Cone beam computed tomography; Cross-sectional anatomy; Software; Three-dimensional imaging.