2D-to-3D: Predicting three-dimensional (3D) cephalometric measurements from two conventional X-ray images : From 2D to 3D with a computational tool without using computed tomography.

Abstract

The use of computed tomography (CT) or cone-beam CT (CBCT) enables three-dimensional (3D) cephalometric analysis but exposes patients to high levels of radiation, while conventional X‑ray imaging provides projected measurements for two-dimensional (2D) cephalometric analysis with lower radiation exposure. The presented tool overcomes the disadvantages of both imaging modalities for performing 3D cephalometric analysis. Our tool aims to evolve a 3D framework from the fusion of conventional lateral and posteroanterior (PA) X‑ray images for performing 3D cephalometric analysis. Lateral and PA image datasets of 27 patients were used to validate the proposed tool. A total of 29 cephalometric landmarks were plotted on each image, and therefore, a total of 406 measurements were computed on each patient's image. An image registration-based method was implemented using the MATLAB (MathWorks, Natick, MA, USA) programming language for the prediction of 3D landmarks and computation of cephalometric measurements. The same 406 measurements were compared with measurements made on corresponding CT images. Statistical parameters, e.g., average error, standard deviation, t‑test, and correlation coefficient were computed to evaluate the performance of the proposed tool. The average mean error and average standard deviation of the 406 measurements were 2.40 and 1.77 mm, respectively. Using the proposed tool, a 3D cephalometric analysis can be performed using two X‑ray images, avoiding the need for CT/CBCT images. The proposed tool is the first method to evolve 3D cephalometric landmarks and measurements from two conventional X‑ray images, and the overall results were satisfactory.