Cephalometric Landmark Regression with Convolutional Neural Networks on 3D Computed Tomography Data
- PDF / 1,324,466 Bytes
- 11 Pages / 612 x 792 pts (letter) Page_size
- 68 Downloads / 186 Views
Cephalometric Landmark Regression with Convolutional Neural Networks on 3D Computed Tomography Data D. Lachinova,b,*, A. Getmanskayaa,**, and V. Turlapova,*** a
Institute of Information Technologies, Mathematics and Mechanics, Lobachevsky State University, Nizhny Novgorod, Russia b Department of Ophthalmology and Optometry, Medical University of Vienna, Vienna, Austria * e-mail: [email protected] ** e-mail: [email protected] *** e-mail: [email protected] Abstract—In this paper, we address the problem of automatic three-dimensional cephalometric analysis. Cephalometric analysis performed on lateral radiographs doesn’t fully exploit the structure of 3D objects due to projection onto the lateral plane. With the development of three-dimensional imaging techniques such as CT, several analysis methods have been proposed that extend to the 3D case. The analysis based on these methods is invariant to rotations and translations and can describe difficult skull deformation, where 2D cephalometry has no use. In this paper, we provide a wide overview of existing approaches for cephalometric landmark regression. Moreover, we perform a series of experiments with state of the art 3D convolutional neural network (CNN) based methods for keypoint regression: direct regression with CNN, heatmap regression and Softargmax regression. For the first time, we extensively evaluate the described methods and demonstrate their effectiveness in the estimation of Frankfort Horizontal and cephalometric points locations for patients with severe skull deformations. We demonstrate that Heatmap and Softargmax regression models provide sufficient regression error for medical applications (less than 4 mm). Moreover, the Softargmax model achieves 1.15° inclination error for the Frankfort horizontal. For the fair comparison with the prior art, we also report results projected on the lateral plane. Keywords: 3D cephalometry, automatic cephalometry, keypoint regression, neural networks, deep learning DOI: 10.1134/S1054661820030165
1. INTRODUCTION In 1995 Jacobson in [1] suggested applying a threedimensional cephalometric analysis approach to orthodontic treatment. They started with the Digigraph—a device for recording soft tissue cephalometric points in 3D. Subsequent studies [2–4] show that threedimensional cephalometry is an important next step with many potential improvements over conventional two-dimensional radiographic approaches. The clear advantage of 3D analysis is the diagnosis and analysis accuracy of craniomaxillofacial deformities. This is due to the majority of congenital and acquired craniomaxillofacial deformities are asymmetric, and the deformation they correspond to is three-dimensional [5]. In turn, 3D analysis has its limitations: 3D scan study is time-consuming and perceptually hard. Moreover, Smektala et al. [6] argue that 3D cephalometry is not accurate or reproducible enough to be a reliable diagnostic technique. These flaws can be noticed while using linear and angular measurements in the tradi
Data Loading...
