Cephalometric Landmark Regression with Convolutional Neural Networks on 3D Computed Tomography Data

TL;DR

This paper evaluates deep learning methods for 3D cephalometric landmark detection in CT scans, demonstrating that heatmap and Softargmax regression achieve clinically acceptable accuracy even in severe skull deformities.

Contribution

It provides the first extensive comparison of CNN-based keypoint regression methods for 3D cephalometry, highlighting their effectiveness in challenging cases.

Findings

Heatmap and Softargmax models achieve less than 4 mm regression error.

Softargmax model attains 1.15° inclination error for Frankfort horizontal.

Methods are invariant to rotations and translations, suitable for severe skull deformations.

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 1.15o inclination error for the Frankfort horizontal. For the fair comparison with the prior art, we also report results projected on the lateral plane.