Semi-automatic Cephalometric Landmark Detection on X-ray Images Using Deep Learning Method

Abstract

Accurate quantitative cephalometry plays a significant role in both diagnosis and treatment. However, it is tedious work and time consuming to label cephalometric landmarks manually. It also requires professional doctors to implement. Deep learning has been a hot topic in recent years and has achieved great success in many image classification and image regression problems. In this paper, we propose a semi-automatic method for detection of cephalometric landmarks using deep learning. We first roughly extract an ROI region for each landmark manually. Then we utilize Resnet50, which is a state-of-the-art convolutional neural network, to detect the landmark in the ROI region. The network directly output the coordinates of the landmark. The dataset we used in this paper is a public dataset—ISBI 2015 grand challenge in dental x-ray image analysis. Experiments demonstrated that the proposed method achieved better results compared with state-of-the-art methods.