Abstract
Cephalometric landmark detection is a crucial step in orthodontic and orthognathic treatments. To detect cephalometric landmarks accurately, we propose a novel multi-head attention neural network (CephaNN). CephaNN is an end-to-end network based on the heatmaps of annotated landmarks, and it consists of two parts, the multi-head part and the attention part. In the multi-head part, we adopt multi-head subnets to gain comprehensive knowledge of various subspaces of a cephalogram. The intermediate supervision is applied to accelerate the convergence. Based on the feature maps learned from the multi-head Part, the attention part applies the multi-attention mechanism to obtain a refined detection. For solving the class imbalance problem, we propose a region enhancing (RE) loss, to enhance the efficient regions on the regressed heatmaps. Experiments in the benchmark dataset demonstrate that CephaNN is state-of-the-art with the detection accuracy of 87.61% in the clinically accepted 2.0-mm range. Furthermore, CephaNN is efficient in classifying the anatomical types and robust in a real application on a 75-landmark dataset.
Highlights
Cephalometric landmark annotation is an important procedure in orthodontics and orthognathic treatment [1]
In the Test1 dataset, CephaNN with ResNeXt50 has the best performance in the 2.0-mm region with success detection rate (SDR) improving 1.89%, 2.52% and mean radial error (MRE) decreasing 0.13 mm, 0.12 mm compared with ResNet50 and VGG19
In the multi-head part, we adopt multi-head subnets to learn different features from various aspects, and we apply intermediate supervision to improve the performance and convergence speed
Summary
Cephalometric landmark annotation is an important procedure in orthodontics and orthognathic treatment [1]. In 2014 and 2015, Wang et al [2], [3] organized two automatic cephalometric landmark detection challenges They provided a public dataset containing 400 high-resolution X-ray images with annotations at the IEEE ISBI. Lindner and Cootes [5] proposed an automatic detection algorithm based on the Random Forest regression-voting method, which won the automatic cephalometric landmark detection challenge Their method requires complex human-designed features for better performance. Despite the significant progress made by deep learning methods, there are still some problems in cephalometric landmark detection, such as class imbalance and coarse detected results. We propose a novel multi-head attention neural network, CephaNN, for cephalometric landmark detection. CephaNN is an end-to-end network based on landmark heatmaps, which consists of the multi-head part and the attention part, for a coarse-to-fine detection.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have