Fully automated quantitative cephalometry using convolutional neural networks

Abstract

Abstract. Quantitative cephalometry plays an essential role in clinical diagnosis, treatment, and surgery. Development of fully automated techniques for these procedures is important to enable consistently accurate computerized analyses. We study the application of deep convolutional neural networks (CNNs) for fully automated quantitative cephalometry for the first time. The proposed framework utilizes CNNs for detection of landmarks that describe the anatomy of the depicted patient and yield quantitative estimation of pathologies in the jaws and skull base regions. We use a publicly available cephalometric x-ray image dataset to train CNNs for recognition of landmark appearance patterns. CNNs are trained to output probabilistic estimations of different landmark locations, which are combined using a shape-based model. We evaluate the overall framework on the test set and compare with other proposed techniques. We use the estimated landmark locations to assess anatomically relevant measurements and classify them into different anatomical types. Overall, our results demonstrate high anatomical landmark detection accuracy (∼1% to 2% higher success detection rate for a 2-mm range compared with the top benchmarks in the literature) and high anatomical type classification accuracy (∼76% average classification accuracy for test set). We demonstrate that CNNs, which merely input raw image patches, are promising for accurate quantitative cephalometry.

Automated Identification of Cephalometric Landmarks: Part 2- Might It Be Better Than human?

“…3 The performance of an automated identification system (AI) has traditionally been compared by the successful detection rates of 19 skeletal landmarks with a 2-mm range, which has conventionally been accepted as a clinical error range at AI performance competitions. [4][5][6] Rather than again comparing certain AI techniques to other AI techniques to determine which were more accurate, the present study proposed a new automatic identification method and tested whether this new AI method was better and more reliable than clinically experienced human experts. This could be more interesting and actually applicable to clinicians.…”

Automatic three-dimensional cephalometric annotation system using three-dimensional convolutional neural networks: a developmental trial

“…It has shown outstanding performance in solving many problems in computer vision and biomedical applications. (Giordano et al, 2005;LeCun, Bengio, & Hinton, 2015;Wang et al, 2016) Lindner et al (Lindner et al, 2016) Arik et al (Arik et al, 2017), conducted deep learning-based 2D cephalometry using CNN, a rapidly developing deep learning algorithm that uses a variation of multilayer perceptrons (LeCun, 2015), inspired by the connectivity of the biological nervous system. (Huang & LeCun, 2006) It is especially suitable for image processing and recognition in that it exploits spatial correlations by imposing local connectivity patterns.…”

Cephalometric Landmark Detection by Attentive Feature Pyramid Fusion and Regression-Voting

“…We compare our approach with the top two methods [7,20] in IEEE ISBI 2015 Challenge and two new approaches proposed by Arik et al [11] and Payer et al [22]. We also remove the AFPF module and the attention mechanism respectively to do the ablation study.…”

“…Two frameworks [6,7] combining the random forests regression-voting and the statistical shape analysis techniques perform well in the IEEE ISBI 2014 and 2015 Challenges [8,9]. Since then, almost all the methods are benchmarked against the Grand Challenges dataset [10,11,22,23]. There are also some hybrid-based methods [12] integrating different techniques mentioned above.…”