Automated cephalometric landmark detection with confidence regions using Bayesian convolutional neural networks

Abstract

Background Despite the integral role of cephalometric analysis in orthodontics, there have been limitations regarding the reliability, accuracy, etc. of cephalometric landmarks tracing. Attempts on developing automatic plotting systems have continuously been made but they are insufficient for clinical applications due to low reliability of specific landmarks. In this study, we aimed to develop a novel framework for locating cephalometric landmarks with confidence regions using Bayesian Convolutional Neural Networks (BCNN). Methods We have trained our model with the dataset from the ISBI 2015 grand challenge in dental X-ray image analysis. The overall algorithm consisted of a region of interest (ROI) extraction of landmarks and landmarks estimation considering uncertainty. Prediction data produced from the Bayesian model has been dealt with post-processing methods with respect to pixel probabilities and uncertainties. Results Our framework showed a mean landmark error (LE) of 1.53 ± 1.74 mm and achieved a successful detection rate (SDR) of 82.11, 92.28 and 95.95%, respectively, in the 2, 3, and 4 mm range. Especially, the most erroneous point in preceding studies, Gonion, reduced nearly halves of its error compared to the others. Additionally, our results demonstrated significantly higher performance in identifying anatomical abnormalities. By providing confidence regions (95%) that consider uncertainty, our framework can provide clinical convenience and contribute to making better decisions. Conclusion Our framework provides cephalometric landmarks and their confidence regions, which could be used as a computer-aided diagnosis tool and education.

Multistage Probabilistic Approach for the Localization of Cephalometric Landmarks

“…In terms of SDR, the proposed method takes the first place when τ is 2.0 mm and 2.5 mm, the second place when τ is 3.0 mm and the third place when τ is 4.0 mm in the Test2. For Test1, the method in [11] shows the best SDR for all thresholds, however, our method follows their results when τ = 2.0 mm, which is a clinically acceptable error value in cephalometric analysis [8]. Also note that Test1 is used as a validation set in the experimental settings.…”

“…In the second stage, a set of U-Nets compute refined heatmaps of landmarks using high-resolution patches around coarse estimations. Lee et al [8] attempted to use confidence maps in this two stage approach. The first stage gives the rough estimates of landmark positions using down-sampled input images.…”

“…Automated methods also suffer from inaccurate localization [4]. Actually, the difficulty in the localization of gonion (landmark 10, one of the bilateral landmarks of the mandible) has been reported in [3], [8]. This is usually caused by asymmetry of the mandible, and we need to consider related anatomic structures for the accurate localization.…”

“…The refinement stage improves the positions of bilateral landmarks of the mandible (x P b ) with linear estimators (g b (•, •)) using x G and x L . and the probabilistic interpretation of predicted heatmap and objective functions is not clear [8].…”

Automatic Cephalometric Landmark Identification System Based on the Multi-Stage Convolutional Neural Networks with CBCT Combination Images

“…There were no significant differences between the manual and automatic prediction in the cephalometric analysis, but the average prediction errors recorded 17.02 in pixel (approximately 4.50 mm) [ 23 ]. In 2020, many studies introduced new CNNs algorithms or methods [ 44 , 45 , 46 ]. Kunz et al [ 44 ] used customized Keras and Tensorflow, similar to us.…”