Accuracy of automated 3D cephalometric landmarks by deep learning algorithms: systematic review and meta-analysis

Abstract

Objectives The aim of the present systematic review and meta-analysis is to assess the accuracy of automated landmarking using deep learning in comparison with manual tracing for cephalometric analysis of 3D medical images. Methods PubMed/Medline, IEEE Xplore, Scopus and ArXiv electronic databases were searched. Selection criteria were: ex vivo and in vivo volumetric data images suitable for 3D landmarking (Problem), a minimum of five automated landmarking performed by deep learning method (Intervention), manual landmarking (Comparison), and mean accuracy, in mm, between manual and automated landmarking (Outcome). QUADAS-2 was adapted for quality analysis. Meta-analysis was performed on studies that reported as outcome mean values and standard deviation of the difference (error) between manual and automated landmarking. Linear regression plots were used to analyze correlations between mean accuracy and year of publication. Results The initial electronic screening yielded 252 papers published between 2020 and 2022. A total of 15 studies were included for the qualitative synthesis, whereas 11 studies were used for the meta-analysis. Overall random effect model revealed a mean value of 2.44 mm, with a high heterogeneity (I2 = 98.13%, τ2 = 1.018, p-value < 0.001); risk of bias was high due to the presence of issues for several domains per study. Meta-regression indicated a significant relation between mean error and year of publication (p value = 0.012). Conclusion Deep learning algorithms showed an excellent accuracy for automated 3D cephalometric landmarking. In the last two years promising algorithms have been developed and improvements in landmarks annotation accuracy have been done.

Automated Landmark Annotation for Morphometric Analysis of Distal Femur and Proximal Tibia

“…Additionally, considering the data-driven nature of our method, a larger training dataset could be beneficial for automatic landmarking accuracy. However, similar validation studies report acceptable to excellent results based on similar sample sizes [20,[24][25][26].…”

Fully automated landmarking and facial segmentation on 3D photographs

“…An automated landmarking method, based on multi-stage deep reinforcement learning and volume-rendered imaging, was proposed by Kang et al and yielded a precision of 1.96 ± 0.78 mm 20 . A systematic review by Serafin et al found a mean precision of 2.44 mm for the prediction of 3D hard-tissue landmarks from CT and CBCT images 5 .…”

“…Cephalometric analysis can be performed on 3D stereophotographs to extract information about the position of individual landmarks or distances and angles between several landmarks, with the purpose of objectifying clinical observations 5 . Despite being a commonly used diagnostic tool in the craniofacial region, landmarking often remains a manual task that is time-consuming, prone to observer variability, and affected by observer fatigue and skill level 6 , 7 .…”

“…Several studies have described the use of deep learning algorithms for the automation of hard-tissue landmark extraction for cephalometric analysis 5 , 8 . Studies that include soft-tissue landmarks utilize (projective) 2D imaging, are pose dependent, or require manual input 9 , 10 .…”

“…Only a limited number of studies were performed on the automated extraction of facial soft-tissue landmarks from 3D photographs 11 , 12 . Nevertheless, since the studies in question did not integrate deep learning algorithms into their methodologies, and considering that deep learning has demonstrated enhanced accuracy in hard tissue landmarking, the effectiveness of soft tissue landmarking could potentially see improvements through the integration of deep learning techniques 5 . Therefore, this study aimed to develop and validate an automated approach for the extraction of soft-tissue facial landmarks from 3D photographs using deep learning.…”