Artificial intelligence system for automated landmark localization and analysis of cephalometry

Abstract

Objectives: Cephalometric analysis is essential for diagnosis, treatment planning and outcome assessment of orthodontics and orthognathic surgery. Utilizing artificial intelligence (AI) to achieve automated landmark localization has proved feasible and convenient. However, current systems remain insufficient for clinical application, as patients exhibit various malocclusions in cephalograms produced by different manufacturers while limited cephalograms were applied to train AI in these systems. Methods: A robust and clinically applicable artificial intelligence system was proposed for automatic cephalometric analysis. First, 9870 cephalograms taken by different radiography machines with various malocclusions of patients were collected from 20 medical institutions. Then 30 landmarks of all these cephalogram samples were manually annotated to train an artificial intelligence system, composed of a two-stage convolutional neural network (CNN) and a software-as-a-service (SaaS) system. Further, more than 100 orthodontists participated to refine the AI-output landmark localizations and re-train this system. Results: The average landmark prediction error of this system was as low as 0.94 ± 0.74 mm and the system achieved an average classification accuracy of 89.33%. Conclusions: An automatic cephalometric analysis system based on CNN was proposed, which can realize automatic landmark location and cephalometric measurements classification. This system showed promise in improving diagnostic efficiency in clinical circumstances.

Application of Artificial Intelligence (AI) in a Cephalometric Analysis: A Narrative Review

“…Table 1 presents the summarised studies on the application of AI in cephalometric analysis. In total, 23 articles were included based on both AI algorithms designed by their authors for the purpose of a specific study [ 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 ] and web-based software available on search engines and mobile applications [ 26 , 27 , 28 , 29 , 30 , 31 , 32 , 33 ]. The studies focused on comparing the reliability of AI algorithms in localising cephalometric landmarks on lateral cephalometric radiographs with the manual tracing of these points; differences between various algorithms were also examined [ 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 , 30 , 31 , 32 , 33 ].…”

“…Recently, Jiang et al developed a novel and accurate system for automatic cephalometric landmark location and analysis based on a two-stage cascade CephNet system [ 33 ]. The system consisted of two-stage neural networks, from which the first aimed to detect 10 regions of interest (ROI), each containing 1–9 landmarks, and in the second stage, the landmarks were accurately located in the ROIs.…”

Deep Learning-based Assessment of Facial Asymmetry Using U-Net Deep Convolutional Neural Network Algorithm

“…Several studies on ANN in orthodontics have reported their results on automated cephalometric landmark identification 12,18–25 . Furthermore, commercial orthodontic products for orthodontic diagnosis and treatment evaluation are already in clinical use.…”

“…Several studies on ANN in orthodontics have reported their results on automated cephalometric landmark identification. 12,[18][19][20][21][22][23][24][25] Furthermore, commercial orthodontic products for orthodontic diagnosis and treatment evaluation are already in clinical use. Clinicians can choose these AI products and benefit from orthodontic treatment and clinical practice.…”

Evaluation of deep learning and convolutional neural network algorithms accuracy for detecting and predicting anatomical landmarks on 2D lateral cephalometric images: A systematic review and meta-analysis

“… Author/year Country Architecture Objective Sample size SDR (successful detection rate) Alshamrani et al (2022) ( Alshamrani et al, 2022 ) Saudi Arabia CNN (autoencoder-based Inception layers) Generate a Bjork–Jarabak and Ricketts cephalometrics automatically. 100 Basic autoencoder model trained on Set 1 2.0 mm: 64% 2.5 mm: 69% 3.0 mm: 72% 4.0 mm: 77% 150 Model autoencoder wider Paddup box set 2 2.0 mm: 71% 2.5 mm: 75% 3.0 mm: 78% 4.0 mm: 84% El-Fegh et al (2008) ( El-Fegh et al, 2008 ) Libya/ Canada CNN A new approach to cephalometric X-ray landmark localization > 80 2.0 mm: 91% El-Feghi et al (2003) ( El-Feghi et al, 2003 ) Canada MLP A novel algorithm based on the use of the Multi-layer Perceptron (MLP) to locate landmarks on a digitized X-ray of the skull 134 2.0 mm: 91.6% Hwang et al (2021) ( Hwang et al, 2021 ) South Korea CNN (YOLO version 3) To compare an automated cephalometric analysis based on the latest deep learning method 200 2.0 mm: 75.45% 2.5 mm: 83.66% 3.0 mm: 88.92% 4.0 mm: 94.24% Jiang et al (2023) ( Jiang et al, 2023 ) China CNN (A cascade framework “CephNet”) Utilizing artificial intelligence (AI) to achieve automated landmark localization in patients with various malocclusions 259 1.0 mm: 66.15% 2.0 mm: 91.73% 3.0 mm: 97.99% Kafieh et al (2009) ( Kafieh et al, 2009 ) Iran ASM As a new method for automatic landmark detection in cephalometry, they propose two different methods for bony structure discrimination in cephalograms. 63 1.0 mm: 24.00% 2.0 mm: 61.00% ...…”