Accuracy of an automated method of 3D soft tissue landmark detection

Photographic Parameters in Three-Dimensional Facial Image Acquisition. A Scoping Review

OBJECTIVE

Orthognathic surgery is a viable and reproducible treatment for facial deformities. Despite the precision of the skeletal planning of surgical procedures, there is little information about the relations between hard and soft tissues in three-dimensional (3D) analysis, resulting in unpredictable soft tissue outcomes. Three-dimensional photography is a viable tool for soft tissue analysis because it is easy to use, has wide availability, low cost, and is harmless. This review aims to establish parameters for acquiring consistent and reproducible 3D facial images.

METHODS

A scoping review was conducted across PubMed, SCOPUS, Scientific Electronic Library Online (SciELO), and Web of Science databases, adhering to "Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews" guidelines. Articles presenting 3D facial photographs in the diagnostic phase were considered.

RESULTS

A total of 79 articles were identified, of which 29 were selected for analysis.

CONCLUSION

The predominant use of automated systems like 3dMD and VECTRA M3 was noted. User positioning has highest agreement among authors. Noteworthy aspects include the importance of proper lighting, facial expression, and dental positioning, with observed discrepancies and inconsistencies among authors. Finally, the authors proposed a 3D image acquisition protocol based on this research findings.

Fully automated landmarking and facial segmentation on 3D photographs

Three-dimensional facial stereophotogrammetry provides a detailed representation of craniofacial soft tissue without the use of ionizing radiation. While manual annotation of landmarks serves as the current gold standard for cephalometric analysis, it is a time-consuming process and is prone to human error. The aim in this study was to develop and evaluate an automated cephalometric annotation method using a deep learning-based approach. Ten landmarks were manually annotated on 2897 3D facial photographs. The automated landmarking workflow involved two successive DiffusionNet models. The dataset was randomly divided into a training and test dataset. The precision of the workflow was evaluated by calculating the Euclidean distances between the automated and manual landmarks and compared to the intra-observer and inter-observer variability of manual annotation and a semi-automated landmarking method. The workflow was successful in 98.6% of all test cases. The deep learning-based landmarking method achieved precise and consistent landmark annotation. The mean precision of 1.69 ± 1.15 mm was comparable to the inter-observer variability (1.31 ± 0.91 mm) of manual annotation. Automated landmark annotation on 3D photographs was achieved with the DiffusionNet-based approach. The proposed method allows quantitative analysis of large datasets and may be used in diagnosis, follow-up, and virtual surgical planning.

The Reliability of Three-Dimensional Landmark-Based Craniomaxillofacial and Airway Cephalometric Analysis

Cephalometric analysis is a standard diagnostic tool in orthodontics and craniofacial surgery. Today, as conventional 2D cephalometry is limited and susceptible to analysis bias, a more reliable and user-friendly three-dimensional system that includes hard tissue, soft tissue, and airways is demanded in clinical practice. We launched our study to develop such a system based on CT data and landmarks. This study aims to determine whether the data labeled through our process is highly qualified and whether the soft tissue and airway data derived from CT scans are reliable. We enrolled 15 patients (seven males, eight females, 26.47 ± 3.44 years old) diagnosed with either non-syndromic dento-maxillofacial deformities or OSDB in this study to evaluate the intra- and inter-examiner reliability of our system. A total of 126 landmarks were adopted and divided into five sets by region: 28 cranial points, 25 mandibular points, 20 teeth points, 48 soft tissue points, and 6 airway points. All the landmarks were labeled by two experienced clinical practitioners, either of whom had labeled all the data twice at least one month apart. Furthermore, 78 parameters of three sets were calculated in this study: 42 skeletal parameters (23 angular and 19 linear), 27 soft tissue parameters (9 angular and 18 linear), and 9 upper airway parameters (2 linear, 4 areal, and 3 voluminal). Intraclass correlation coefficient (ICC) was used to evaluate the inter-examiner and intra-examiner reliability of landmark coordinate values and measurement parameters. The overwhelming majority of the landmarks showed excellent intra- and inter-examiner reliability. For skeletal parameters, angular parameters indicated better reliability, while linear parameters performed better for soft tissue parameters. The intra- and inter-examiner ICCs of airway parameters referred to excellent reliability. In summary, the data labeled through our process are qualified, and the soft tissue and airway data derived from CT scans are reliable. Landmarks that are not commonly used in clinical practice may require additional attention while labeling as they are prone to poor reliability. Measurement parameters with values close to 0 tend to have low reliability. We believe this three-dimensional cephalometric system would reach clinical application.

An Automated Method of 3D Facial Soft Tissue Landmark Prediction Based on Object Detection and Deep Learning

BACKGROUND

Three-dimensional facial soft tissue landmark prediction is an important tool in dentistry, for which several methods have been developed in recent years, including a deep learning algorithm which relies on converting 3D models into 2D maps, which results in the loss of information and precision.

METHODS

This study proposes a neural network architecture capable of directly predicting landmarks from a 3D facial soft tissue model. Firstly, the range of each organ is obtained by an object detection network. Secondly, the prediction networks obtain landmarks from the 3D models of different organs.

RESULTS

The mean error of this method in local experiments is 2.62±2.39, which is lower than that in other machine learning algorithms or geometric information algorithms. Additionally, over 72% of the mean error of test data falls within ±2.5 mm, and 100% falls within 3 mm. Moreover, this method can predict 32 landmarks, which is higher than any other machine learning-based algorithm.

CONCLUSIONS

According to the results, the proposed method can precisely predict a large number of 3D facial soft tissue landmarks, which gives the feasibility of directly using 3D models for prediction.

Artificial Intelligence Systems Assisting in the Assessment of the Course and Retention of Orthodontic Treatment

This scoping review examines the contemporary applications of advanced artificial intelligence (AI) software in orthodontics, focusing on its potential to improve daily working protocols, but also highlighting its limitations. The aim of the review was to evaluate the accuracy and efficiency of current AI-based systems compared to conventional methods in diagnosing, assessing the progress of patients' treatment and follow-up stability. The researchers used various online databases and identified diagnostic software and dental monitoring software as the most studied software in contemporary orthodontics. The former can accurately identify anatomical landmarks used for cephalometric analysis, while the latter enables orthodontists to thoroughly monitor each patient, determine specific desired outcomes, track progress, and warn of potential changes in pre-existing pathology. However, there is limited evidence to assess the stability of treatment outcomes and relapse detection. The study concludes that AI is an effective tool for managing orthodontic treatment from diagnosis to retention, benefiting both patients and clinicians. Patients find the software easy to use and feel better cared for, while clinicians can make diagnoses more easily and assess compliance and damage to braces or aligners more quickly and frequently.

Three-dimensional soft tissue landmark detection with marching cube algorithm

Current method of analyzing three-dimensional soft tissue data, especially in the frontal view, is subjective and has poor reliability. To overcome this limitation, the present study aimed to introduce a new method of analyzing soft tissue data reconstructed by marching cube algorithm (Program S) and compare it with a commercially available program (Program A). Cone-beam computed tomography images of 42 patients were included. Two orthodontists digitized six landmarks (pronasale, columella, upper and lower lip, right and left cheek) twice using both programs in two-week intervals, and the reliability was compared. Furthermore, computer-calculated point (CC point) was developed to evaluate whether human error could be reduced. The results showed that the intra- and inter-examiner reliability of Program S (99.7-100% and 99.9-100%, respectively) were higher than that of Program A (64.0-99.9% and 76.1-99.9%, respectively). Moreover, the inter-examiner difference of coordinate values and distances for all six landmarks in Program S was lower than Program A. Lastly, CC point was provided as a consistent single point. Therefore, it was validated that this new methodology can increase the intra- and inter-examiner reliability of soft tissue landmark digitation and CC point can be used as a landmark to reduce human error.

3D Surface Topographic Optical Scans Yield Highly Reliable Global Spine Range of Motion Measurements in Scoliotic and Non-Scoliotic Adolescents

Background: Adolescent idiopathic scoliosis results in three dimensional changes to a patient’s body, which may change a patient’s range of motion. Surface topography is an emerging technology to evaluate three dimensional parameters in patients with scoliosis. The goal of this paper is to introduce novel and reliable surface topographic measurements for the assessment of global coronal and sagittal range of motion of the spine in adolescents, and to determine if these measurements can distinguish between adolescents with lumbar scoliosis and those without scoliosis. Methods: This study is a retrospective cohort study of a prospectively collected registry. Using a surface topographic scanner, a finger to floor and lateral bending scans were performed on each subject. Inter- and intra-rater reliabilities were assessed for each measurement. ANOVA analysis was used to test comparative hypotheses. Results: Inter-rater reliability for lateral bending fingertip asymmetry (LBFA) and lateral bending acromia asymmetry (LBAA) displayed poor reliability, while the coronal angle asymmetry (CAA), coronal angle range of motion (CAR), forward bending finger to floor (FBFF), forward bending acromia to floor (FBAF), sagittal angle (SA), and sagittal angle normalized (SAN) demonstrated good to excellent reliability. There was a significant difference between controls and lumbar scoliosis patients for LBFA, LBAA, CAA and FBAF (p-values < 0.01). Conclusion: Surface topography yields a reliable and rapid process for measuring global spine range of motion in the coronal and sagittal planes. Using these tools, there was a significant difference in measurements between patients with lumbar scoliosis and controls. In the future, we hope to be able to assess and predict perioperative spinal mobility changes.

Use of Advanced Artificial Intelligence in Forensic Medicine, Forensic Anthropology and Clinical Anatomy

Three-dimensional convolutional neural networks (3D CNN) of artificial intelligence (AI) are potent in image processing and recognition using deep learning to perform generative and descriptive tasks. Compared to its predecessor, the advantage of CNN is that it automatically detects the important features without any human supervision. 3D CNN is used to extract features in three dimensions where input is a 3D volume or a sequence of 2D pictures, e.g., slices in a cone-beam computer tomography scan (CBCT). The main aim was to bridge interdisciplinary cooperation between forensic medical experts and deep learning engineers, emphasizing activating clinical forensic experts in the field with possibly basic knowledge of advanced artificial intelligence techniques with interest in its implementation in their efforts to advance forensic research further. This paper introduces a novel workflow of 3D CNN analysis of full-head CBCT scans. Authors explore the current and design customized 3D CNN application methods for particular forensic research in five perspectives: (1) sex determination, (2) biological age estimation, (3) 3D cephalometric landmark annotation, (4) growth vectors prediction, (5) facial soft-tissue estimation from the skull and vice versa. In conclusion, 3D CNN application can be a watershed moment in forensic medicine, leading to unprecedented improvement of forensic analysis workflows based on 3D neural networks.

Influence of the Depth of the Convolutional Neural Networks on an Artificial Intelligence Model for Diagnosis of Orthognathic Surgery

The aim of this study was to investigate the relationship between image patterns in cephalometric radiographs and the diagnosis of orthognathic surgery and propose a method to improve the accuracy of predictive models according to the depth of the neural networks. The study included 640 and 320 patients requiring non-surgical and surgical orthodontic treatments, respectively. The data of 150 patients were exclusively classified as a test set. The data of the remaining 810 patients were split into five groups and a five-fold cross-validation was performed. The convolutional neural network models used were ResNet-18, 34, 50, and 101. The number in the model name represents the difference in the depth of the blocks that constitute the model. The accuracy, sensitivity, and specificity of each model were estimated and compared. The average success rate in the test set for the ResNet-18, 34, 50, and 101 was 93.80%, 93.60%, 91.13%, and 91.33%, respectively. In screening, ResNet-18 had the best performance with an area under the curve of 0.979, followed by ResNets-34, 50, and 101 at 0.974, 0.945, and 0.944, respectively. This study suggests the required characteristics of the structure of an artificial intelligence model for decision-making based on medical images.