Automatic Landmark Identification Research Articles

Cephalometric analysis performance discrepancy between orthodontists and an artificial intelligence model using lateral cephalometric radiographs.

The purpose of the present clinical study was to compare the Ricketts and Steiner cephalometric analysis obtained by two experienced orthodontists and artificial intelligence (AI)-based software program and measure the orthodontist variability. A total of 50 lateral cephalometric radiographs from 50 patients were obtained. Two groups were created depending on the operator performing the cephalometric analysis: orthodontists (Orthod group) and an AI software program (AI group). In the Orthod group, two independent experienced orthodontists performed the measurements by performing a manual identification of the cephalometric landmarks and a software program (NemoCeph; Nemotec) to calculate the measurements. In the AI group, an AI software program (CephX; ORCA Dental AI) was selected for both the automatic landmark identification and cephalometric measurements. The Ricketts and Steiner cephalometric analyses were assessed in both groups including a total of 24 measurements. The Shapiro-Wilk test showed that the data was normally distributed. The t-test was used to analyze the data (α = 0.05). The t-test analysis showed significant measurement discrepancies between the Orthod and AI group in seven of the 24 cephalometric parameters tested, namely the corpus length (p = 0.003), mandibular arc (p < 0.001), lower face height (p = 0.005), overjet (p = 0.019), and overbite (p = 0.022) in the Ricketts cephalometric analysis and occlusal to SN (p = 0.002) and GoGn-SN (p < 0.001) in the Steiner cephalometric analysis. The intraclass correlation coefficient (ICC) between both orthodontists of the Orthod group for each cephalometric measurement was calculated. Significant discrepancies were found in seven of the 24 cephalometric measurements tested between the orthodontists and the AI-based program assessed. The intra-operator reliability analysis showed reproducible measurements between both orthodontists, except for the corpus length measurement. The artificial intelligence software program tested has the potential to automatically obtain cephalometric analysis using lateral cephalometric radiographs; however, additional studies are needed to further evaluate the accuracy of this AI-based system.

Automatic identification of posteroanterior cephalometric landmarks using a novel deep learning algorithm: a comparative study with human experts

This study aimed to propose a fully automatic posteroanterior (PA) cephalometric landmark identification model using deep learning algorithms and compare its accuracy and reliability with those of expert human examiners. In total, 1032 PA cephalometric images were used for model training and validation. Two human expert examiners independently and manually identified 19 landmarks on 82 test set images. Similarly, the constructed artificial intelligence (AI) algorithm automatically identified the landmarks on the images. The mean radial error (MRE) and successful detection rate (SDR) were calculated to evaluate the performance of the model. The performance of the model was comparable with that of the examiners. The MRE of the model was 1.87 ± 1.53 mm, and the SDR was 34.7%, 67.5%, and 91.5% within error ranges of < 1.0, < 2.0, and < 4.0 mm, respectively. The sphenoid points and mastoid processes had the lowest MRE and highest SDR in auto-identification; the condyle points had the highest MRE and lowest SDR. Comparable with human examiners, the fully automatic PA cephalometric landmark identification model showed promising accuracy and reliability and can help clinicians perform cephalometric analysis more efficiently while saving time and effort. Future advancements in AI could further improve the model accuracy and efficiency.

Automatic landmark identification in cone-beam computed tomography.

To present and validate an open-source fully automated landmark placement (ALICBCT) tool for cone-beam computed tomography scans. One hundred and forty-three large and medium field of view cone-beam computed tomography (CBCT) were used to train and test a novel approach, called ALICBCT that reformulates landmark detection as a classification problem through a virtual agent placed inside volumetric images. The landmark agents were trained to navigate in a multi-scale volumetric space to reach the estimated landmark position. The agent movements decision relies on a combination of DenseNet feature network and fully connected layers. For each CBCT, 32 ground truth landmark positions were identified by 2 clinician experts. After validation of the 32 landmarks, new models were trained to identify a total of 119 landmarks that are commonly used in clinical studies for the quantification of changes in bone morphology and tooth position. Our method achieved a high accuracy with an average of 1.54 ± 0.87 mm error for the 32 landmark positions with rare failures, taking an average of 4.2second computation time to identify each landmark in one large 3D-CBCT scan using a conventional GPU. The ALICBCT algorithm is a robust automatic identification tool that has been deployed for clinical and research use as an extension in the 3D Slicer platform allowing continuous updates for increased precision.

Automated landmark identification for diagnosis of the deformity using a cascade convolutional neural network (FlatNet) on weight-bearing lateral radiographs of the foot

Landmark detection in flatfoot radiographs is crucial in analyzing foot deformity. Here, we evaluated the accuracy and efficiency of the automated identification of flatfoot landmarks using a newly developed cascade convolutional neural network (CNN) algorithm, Flatfoot Landmarks AnnoTating Network (FlatNet). A total of 1200 consecutive weight-bearing lateral radiographs of the foot were acquired. The first 1050 radiographs were used as the training and tuning, and the following 150 radiographs were used as the test sets, respectively. An expert orthopedic surgeon (A) manually labeled ground truths for twenty-five anatomical landmarks. Two orthopedic surgeons (A and B, each with eight years of clinical experience) and a general physician (GP) independently identified the landmarks of the test sets using the same method. After two weeks, observers B and GP independently identified the landmarks once again using the developed deep learning CNN model (DLm). The X- and Y-coordinates and the mean absolute distance were evaluated. The average differences (mm) from the ground truth were 0.60 ± 0.57, 1.37 ± 1.28, and 1.05 ± 1.23 for the X-coordinate, and 0.46 ± 0.59, 0.97 ± 0.98, and 0.73 ± 0.90 for the Y-coordinate in DLm, B, and GP, respectively. The average differences (mm) from the ground truth were 0.84 ± 0.73, 1.90 ± 1.34, and 1.42 ± 1.40 for the absolute distance in DLm, B, and GP, respectively. Under the guidance of the DLm, the overall differences (mm) from the ground truth were enhanced to 0.87 ± 1.21, 0.69 ± 0.74, and 1.24 ± 1.31 for the X-coordinate, Y-coordinate, and absolute distance, respectively, for observer B. The differences were also enhanced to 0.74 ± 0.73, 0.57 ± 0.63, and 1.04 ± 0.85 for observer GP. The newly developed FlatNet exhibited better accuracy and reliability than the observers. Furthermore, under the FlatNet guidance, the accuracy and reliability of the human observers generally improved.

Comparison between a human judge and automatic landmark identification on digital models

Accurate landmark identification is a prerequisite for accurate and reliable biomedical image analysis. Orthodontic study models are valuable tools for diagnosis, treatment planning, and maintaining complete records. The purpose of this study was to evaluate the reliability and validity of a software program (Align Technology, Inc) as a tool for automatic landmark location. Using digital intraoral scans of 10 dental arches, 4 calibrated human judges independently located cusp tips and interproximal contacts. The same landmarks were automatically identified by the software. Intraclass correlation coefficient (Cronbach α), absolute mean errors, and regression analysis were calculated. In addition, Bland-Altman 95% confidence limits were also applied to the data to graphically display agreement on landmark identification between the human judges and the software. The intraclass correlation coefficient between the software and the human judges' average for the x-, y-, and z-coordinates for all landmarks was excellent, at 1.0, 1.0, and 0.98, respectively. The regression analysis and Bland-Altman plots show no systematic errors for agreement on landmark identification between the human judges and the software. Landmark location was nearly identical between the software and the human judges, making the methods interchangeable.

Automatic landmark identification for surgical 3d-navigation- A proposed method for marker-free dental surgical navigation systems.

This paper proposes a conceptual method to calculate the pose of a stereo-vision camera relative to an artificial mandible without additional markers. The general method for marker-free navigation has four steps: 1) parallel image acquisition by a stereo-vision camera, 2) automatic identification of 2d point pairs (landmark pairs) in a left and a right image, 3) calculation of related 3d points in the joint camera coordinate system and 4) matching of 3d points generated to a preoperative 3d model (i.e.,CT data based). To identify and compare landmarks in the acquired stereo images, well-known algorithms for landmark detection, description and matching were compared within the developed approach. Finally, the BRISK algorithm (Leutenegger S, Chli M, Siegwart RY. BRISK: Binary Robust invariant scalable keypoints. Proceedings of the IEEE International Conference on Computer Vision; 2011: 2548-2555) was used. The proposed method was implemented in MATLAB® and validated in vitro with one artificial mandible. The accuracy evaluation of the camera positions calculated resulted in an average deviation error of 1.45mm±0.76mm to the real camera displacement. This value was calculated using only stereo images with over 100 reconstructed landmark pairs each. This provides the basis for marker-free navigation.

Cephalometric Analysis in Orthodontics Using Artificial Intelligence—A Comprehensive Review

Artificial intelligence (AI) is a branch of science concerned with developing programs and computers that can gather data, reason about it, and then translate it into intelligent actions. AI is a broad area that includes reasoning, typical linguistic dispensation, machine learning, and planning. In the area of medicine and dentistry, machine learning is currently the most widely used AI application. This narrative review is aimed at giving an outline of cephalometric analysis in orthodontics using AI. Latest algorithms are developing rapidly, and computational resources are increasing, resulting in increased efficiency, accuracy, and reliability. Current techniques for completely automatic identification of cephalometric landmarks have considerably improved efficiency and growth prospects for their regular use. The primary considerations for effective orthodontic treatment are an accurate diagnosis, exceptional treatment planning, and good prognosis estimation. The main objective of the AI technique is to make dentists' work more precise and accurate. AI is increasingly being used in the area of orthodontic treatment. It has been evidenced to be a time-saving and reliable tool in many ways. AI is a promising tool for facilitating cephalometric tracing in routine clinical practice and analyzing large databases for research purposes.

Automated landmark identification on cone-beam computed tomography: Accuracy and reliability.

To evaluate the accuracy and reliability of a fully automated landmark identification (ALI) system as a tool for automatic landmark location compared with human judges. A total of 100 cone-beam computed tomography (CBCT) images were collected. After the calibration procedure, two human judges identified 53 landmarks in the x, y, and z coordinate planes on CBCTs using Checkpoint Software (Stratovan Corporation, Davis, Calif). The ground truth was created by averaging landmark coordinates identified by two human judges for each landmark. To evaluate the accuracy of ALI, the mean absolute error (mm) at the x, y, and z coordinates and mean error distance (mm) between the human landmark identification and the ALI were determined, and a successful detection rate was calculated. Overall, the ALI system was as successful at landmarking as the human judges. The ALI's mean absolute error for all coordinates was 1.57 mm on average. Across all three coordinate planes, 94% of the landmarks had a mean absolute error of less than 3 mm. The mean error distance for all 53 landmarks was 3.19 ± 2.6 mm. When applied to 53 landmarks on 100 CBCTs, the ALI system showed a 75% success rate in detecting landmarks within a 4-mm error distance range. Overall, ALI showed clinically acceptable mean error distances except for a few landmarks. The ALI was more precise than humans when identifying landmarks on the same image at different times. This study demonstrates the promise of ALI in aiding orthodontists with landmark identifications on CBCTs.

Accuracy of Automatic Cephalometric Analysis Programs on Lateral Cephalograms of Preadolescent Children

The aim of this study was to evaluate the accuracy of 3 different automatic landmark identification programs on lateral cephalgrams and the clinical acceptability in pediatric dentistry. Sixty digital cephalometric radiographs of 7 to 12 years old healthy children were randomly selected. Fourteen landmarks were chosen for assessment and the mean of 3 measurements of each landmark by a single examiner was defined as the baseline landmarks. The mean difference between an automatically identified landmark and the baseline landmark was measured for each landmark on each image. The total mean difference of 3 automatic programs compared to the baseline landmarks were 2.53 ± 1.63 mm. Errors among 3 programs were not significantly different for 12 of 14 landmarks except Orbitale and Gonion. The automatic landmark identification programs showed significant higher mean detection errors than the manual method. The programs couldn’t be used as the 1st tool to replace human examiners. But considering short consuming time, these results indicate that all 3 programs have sufficient validity to be used in pediatric dental clinic.

Geometric and dosimetric accuracy of deformable image registration between average-intensity images for 4DCT-based adaptive radiotherapy for non-small cell lung cancer.

PurposeRe‐planning for four‐dimensional computed tomography (4DCT)‐based lung adaptive radiotherapy commonly requires deformable dose mapping between the planning average‐intensity image (AVG) and the newly acquired AVG. However, such AVG‐AVG deformable image registration (DIR) lacks accuracy assessment. The current work quantified and compared geometric accuracies of AVG‐AVG DIR and corresponding phase‐phase DIRs, and subsequently investigated the clinical impact of such AVG‐AVG DIR on deformable dose mapping.Methods and MaterialsHybrid intensity‐based AVG‐AVG and phase‐phase DIRs were performed between the planning and mid‐treatment 4DCTs of 28 non‐small cell lung cancer patients. An automated landmark identification algorithm detected vessel bifurcation pairs in both lungs. Target registration error (TRE) of these landmark pairs was calculated for both DIR types. The correlation between TRE and respiratory‐induced landmark motion in the planning 4DCT was analyzed. Global and local dose metrics were used to assess the clinical implications of AVG‐AVG deformable dose mapping with both DIR types.ResultsTRE of AVG‐AVG and phase‐phase DIRs averaged 3.2 ± 1.0 and 2.6 ± 0.8 mm respectively (p < 0.001). Using AVG‐AVG DIR, TREs for landmarks with <10 mm motion averaged 2.9 ± 2.0 mm, compared to 3.1 ± 1.9 mm for the remaining landmarks (p < 0.01). Comparatively, no significant difference was demonstrated for phase‐phase DIRs. Dosimetrically, no significant difference in global dose metrics was observed between doses mapped with AVG‐AVG DIR and the phase‐phase DIR, but a positive linear relationship existed (p = 0.04) between the TRE of AVG‐AVG DIR and local dose difference.ConclusionsWhen the region of interest experiences <10 mm respiratory‐induced motion, AVG‐AVG DIR may provide sufficient geometric accuracy; conversely, extra attention is warranted, and phase‐phase DIR is recommended. Dosimetrically, the differences in geometric accuracy between AVG‐AVG and phase‐phase DIRs did not impact global lung‐based metrics. However, as more localized dose metrics are needed for toxicity assessment, phase‐phase DIR may be required as its lower mean TRE improved voxel‐based dosimetry.

Comparison of cephalometric measurements between conventional and automatic cephalometric analysis using convolutional neural network

ObjectiveThe rapid development of artificial intelligence technologies for medical imaging has recently enabled automatic identification of anatomical landmarks on radiographs. The purpose of this study was to compare the results of an automatic cephalometric analysis using convolutional neural network with those obtained by a conventional cephalometric approach.Material and methodsCephalometric measurements of lateral cephalograms from 35 patients were obtained using an automatic program and a conventional program. Fifteen skeletal cephalometric measurements, nine dental cephalometric measurements, and two soft tissue cephalometric measurements obtained by the two methods were compared using paired t test and Bland-Altman plots.ResultsA comparison between the measurements from the automatic and conventional cephalometric analyses in terms of the paired t test confirmed that the saddle angle, linear measurements of maxillary incisor to NA line, and mandibular incisor to NB line showed statistically significant differences. All measurements were within the limits of agreement based on the Bland-Altman plots. The widths of limits of agreement were wider in dental measurements than those in the skeletal measurements.ConclusionsAutomatic cephalometric analyses based on convolutional neural network may offer clinically acceptable diagnostic performance. Careful consideration and additional manual adjustment are needed for dental measurements regarding tooth structures for higher accuracy and better performance.

Proposition of local automatic algorithm for landmark detection in 3D cephalometry

This study proposes a new contribution to solve the problem of automatic landmarks detection in three-dimensional cephalometry. 3D images obtained from CBCT (cone beam computed tomography) equipment were used for automatic identification of twelve landmarks. The proposed method is based on a local geometry and intensity criteria of skull structures. After the step of preprocessing and binarization, the algorithm segments the skull into three structures using the geometry information of nasal cavity and intensity information of the teeth. Each targeted landmark was detected using local geometrical information of the volume of interest containing this landmark. The ICC and confidence interval (95% CI) for each direction were 0, 91 (0.75 to 0.96) for x- direction; 0.92 (0.83 to 0.97) for y-direction; 0.92 (0.79 to 0.97) for z-direction. The mean error of detection was calculated using the Euclidian distance between the 3D coordinates of manually and automatically detected landmarks. The overall mean error of the algorithm was 2.76 mm with a standard deviation of 1.43 mm. Our proposed approach for automatic landmark identification in 3D cephalometric was capable of detecting 12 landmarks on 3D CBCT images which can be facilitate the use of 3D cephalometry to orthodontists.

Evaluation of a multi-stage convolutional neural network-based fully automated landmark identification system using cone-beam computed tomographysynthesized posteroanterior cephalometric images.

ObjectiveTo evaluate the accuracy of a multi-stage convolutional neural network (CNN) model-based automated identification system for posteroanterior (PA) cephalometric landmarks.MethodsThe multi-stage CNN model was implemented with a personal computer. A total of 430 PA-cephalograms synthesized from cone-beam computed tomography scans (CBCT-PA) were selected as samples. Twenty-three landmarks used for Tweemac analysis were manually identified on all CBCT-PA images by a single examiner. Intra-examiner reproducibility was confirmed by repeating the identification on 85 randomly selected images, which were subsequently set as test data, with a two-week interval before training. For initial learning stage of the multi-stage CNN model, the data from 345 of 430 CBCT-PA images were used, after which the multi-stage CNN model was tested with previous 85 images. The first manual identification on these 85 images was set as a truth ground. The mean radial error (MRE) and successful detection rate (SDR) were calculated to evaluate the errors in manual identification and artificial intelligence (AI) prediction.ResultsThe AI showed an average MRE of 2.23 ± 2.02 mm with an SDR of 60.88% for errors of 2 mm or lower. However, in a comparison of the repetitive task, the AI predicted landmarks at the same position, while the MRE for the repeated manual identification was 1.31 ± 0.94 mm.ConclusionsAutomated identification for CBCT-synthesized PA cephalometric landmarks did not sufficiently achieve the clinically favorable error range of less than 2 mm. However, AI landmark identification on PA cephalograms showed better consistency than manual identification.

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

This study was designed to develop and verify a fully automated cephalometry landmark identification system, based on multi-stage convolutional neural networks (CNNs) architecture, using a combination dataset. In this research, we trained and tested multi-stage CNNs with 430 lateral and 430 MIP lateral cephalograms synthesized by cone-beam computed tomography (CBCT) to make a combination dataset. Fifteen landmarks were manually and respectively identified by experienced examiner, at the preprocessing phase. The intra-examiner reliability was high (ICC = 0.99) in manual identification. The results of prediction of the system for average mean radial error (MRE) and standard deviation (SD) were 1.03 mm and 1.29 mm, respectively. In conclusion, different types of image data might be the one of factors that affect the prediction accuracy of a fully-automated landmark identification system, based on multi-stage CNNs.

A fully deep learning model for the automatic identification of cephalometric landmarks.

PurposeThis study aimed to propose a fully automatic landmark identification model based on a deep learning algorithm using real clinical data and to verify its accuracy considering inter-examiner variability.Materials and MethodsIn total, 950 lateral cephalometric images from Yonsei Dental Hospital were used. Two calibrated examiners manually identified the 13 most important landmarks to set as references. The proposed deep learning model has a 2-step structure—a region of interest machine and a detection machine—each consisting of 8 convolution layers, 5 pooling layers, and 2 fully connected layers. The distance errors of detection between 2 examiners were used as a clinically acceptable range for performance evaluation.ResultsThe 13 landmarks were automatically detected using the proposed model. Inter-examiner agreement for all landmarks indicated excellent reliability based on the 95% confidence interval. The average clinically acceptable range for all 13 landmarks was 1.24 mm. The mean radial error between the reference values assigned by 1 expert and the proposed model was 1.84 mm, exhibiting a successful detection rate of 36.1%. The A-point, the incisal tip of the maxillary and mandibular incisors, and ANS showed lower mean radial error than the calibrated expert variability.ConclusionThis experiment demonstrated that the proposed deep learning model can perform fully automatic identification of cephalometric landmarks and achieve better results than examiners for some landmarks. It is meaningful to consider between-examiner variability for clinical applicability when evaluating the performance of deep learning methods in cephalometric landmark identification.

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

Due to technological advances, the quantification of facial form can now be done via three-dimensional (3D) photographic systems such as stereophotogrammetry. To enable comparison with traditional cephalometry, soft-tissue anatomical landmark definitions have been modified to incorporate the third dimension. Annotating these landmarks manually, however, is still a time-consuming and arduous process. To develop an automated algorithm to accurately identify anatomical landmarks on three-dimensional soft tissue images. Thirty 3dMD images were selected from a private orthodontic practice consisting of 15 males and 15 females between 9 and 17 years of age. The soft-tissue 3D images were aligned along a reference plane to setup a Cartesian coordinate system. Screened by 2 observers, 21 landmarks were manually annotated and their coordinates defined. An automated landmark identification algorithm, based on their anatomical definitions, was developed to compare the landmark validity against the manually identified counterpart. Twenty-one landmarks were analysed in detail. Inter-observer and intra-observer reliability using ICC was >0.9. The average difference and standard deviation between manual and automated methods for all landmarks was 3.2 and 1.64 mm, respectively. Sixteen out of twenty-one landmarks had a mean difference less than 4 mm. The landmarks of greatest agreement (≤2 mm) were mainly in the midline: pronasale, subnasale, subspinale, labiale superius, stomion, with the exception of chelion right. Five linear facial measurements were found to have moderate to good agreement between the manual and automated identification methods. The developed algorithm was determined to be clinically relevant in the detection of midsagittal landmarks and associated measurements within the studied sample of adolescent Caucasian subjects.

A robust method for automatic identification of femoral landmarks, axes, planes and bone coordinate systems using surface models

The identification of femoral landmarks is a common procedure in multiple academic fields. Femoral bone coordinate systems are used particularly in orthopedics and biomechanics, and are defined by landmarks, axes and planes. A fully automatic detection overcomes the drawbacks of a labor-intensive manual identification. In this paper, a new automatic atlas- and a priori knowledge-based approach that processes femoral surface models, called the A&A method, was evaluated. The A&A method is divided in two stages. Firstly, a single atlas-based registration maps landmarks and areas from a template surface to the subject. In the second stage, landmarks, axes and planes that are used to construct several femoral bone coordinate systems are refined using a priori knowledge. Three common femoral coordinate systems are defined by the landmarks detected. The A&A method proved to be very robust against a variation of the spatial alignment of the surface models. The results of the A&A method and a manual identification were compared. No significant rotational differences existed for the bone coordinate system recommended by the International Society of Biomechanics. Minor significant differences of maximally 0.5° were observed for the two other coordinate systems. This might be clinically irrelevant, depending on the context of use and should, therefore, be evaluated by the potential user regarding the specific application. The entire source code of the A&A method and the data used in the study is open source and can be accessed via https://github.com/RWTHmediTEC/FemoralCoordinateSystem.

A robust method for automatic identification of landmarks on surface models of the pelvis

The recognition of bony landmarks of the pelvis is a required operation in patient-specific orthopedics, subject-specific biomechanics or morphometrics. A fully automatic detection is preferable to a subjective and time-consuming manual identification. In this paper, a new approach, called the iterative tangential plane (ITP) method, for fully automatic identification of landmarks on surface models of the pelvis is introduced. The method includes the landmarks to construct the two most established anatomical reference frames of the pelvis: the anterior pelvic plane (APP) coordinate system and superior inferior spine plane (SISP) coordinate system. The ITP method proved to be robust against the initial alignment of the pelvis in space. A comparison to a manual identification was performed that showed minor but significant (p < 0.05) median differences below 3 mm for the position of the landmarks and below 1° for the orientation of the APP coordinate system. Whether these differences are acceptable, has to be evaluated for each specific use case. There were no significant differences for the orientation of the SISP coordinate system recommended by the International Society of Biomechanics.

Automated identification of cephalometric landmarks: Part 1-Comparisons between the latest deep-learning methods YOLOV3 and SSD.

To compare the accuracy and computational efficiency of two of the latest deep-learning algorithms for automatic identification of cephalometric landmarks. A total of 1028 cephalometric radiographic images were selected as learning data that trained You-Only-Look-Once version 3 (YOLOv3) and Single Shot Multibox Detector (SSD) methods. The number of target labeling was 80 landmarks. After the deep-learning process, the algorithms were tested using a new test data set composed of 283 images. Accuracy was determined by measuring the point-to-point error and success detection rate and was visualized by drawing scattergrams. The computational time of both algorithms was also recorded. The YOLOv3 algorithm outperformed SSD in accuracy for 38 of 80 landmarks. The other 42 of 80 landmarks did not show a statistically significant difference between YOLOv3 and SSD. Error plots of YOLOv3 showed not only a smaller error range but also a more isotropic tendency. The mean computational time spent per image was 0.05 seconds and 2.89 seconds for YOLOv3 and SSD, respectively. YOLOv3 showed approximately 5% higher accuracy compared with the top benchmarks in the literature. Between the two latest deep-learning methods applied, YOLOv3 seemed to be more promising as a fully automated cephalometric landmark identification system for use in clinical practice.

The effect of automated landmark identification on morphometric analyses.

Morphometric analysis of anatomical landmarks allows researchers to identify specific morphological differences between natural populations or experimental groups, but manually identifying landmarks is time-consuming. We compare manually and automatically generated adult mouse skull landmarks and subsequent morphometric analyses to elucidate how switching from manual to automated landmarking will impact morphometric analysis results for large mouse (Mus musculus) samples (n=1205) that represent a wide range of 'normal' phenotypic variation (62 genotypes). Other studies have suggested that the use of automated landmarking methods is feasible, but this study is the first to compare the utility of current automated approaches to manual landmarking for a large dataset that allows the quantification of intra- and inter-strain variation. With this unique sample, we investigated how switching to a non-linear image registration-based automated landmarking method impacts estimated differences in genotype mean shape and shape variance-covariance structure. In addition, we tested whether an initial registration of specimen images to genotype-specific averages improves automatic landmark identification accuracy. Our results indicated that automated landmark placement was significantly different than manual landmark placement but that estimated skull shape covariation was correlated across methods. The addition of a preliminary genotype-specific registration step as part of a two-level procedure did not substantially improve on the accuracy of one-level automatic landmark placement. The landmarks with the lowest automatic landmark accuracy are found in locations with poor image registration alignment. The most serious outliers within morphometric analysis of automated landmarks displayed instances of stochastic image registration error that are likely representative of errors common when applying image registration methods to micro-computed tomography datasets that were initially collected with manual landmarking in mind. Additional efforts during specimen preparation and image acquisition can help reduce the number of registration errors and improve registration results. A reduction in skull shape variance estimates were noted for automated landmarking methods compared with manual landmarking. This partially reflects an underestimation of more extreme genotype shapes and loss of biological signal, but largely represents the fact that automated methods do not suffer from intra-observer landmarking error. For appropriate samples and research questions, our image registration-based automated landmarking method can eliminate the time required for manual landmarking and have a similar power to identify shape differences between inbred mouse genotypes.