Cephalometric Landmark Detection by AttentiveFeature Pyramid Fusion and Regression-Voting

Daniel Lee; Nenglun Chen; Runnan Chen; Wenping Wang; Yuexin Ma

REVIEW 4 major objections 5 minor 26 references

Cephalometric Landmark Detection by AttentiveFeature Pyramid Fusion and Regression-Voting

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Cephalometric landmark AI reaches 1.17 mm error, beats prior

desk verdict A real benchmark improvement on ISBI 2015, but the 7-11% claim is too broad and the unquantified label-noise floor weakens the absolute accuracy story. read the letter →

arxiv 1908.08841 v1 pith:R64BCBGW submitted 2019-08-23 cs.CV

Runnan Chen , Yuexin Ma , Nenglun Chen , Daniel Lee , Wenping Wang This is my paper · ORCID

classification cs.CV

keywords cephalometriclandmarksdeeplearningself-attentionfeaturepyramidfusionregression-votinglandmarkdetectionmedicalimagingheatmapsandoffset

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Accurate cephalometric analysis depends on locating 19 craniofacial landmarks on X-ray images, a task that is laborious by hand and hard to automate because different landmarks need different resolutions and semantic context. The paper proposes an end-to-end deep learning network built from two ideas: an attentive feature pyramid fusion (AFPF) module that recombines multi-scale features and learns a per-landmark attention weighting, and a prediction module that combines heat maps with offset maps to vote for each landmark's position. The authors claim this design improves every evaluation metric by 7%–11% over the previous best method on the standard public cephalometric benchmark. Their reported mean radial error — the average distance between predicted and doctor-marked positions — is 1.17 mm on the validation set and 1.48 mm on a held-out test set, with 2 mm success rates of 86.67% and 75.05%. If accurate, this would make the method the most accurate published automatic detector on this dataset and fast enough for real-time clinical use.

What carries the argument

The central mechanism is the Attentive Feature Pyramid Fusion module (AFPF). It takes feature maps from several depths of a backbone network, rescales them to a common size with lateral connections and upsampling, concatenates them, and passes them through a dilated convolutional block to form a feature pyramid. For each of the 19 landmarks a self-attention vector is computed as $\mathbf{a}_k = \mathrm{softmax}(\mathbf{W}_{k1}\tanh(\mathbf{W}_{k2}\tilde{\mathbf{F}}))$ and applied as a channel-wise multiplication to select the channels relevant to that landmark; a $1\times1$ convolution then emits one heat map and two offset maps per landmark. In the prediction stage, pixels with the largest heat-map values cast votes for the position their offset maps point to, and the pixel receiving the most votes is the predicted landmark. The training loss is a weighted sum of a logistic loss on heat maps and an L1 loss on offset maps, with the heat-map term weighted by $\alpha=2/3$. This combination — multi-scale fusion, per-landmark attention, and vote aggregation — is the machinery the paper claims produces the accuracy gain.

What would settle it

Re-measure performance against a third clinician's independent annotations on the same test images and compare each prediction with the single-observer label; if the mean distance between doctors is already around 1 mm or more, the reported 1.17 mm accuracy would plausibly be limited by label noise rather than by the network's design.

Watch

Extended reading notes

Core claim

The paper's central claim is that no single network layer supplies both the spatial resolution and the semantic content that all 19 landmarks need, and that an explicit fusion of multi-scale features, weighted per landmark by attention, removes this limitation. The paper asserts that AFPF fuses feature maps from multiple backbone layers into a common pyramid, applies a dilated convolution for multi-scale context, and then uses a self-attention weight vector per landmark so each landmark reads the fusion feature it needs. A regression-voting step turns these features into final positions by letting every pixel whose heat-map response is large vote for the location indicated by its offset map. On the public benchmark this yields a mean radial error of 1.17 mm and 1.48 mm on the two test partitions and 2 mm successful detection rates of 86.67% and 75.05%, surpassing the compared methods by 7% to 11% on all metrics. The paper's ablations show that removing AFPF or its attention mechanism degrades accuracy, and that inserting AFPF into other backbone networks improves them, which it reads as evidence that the fusion-and-attention design is what carries the gain.

Load-bearing premise

The load-bearing assumption is that the average of two doctors' landmark annotations used as ground truth is accurate enough to measure a 1.17 mm mean error; the paper does not quantify how much the two doctors disagree.

Editorial extensions

If this is right

Clinicians could get real-time automated landmark marking: the pipeline processes one image in about 70 ms on a GPU, and most landmarks land within the 2 mm clinical tolerance.
Because AFPF also improves ResNet50 and Inception backbones in the paper's experiments, the module is a reusable component that could raise accuracy in other landmark-localization networks.
The reported results on images from a device never seen in training (mean error 0.88 mm) suggest the method may work across clinics without per-device retraining.
The heat-map/offset-map voting scheme outperforms earlier methods even without the fusion module, so the voting mechanism alone is a useful design for landmark detection.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The per-landmark attention weights could be visualized to identify which anatomical context each landmark depends on; that would test whether the network behaves like a clinician or exploits image artifacts.
Since the ground truth averages two doctors' annotations, the reported 1.17 mm error may be close to the human-agreement floor; comparing predictions against each doctor separately would reveal how much of the residual is annotation noise.
The same fusion-plus-attention recipe could be transplanted to other medical landmark tasks with heterogeneous target sizes, such as fetal ultrasound or retinal imaging, where a 7–11% gain would be clinically meaningful.
One could deliberately distort or remove high-frequency texture in local patches to see whether the attention weights shift; this would test whether the network's resolution preference is causal or incidental.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Quick take: the headline numbers are probably a genuine advance on the ISBI 2015 benchmark, but the "7%~11% for all evaluation metrics" claim is not true as stated, and the evaluation never quantifies the label noise floor. Good engineering, overstated packaging.

The genuinely new thing is the AFPF module: fusing multi-scale features with per-landmark channel attention, then doing heatmap-plus-offset regression-voting. The ablations in Table 1 do their job — removing AFPF costs about 0.18 mm MRE on Test Dataset 1, and removing attention costs another 0.05 mm. The module also transfers to ResNet50 and Inception, which suggests it is a real component, not a one-off trick. Speed is fine at 70 ms on a GTX 1080 Ti, and the method is tested on a public benchmark used by the field. That is a reasonable contribution.

The weak spots are clear. First, the abstract says "7%~11% for all evaluation metrics," but Table 1 shows APC gains of only 2.64 and 0.96 percentage points over the best prior rows. The 7-11% range holds for the SDR metrics, not for APC. That is an overclaim and should be corrected. Second, the benchmark ground truth is the average of two doctors' annotations, and the paper never reports how far apart those two sets are. If the inter-observer distance is around 1 mm, the reported 1.17 mm MRE sits near the label-noise floor. The extended experiments use a single dentist's labels and claim this avoids inter-observer errors — actually it removes disagreement but adds a new annotation bias, so without reliability data those numbers do not fully support the generalization claim. Third, no code or trained models, and no repeated-run variance, so the SOTA claim is not independently checkable. The comparison to Payer et al. is also only in text, not in the main table.

None of this destroys the central result. The ranking on the official benchmark is probably robust because baselines are all measured on the same reference average. But the absolute accuracy claim and the "7-11%" phrasing need a rewrite, and the label-noise question needs at least a paragraph.

This paper is for people working on cephalometric analysis and, more broadly, on evaluation pitfalls in landmark-detection benchmarks. A good referee would push on the noise floor and the APC overclaim, but the paper deserves a serious peer review, not a desk reject. My recommendation: send it out, with an explicit request that the authors correct the scope of the headline claim and report inter-observer variability if it is available.

Referee Report

4 major / 5 minor

Summary. The paper proposes an end-to-end deep learning framework for automatic cephalometric landmark detection from lateral cephalometric radiographs. The architecture consists of a VGG-19 backbone, a novel attentive feature pyramid fusion (AFPF) module that fuses multi-resolution feature maps and applies per-landmark self-attention, and a prediction module that combines heat maps and offset maps with pixel-wise regression-voting. The method is evaluated on the ISBI 2015 Challenge dataset, reporting a mean radial error (MRE) of 1.17 mm on Test Dataset 1 and 1.48 mm on Test Dataset 2, with 2 mm SDR of 86.67% and 75.05%, respectively, and claims a 7%–11% improvement over the state of the art across all metrics. Additional ablations demonstrate the contribution of the AFPF module and its flexibility with different backbones, and extended experiments on private datasets from multiple devices are used to support generalization claims.

Significance. If the reported results hold, the proposed framework would be the most accurate published landmark detector on the ISBI 2015 benchmark, and the AFPF module appears to be a versatile component that improves accuracy across multiple backbone networks. The paper's strengths include a clear ablation study showing the incremental contributions of the AFPF module and the self-attention mechanism, a demonstration of architectural flexibility (ResNet50 and Inception), and a direct evaluation on unseen data sources from different devices, which is practically valuable. The reported inference time (70 ms on a GPU) supports clinical feasibility. These strengths are partially offset by the absence of code, repeated-run statistics, and a quantitative characterization of the label-noise floor, which are needed to fully substantiate the state-of-the-art claim.

major comments (4)

[Section 3.1, Table 1] The evaluation uses the average of annotations from two doctors as ground truth, but the paper never quantifies the inter-observer variability between these two annotations. In cephalometric landmark detection, typical inter-observer MRE is of the order of 1–2 mm, so the reported MRE values of 1.17 mm and 1.48 mm may be close to the label-noise floor. Without reporting the distance between the two doctors' annotations (mean, standard deviation, and per-landmark breakdown), it is impossible to assess whether the observed differences from baselines reflect algorithmic superiority or noise in the reference standard. This is load-bearing because the central SOTA claim depends on the reliability of the ground truth.
[Abstract, Section 1, Table 1] The claim that the method improves accuracy by '7%~11% for all the evaluation metrics over the state-of-the-art method' is not supported by Table 1. For APC, the improvement over the best prior row is only 2.64 percentage points (79.05 vs 76.41) on Test Dataset 1 and 0.96 percentage points (81.95 vs 80.99) on Test Dataset 2, which is outside the claimed 7–11% range. For MRE, the relative improvement over Lindner et al. is about 30% (1.67 mm to 1.17 mm). The abstract and introduction should be revised to state the per-metric improvements precisely and avoid the blanket 'all evaluation metrics' claim.
[Section 3.3, Table 1] No repeated-run statistics or significance tests are reported. Several margins are small, notably the 2 mm SDR on Test Dataset 2 (75.05 vs 74.21 for the no-attention ablation) and APC on Test Dataset 2 (81.95 vs 80.61 for the no-attention ablation). With only 100 test images, these differences may not be statistically distinguishable. The authors should provide standard deviations over multiple training runs or perform paired significance tests to support the claim that the full method is superior to its ablations and to prior methods.
[Section 3.2, Table 1] Payer et al. [22] is listed as a baseline, but its results are not included in Table 1; instead, only combined two-test-set accuracies are discussed in the text. Since Payer et al. is a recent CNN-based landmark detection method that is directly comparable, the omission makes the head-to-head comparison incomplete. The authors should either include Payer et al.'s per-test-set numbers in Table 1 (if available) or clearly justify why a direct comparison is not possible.

minor comments (5)

[Section 3.4] The text reports 'MSE is 1.03mm' and 'MSE is 0.88mm'; since these values are in millimeters and appear to denote mean radial error, 'MSE' is a likely typo for 'MRE'.
[Figure 1] The module-area colors and the text labels in the figure are difficult to discern in a grayscale print; consider adding explicit module boundaries or labels.
[Equation (1)] The dimensions of W_k1 and W_k2 are not specified; stating the input/output sizes would make the attention formulation clearer.
[Section 2.2] The choice of alpha = 2/3 is described as empirical, but no sensitivity analysis is reported; a brief note on how the results vary with alpha would increase confidence in the robustness of the loss balance.
[References] Several references are incorrectly formatted, e.g., [22] lists authors as 'P. Christian, Š. Darko, B. Horst, and U. Martin' instead of Christian Payer, Darko Štern, Horst Bischof, and Martin Urschler; similarly [23] appears garbled. These should be corrected.

Circularity Check

0 steps flagged · score 2.0 of 10

Empirical benchmark comparison; only a minor non-load-bearing self-citation, no circular derivation.

full rationale

The paper's central claim is an empirical benchmark comparison on the ISBI 2015 dataset, not a derivation from assumptions that covertly contain the conclusion. The pipeline is a standard deep-learning architecture: VGG-19 features, an attentive feature pyramid fusion module, and heat-map/offset-map regression-voting. The prediction equations (Eq. 1-4) are learned attention weights and Hough-style voting; they do not define the reported accuracy numbers in terms of themselves. The method is evaluated against fixed external benchmarks (Table 1) and extended datasets, so the claimed MRE/SDR improvements are testable empirical outputs, not renaming or fitted parameters passed off as predictions. The only self-citation is reference [26] (TrafficPredict), by Yuexin Ma, Wenping Wang, and others, cited together with Vaswani et al. [19] for the self-attention mechanism. That citation is not load-bearing: the attention mechanism is independently established by [19], and the paper's benchmark results do not depend on any result from [26]. The abstract's '7%~11% for all evaluation metrics' is not fully reflected in the APC columns of Table 1 (gains of 2.64 and 0.96 percentage points), and the ground truth, an average of two doctors' annotations, has unquantified inter-observer variability; however, these are internal-consistency and benchmark-validity concerns, not circular reasoning. No circular step is present.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The central claim is an empirical accuracy claim, not a derivation. Its free parameters are the manually chosen loss weight alpha, the disk radius R, and the fixed input resolution. The main domain assumptions are that the challenge's averaged ground truth is reliable, that ImageNet-pretrained features transfer, and that a uniform disk is a sufficient training target. No invented physical entities are introduced.

free parameters (3)

alpha (loss balance weight) = 2/3
Set 'empirically' in Equation 3 to balance heatmap and offset losses; it is a tuned constant, not derived from first principles.
R (heatmap and voting radius) = 40 pixels
Chosen in Section 2.2 so that there is a minimal activation area on the smallest feature map with stride 32; it controls the binary heatmap ground truth and the size of the voting set.
Input image size = 800 x 640
Fixed resize used for all images in training and testing; no justification is given for this resolution choice.

assumptions (3)

domain assumption The average of the two doctors' manual annotations is treated as the true landmark position (Section 3.1).
All accuracy numbers are computed against this average; if annotation noise is large, the benchmark itself is noisy.
domain assumption ImageNet-pretrained VGG-19 features transfer to grayscale cephalometric radiographs (Section 3).
The backbone is initialized with ImageNet weights and fine-tuned; transferability is assumed rather than demonstrated.
ad hoc to paper A fixed disk of radius R=40 adequately encodes the target region for every landmark and every image (Section 2.2).
The binary heatmap target is a uniform disk; no landmark-specific or image-specific uncertainty is modeled.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Cephalometric Landmark Detection by AttentiveFeature Pyramid Fusion and Regression-Voting." pith.science (2026). https://pith.science/paper/R64BCBGW

@misc{pith2026190808841,
  author       = {Pith},
  title        = {Pith review of: Cephalometric Landmark Detection by AttentiveFeature Pyramid Fusion and Regression-Voting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/R64BCBGW}},
  note         = {Machine review of arXiv:1908.08841}
}

read the original abstract

Marking anatomical landmarks in cephalometric radiography is a critical operation in cephalometric analysis. Automatically and accurately locating these landmarks is a challenging issue because different landmarks require different levels of resolution and semantics. Based on this observation, we propose a novel attentive feature pyramid fusion module (AFPF) to explicitly shape high-resolution and semantically enhanced fusion features to achieve significantly higher accuracy than existing deep learning-based methods. We also combine heat maps and offset maps to perform pixel-wise regression-voting to improve detection accuracy. By incorporating the AFPF and regression-voting, we develop an end-to-end deep learning framework that improves detection accuracy by 7%~11% for all the evaluation metrics over the state-of-the-art method. We present ablation studies to give more insights into different components of our method and demonstrate its generalization capability and stability for unseen data from diverse devices.

Figures

Figures reproduced from arXiv: 1908.08841 by the authors.

**Figure 1.** Overview of our framework. Three consecutive modules, the feature extraction module, the AFPF module, and the prediction module are in blue, yellow and purple areas respectively. Feature maps are reshaped in the green box of AFPF. Attention vectors of heat maps and offset maps for landmarks are in the orange box of AFPF. 2 Methods Given a cephalometric radiography I, the goal is to detect anatomical landmark positio… view at source ↗

**Figure 2.** Data samples of five datasets with 393, 154, 100, 709, 501 samples respectively. Blue points are the ground truth of landmarks. 3.4 Extended Experiments In the real world, images captured by different devices differs significantly. A good algorithm should have the self-adaptive ability even for the data from a new device. To test generalization capability and stability of our approach, we use five datasets ( [PITH_… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 23 canonical work pages

[22]

Christian, ˇS

P. Christian, ˇS. Darko, B. Horst, and U. Martin. Integrating spatial conﬁguration into heatmap regression based cnns for landmark localization. MIA, 2019

work page 2019
[1]

Ricketts, RH

R.M. Ricketts, RH. Roth, SJ. Chaconasand RJ. Schulhof, and GA. Engel. Or- thodontic diagnosis and planning. Denver: RMDS, 1:267p, 1982

work page 1982
[2]

Knowledge-based land- marking of cephalograms

AD Levy-Mandel, AN Venetsanopoulos, and JK Tsotsos. Knowledge-based land- marking of cephalograms. CBR, 19(3):282–309, 1986

work page 1986
[3]

El-Feghi, M.A

I. El-Feghi, M.A. Sid-Ahmed, and M. Ahmadi. Automatic localization of cranio- facial landmarks for assisted cephalometry. Pattern Recognition, 37(3):609–621, 2004

work page 2004
[4]

Cootes, C.J

T.F. Cootes, C.J. Taylor, D.H. Cooper, and J. Graham. Active shape models-their training and application. CVIU, 61(1):38–59, 1995

work page 1995
[5]

Cootes, G.J

T.F. Cootes, G.J. Edwards, and C.J. Taylor. Active appearance models. TPAMI, (6):681–685, 2001

work page 2001
[6]

Ibragimov, B

B. Ibragimov, B. Likar, F. Pernus, and T. Vrtovec. Automatic cephalometric x-ray landmark detection by applying game theory and random forests. In Proc. ISBI Int. Symp. on Biomedical Imaging, 2014

work page 2014
[7]

Lindner and T.F

C. Lindner and T.F. Cootes. Fully automatic cephalometric evaluation using ran- dom forest regression-voting. In ISBI. Citeseer, 2015

work page 2015

Show all 26 references

[8]

C. Wang, C. Huang, M. Hsieh, C. Li, et al. Evaluation and comparison of anatomi- cal landmark detection methods for cephalometric x-ray images: A grand challenge. TMI, 34(9):1890–1900, 2015

1900
[9]

C. Wang, C. Huang, J. Lee, C. Li, S. Chang, et al. A benchmark for comparison of dental radiography analysis algorithms. MIA, 31:63–76, 2016

2016
[10]

Lindner and C

C. Lindner and C. Wang and C. Huang and C. Li, et al. Fully automatic sys- tem for accurate localisation and analysis of cephalometric landmarks in lateral cephalograms.Scientiﬁc reports, 2016

2016
[11]

S. Arik, B. Ibragimov, and L. Xing. Fully automated quantitative cephalometry us- ing convolutional neural networks. Journal of Medical Imaging, 4(1):014501, 2017

2017
[12]

W. Yue, D. Yin, C. Li, G. Wang, and T. Xu. Automated 2-d cephalometric analysis on x-ray images by a model-based approach. TBE, 53(8):1615–1623, 2006

2006
[13]

LeCun, Y

Y. LeCun, Y. Bengio, and G. Hinton. Deep Learning. nature, 521(7553):436, 2015

2015
[14]

Litjens, T

G. Litjens, T. Kooi, B.E. Bejnordi, et al. A survey on deep learning in medical image analysis. MIA, 42:60–88, 2017

2017
[15]

Simonyan and A

K. Simonyan and A. Zisserman. Very deep convolutional networks for large-scale image recognition. arXiv:1409.1556, 2014

2014 arXiv
[16]

K. He, X. Zhang, S. Ren, and J. Sun. Deep residual learning for image recognition. In CVPR, pages 770–778, 2016. Cephalometric Landmark Detection by AFPF and Regression-Voting 9

2016
[17]

Szegedy, W

C. Szegedy, W. Liu, Y. Jia, P. Sermanet, S. Reed, D. Anguelov, et al. Going deeper with convolutions. In CVPR, pages 1–9, 2015

2015
[18]

Yu and V

F. Yu and V. Koltun. Multi-scale context aggregation by dilated convolutions. arXiv:1511.07122, 2015

2015 arXiv
[19]

Vaswani, N

A. Vaswani, N. Shazeer, N. Parmar, et al. Attention is all you need. In NIPS, pages 5998–6008, 2017

2017
[20]

Ibragimov, B

B. Ibragimov, B. Likar, F. Pernus, and T. Vrtovec. Computerized cephalometry by game theory with shape-and appearance-based landmark reﬁnement. ISBI, 2015

2015
[21]

Cardillo and M.A

J. Cardillo and M.A. Sid-Ahmed. An image processing system for locating cranio- facial landmarks. TMI, 13(2):275–289, 1994

1994
[23]

Martin, E

U. Martin, E. Thomas, and ˇS. Darko. Integrating geometric conﬁguration and appearance information into a uniﬁed framework for anatomical landmark local- ization. MIA, 43:23–36, 2018

2018
[24]

Papandreou, T

G. Papandreou, T. Zhu, N. Kanazawa and others. Towards accurate multi-person pose estimation in the wild. In CVPR, volume 3, page 6, 2017

2017
[25]

Krizhevsky, I

A. Krizhevsky, I. Sutskever and G.E. Hinton. Imagenet classiﬁcation with deep convolutional neural networks. In NIPS, page 1097-1105, 2012

2012
[26]

Y. Ma, X. Zhu, S. Zhang, R. Yang, W. Wang, D. Manocha. ”TraﬃcPredict: Tra- jectory prediction for heterogeneous traﬃc-agents.” arXiv:1811.02146, 2018

2018 arXiv

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [22]

Christian, ˇS

P. Christian, ˇS. Darko, B. Horst, and U. Martin. Integrating spatial conﬁguration into heatmap regression based cnns for landmark localization. MIA, 2019

work page 2019

[2] [1]

Ricketts, RH

R.M. Ricketts, RH. Roth, SJ. Chaconasand RJ. Schulhof, and GA. Engel. Or- thodontic diagnosis and planning. Denver: RMDS, 1:267p, 1982

work page 1982

[3] [2]

Knowledge-based land- marking of cephalograms

AD Levy-Mandel, AN Venetsanopoulos, and JK Tsotsos. Knowledge-based land- marking of cephalograms. CBR, 19(3):282–309, 1986

work page 1986

[4] [3]

El-Feghi, M.A

I. El-Feghi, M.A. Sid-Ahmed, and M. Ahmadi. Automatic localization of cranio- facial landmarks for assisted cephalometry. Pattern Recognition, 37(3):609–621, 2004

work page 2004

[5] [4]

Cootes, C.J

T.F. Cootes, C.J. Taylor, D.H. Cooper, and J. Graham. Active shape models-their training and application. CVIU, 61(1):38–59, 1995

work page 1995

[6] [5]

Cootes, G.J

T.F. Cootes, G.J. Edwards, and C.J. Taylor. Active appearance models. TPAMI, (6):681–685, 2001

work page 2001

[7] [6]

Ibragimov, B

B. Ibragimov, B. Likar, F. Pernus, and T. Vrtovec. Automatic cephalometric x-ray landmark detection by applying game theory and random forests. In Proc. ISBI Int. Symp. on Biomedical Imaging, 2014

work page 2014

[8] [7]

Lindner and T.F

C. Lindner and T.F. Cootes. Fully automatic cephalometric evaluation using ran- dom forest regression-voting. In ISBI. Citeseer, 2015

work page 2015