Search

3D stereophotography is rapidly being adopted by medical researchers for analysis of facial forms and features. An essential step for many applications using 3D face data is to first crop the head and face from the raw images. The goal of this paper is to develop a reliable automatic methodology for extracting the face from raw data with texture acquired from a stereo imaging system, based on the medical researchers' specific requirements. We present an automated process, including eye and nose estimation, face detection, Procrustes analysis and final noise removal to crop out the faces and normalize them. The proposed method shows very reliable results on several datasets, including a normal adult dataset and a very challenging dataset consisting of infants with cleft lip and palate.

Morphometrics, the quantitative analysis of shape, is used by craniofacial researchers to study abnormalities in human face shapes. Most of the work in craniofacial morphometrics uses landmark points that are manually marked on 3D face data and processed via a generalized Procrustes analysis. For large data sets this manual process is very time-consuming. Dense sets of pseudo-landmarks have also been proposed and successfully used for classification and clustering, but the two main methods in the literature are very computationally intensive. We have developed a computationally simple method that can compute pseudo-landmark points at different resolutions from 3D meshes of human faces. In this paper, we perform a comparative study employing L1-regularized logistic regression to train a classifier that predicts the sex of 500 normal adult face meshes in order to compare our method to two alternative pseudo-landmark methods and a distance matrix approach.Our results show that our method, which is fully automatic, achieved similar results to the best-scoring methods with no manual landmarking and with much lower computation time. Use of the distance matrix did not improve classification results.

Craniofacial researchers make heavy use of established facial landmarks in their morphometric analyses. For studies on very large facial image datasets, the standard approach of manual landmarking is very labor intensive. With the goal of producing 20 established landmarks, we have developed a geometric methodology that can automatically locate 10 established landmark points and 7 other supporting points on human 3D facial scans. Then, to improve accuracy and produce all 20 landmarks, a deformable matching procedure establishes a dense correspondence from a template 3D mesh with a full set of 20 landmarks to each individual 3D mesh. The 17 geometrically-determined points on the individual 3D mesh are used for the initial correspondence required by the deformable matching. The method is evaluated on 115 3D facial meshes of normal adults, and results are compared to landmarks manually identified by medical experts. Our results show a marked improvement to prior results in the recent literature.

This paper introduces a new tool to quantify and characterize asymmetry in bilaterally paired structures. This method uses deformable registration to produce a dense vector field describing the point correspondences between two images of bilaterally paired structures. The deformation vector field properties are clustered to detect and describe regions of relevant asymmetry. Three methods are provided to analyze the asymmetries: the global asymmetry score uses cluster features to quantify overall asymmetry, the local asymmetry score quantifies asymmetry in user-defined regions of interest, and the asymmetry similarity measure quantifies pairwise similarity of individual asymmetry. The scores and image distances generated by this tool are shown to correlate highly with asymmetry ratings assigned by an expert.

This paper introduces a new method to quantify and characterize shape changes during early facial development without the use of landmarks. Landmarks are traditionally used in morphometric analysis, but very few can be identified reliably across all stages of embryonic development. This method uses deformable registration to produce a dense vector field describing the point correspondences between two images. Low and mid-level features are extracted from the deformable vector field to find regions of organized differences that are biologically relevant. These methods are shown to detect regions of difference when evaluated on chick embryo images warped with small magnitude deformations in regions critical to midfacial development.

Finding point correspondences plays an important role in automatically building statistical shape models from a training set of 3D surfaces. Davies et al. assumed the projected coefficients have a multivariate Gaussian distributions and derived an objective function for the point correspondence problem that uses minimum description length to balance the training errors and generalization ability. Recently, two-dimensional principal component analysis has been shown to achieve better performance than PCA in face recognition. Motivated by the better performance of 2DPCA, we generalize the MDL-based objective function to 2DPCA in this paper. We propose a gradient descent approach to minimize the objective function. We evaluate the generalization abilities of the proposed and original methods in terms of reconstruction errors. From our experimental results on different sets of 3D shapes of different human body organs, the proposed method performs significantly better than the original method.

How to model shape variations plays an important role in active shape models that is widely used in model-based medical image segmentation, and principal component analysis is a common approach for this task. Recently, different tensor-based dimension reduction methods have been proposed and have achieved better performances than PCA in face recognition. However, how they perform in modeling 3D shape variations of organs in terms of reconstruction errors in medical image analysis is still unclear. In this paper, we propose to use tensor-based dimension reduction methods to model shape variations. We empirically compare two-dimensional principal component analysis, the parallel factor model and the Tucker decomposition with PCA in terms of the reconstruction errors. From our experimental results on several different organs such as livers, spleens and kidneys, 2DPCA performs best among the four compared methods, and the performance differences between 2DPCA and the other methods are statistically significant.

Single-suture craniosynostosis is a condition of the sutures of the infant's skull that causes major craniofacial deformities and is associated with an increased risk of cognitive deficits and learning/language disabilities. In this paper we adapt to classification of synostostic head shapes a Bayesian methodology that overcomes the limitations of our previously published shape representation and classification techniques. We evaluate our approach in a series of large-scale experiments and show performance superior to those of standard approaches such as Fourier descriptors, cranial spectrum, and Euclidian-distance-based analyses.