Deep learning-assisted construction of three-demensional facial midsagittal plane / 北京大学学报(医学版)

OBJECTIVE@#To establish a deep learning algorithm that can accurately determine three-dimensional facial anatomical landmarks, multi-view stacked hourglass convolutional neural networks (MSH-CNN) and to construct three-dimensional facial midsagittal plane automatically based on MSH-CNN and weighted Procrustes analysis algorithm.@*METHODS@#One hundred subjects with no obvious facial deformity were collected in our oral clinic. Three-dimensional facial data were scanned by three-dimensional facial scanner. Experts annotated twenty-one facial landmarks and midsagittal plane of each data. Eighty three-dimensional facial data were used as training set, to train the MSH-CNN in this study. The overview of MSH-CNN network architecture contained multi-view rendering and training the MSH-CNN network. The three-dimensional facial data were rendered from ninety-six views that were fed to MSH-CNN and the output was one heatmap per landmark. The result of the twenty-one landmarks was accurately placed on the three-dimensional facial data after a three-dimensional view ray voting process. The remaining twenty three-dimensional facial data were used as test set. The trained MSH-CNN automatically determined twenty-one three-dimensional facial anatomical landmarks of each case of data, and calculated the distance between each MSH-CNN landmark and the expert landmark, which was defined as position error. The midsagittal plane of the twenty subjects' could be automatically constructed, using the MSH-CNN and Procrustes analysis algorithm. To evaluate the effect of midsagittal plane by automatic method, the angle between the midsagittal plane constructed by the automatic method and the expert annotated plane was calculated, which was defined as angle error.@*RESULTS@#For twenty subjects with no obvious facial deformity, the average angle error of the midsagittal plane constructed by MSH-CNN and weighted Procrustes analysis algorithm was 0.73°±0.50°, in which the average position error of the twenty-one facial landmarks automatically determined by MSH-CNN was (1.13±0.24) mm, the maximum position error of the orbital area was (1.31±0.54) mm, and the minimum position error of the nasal area was (0.79±0.36) mm.@*CONCLUSION@#This research combines deep learning algorithms and Procrustes analysis algorithms to realize the fully automated construction of the three-dimensional midsagittal plane, which initially achieves the construction effect of clinical experts. The obtained results constituted the basis for the independent intellectual property software development.

A method for constructing three-dimensional face symmetry reference plane based on weighted shape analysis algorithm / 北京大学学报(医学版)

OBJECTIVE@#To establish a novel method based on three-dimensional (3D) shape analysis and weighted Procrustes analysis (WPA) algorithm to construct a 3D facial symmetry reference plane (SRP), automatically assigning weight to facial anatomical landmarks. The WPA algorithm suitability for commonly observed clinical cases of mandibular deviation were analysed and evaluated.@*METHODS@#Thirty patients with mandibular deviation were recruited for this study. The 3D facial SRPs were extracted independently based on original-mirror alignment method. Thirty-two anatomical landmarks were selected from the overall region by three times to obtain the mean coordinate. The SRP of experimental groups 1 and 2 were using the standard Procrustes analysis (PA) algorithm and WPA algorithm, respectively. A reference plane defined by experts based on regional iterative closest point (ICP) algorithm, served as the ground truth. Three experts manually selecting facial regions with good symmetry for original model, and common region was included in the study. The angle error values between the SRP of WPA algorithm in the experimental group 1 and the truth plane were evaluated in this study, and the SRP of PA algorithm of experimental group 2 was calculated in the same way. Statistics and measurement analysis were used to comprehensively evaluate the clinical suitability of the WPA algorithm to calculate the SRP. A paired t-test analysis (two-tailed) was conducted to compare the angles.@*RESULTS@#The average angle error between the SRP of WPA algorithm and the ground truth was 1.53°±0.84°, which was smaller than that between the SRP of PA and the ground truth (2.06°±0.86°). There were significant differences in the angle errors among the groups (P < 0.05). For the patients with severe mandibular deviation that the distance between pogonion and facial midline greater than 12 mm, the average angle error of the WPA algorithm was 0.86° smaller than that of the PA algorithm.@*CONCLUSION@#The WPA algorithm, based on weighted shape analysis, can provide a more adaptable SRP than the standard PA algorithm when applied to mandibular deviation patients and preliminarily simulate the diagnosis strategies of clinical experts.

Constant fluctuating asymmetry but not directional asymmetry along the geographic distribution of Drosophila antonietae (Diptera, Drosophilidae)

ABSTRACT The population dynamics of a species tends to change from the core to the periphery of its distribution. Therefore, one could expect peripheral populations to be subject to a higher level of stress than more central populations (the center–periphery hypothesis) and consequently should present a higher level of fluctuating asymmetry. To test these predictions we study asymmetry in wing shape of five populations of Drosophila antonietae collected throughout the distribution of the species using fluctuating asymmetry as a proxy for developmental instability. More specifically, we addressed the following questions: (1) what types of asymmetry occur in populations of D. antonietae? (2) Does the level of fluctuating asymmetry vary among populations? (3) Does peripheral populations have a higher fluctuating asymmetry level than central populations? We used 12 anatomical landmarks to quantify patterns of asymmetry in wing shape in five populations of D. antonietae within the framework of geometric morphometrics. Net asymmetry – a composite measure of directional asymmetry + fluctuating asymmetry – varied significantly among populations. However, once net asymmetry of each population is decomposed into directional asymmetry and fluctuating asymmetry, most of the variation in asymmetry was explained by directional asymmetry alone, suggesting that populations of D. antonietae have the same magnitude of fluctuating asymmetry throughout the geographical distribution of the species. We hypothesize that larval development in rotting cladodes might play an important role in explaining our results. In addition, our study underscores the importance of understanding the interplay between the biology of a species and its geographical patterns of asymmetry.

Variación de patrones morfológicos craneofaciales en relaciones esqueléticas Clase I, II y III / Variation of craniofacial morphological patterns in Class I, II, and III skeletal relationships

Introducción: tradicionalmente, los estudios de variaciones morfológicas de componentes craneofaciales para clasificar relaciones esqueléticas han considerado análisis univariados y multivariados mediante variables como distancias, ángulos y planos referenciales. Sin embargo, estos métodos no explican cambios generales de forma y proporcionan una descripción parcial y localizada de estas relaciones. En tanto, los métodos basados en Morfometría Geométrica (MG) en dos o tres dimensiones (2D o 3D,) permiten una comprensión detallada y un examen más sensible de variables. El objetivo fue identificar la variación de patrones morfológicos de la estructura CraneofacialGeneral(CFG) en relaciones esqueléticas I, II y III, utilizando MG-2D. Métodos: se hizo un estudio prospectivo mediante muestreo no probabilístico. Se tomaron 272 radiografías laterales de cráneo (140 hombres/132 mujeres) de individuos colombianos entre 17 y 25 años, y se determinó el error intraexaminador considerando la prueba F-ANOVA como estadístico de prueba. Se hizo Análisis Generalizado de Procrustes (AGP) y detección de datos atípicos por Cuantil Adaptativo. La variación en tamaño se analizó con prueba Kruskal-Wallis, considerando la matriz de Tamaño Centroide (CS) y las diferencias conformacionales con MANOVA. La identificación de patrones craneofaciales se determinó mediante Análisis de Componentes Principales (ACP) y Conglomerados/K-medias. Resultados: en la estructura CFG se encontraron diferencias conformacionales y una capacidad de buena clasificación del 89% (Clase I), 89% (Clase II) y 91% (Clase III). Se identificaron cuatro patrones craneofaciales; tres evidenciaron relaciones esqueléticas clásicas, y el otro identificó un nuevo grupo combinado de Clase I/II. Conclusiones: las diferencias morfológicas en los cuatro patrones identificados fueron evidentes, la MG permitió una visualización explicativa de patrones de variación morfológica, localizando sitios reales en donde ocurren cambios en tamaño y conformación.

Introduction: the studies on morphological variations of craniofacial components to classify skeletal relationships have traditionally included univariate and multivariate analysis using variables such as distances, angles, and reference planes. However, these methods fail to explain general changes in shape and provide partial localized descriptions of these relationships. Whereas methods using two- or threedimensional (2D or 3D) Geometric Morphometrics (GM) allow a detailed understanding and a more sensitive test of variables. The objective of this study was to identify morphological pattern variations of the Overall Craniofacial Structure (OCS) in skeletal relationships I, II, and III using GM-2D. Methods: this was a prospective study using non-probability sampling. It implied taking 272 lateral radiographs of the head of Colombian individuals (140 males/132 females) aged 17 to 25 years, determining intra-examiner error and using F-ANOVA as statistic test. Generalized Procrustes Analysis (GPA) was conducted as well as atypical data detection by Adaptive Quantile. Size variation was analyzed by the Kruskal-Wallis test considering Centroid Size matrix (CS) and conformational differences were analyzed with MANOVA. Craniofacial patterns were identified by Principal Components Analysis (PCA) and K-means/cluster. Results: the OCS showed conformational differences and a good classification capacity of 89% (Class I), 89% (Class II), and 91% (Class III). Four craniofacial patterns were identified; three of them showed typical skeletal relationships and the other pointed out to a new Class I/II combined group. Conclusions: the morphological differences in the four identified patterns were evident; GM allowed an explanatory display of morphological variation patterns, identifying actual sites where changes in size and shape take place.

Comparison of two kinds of methods evaluating the degree of facial asymmetry by three-dimensional data / 北京大学学报(医学版)

Objective:To compare two digital methods of quantitatively accessing the degree of facial asymmetry by three-dimensional data.Methods: The three-dimensional data of 20 subjects were got by the FaceScan, and then were input to the reverse engineering software Imageware 13.0 and Geomagic 12 .Their mirror data were acquired and superimposed with the original data by the methods of interactive closest points ( ICP) and Procrustes analysis ( PA) .The mid-sagittal planes of the two methods were ex-tracted respectively, the degree of facial asymmetry and the distance of 21 automatic landmarks to mid-sagittal plane were calculated and compared.Results:The paired t test was taken and t=1.346, P=0.193.Conclusion:We can safely come to the conclusions that for the subjects with no evident facial asymmetry, there are no significant difference between the PA and the ICP methods for extracting the mid-sagittal plane from three-dimensional data.

Morfometría geométrica y el estudio de las formas biológicas: de la morfología descriptiva a la morfología cuantitativa / Geometric morphometry and the biologic shapes study: from the descriptive morphology to the quantitative morphology

La morfometría es el estudio de la covariación de la forma con factores subyacentes. Su desarrollo en las últimas décadas ha alcanzado áreas de la biología tradicionalmente dedicadas al estudio descriptivo, como las ciencias morfológicas, las que con las nuevas herramientas morfométricas geométricas han logrado no sólo objetivar la evaluación cuantitativa de los cambios morfológicos sino también la evaluación cualitativa a través de la recuperación de la forma en estudio. Esto es posible gracias a la aplicación de técnicas biométricas, instrumentos y programas computacionales que permiten la captura y análisis de datos en forma de matrices de morfocoordenadas que representan la geometría de un espécimen y no se limitan a la obtención de datos lineales de él como medidas de alto o ancho, elementos que carecen de la precisión y la riqueza de los datos geométricos. El análisis morfométrico geométrico consta de tres etapas fundamentales: obtención de los datos primarios, obtención de las variables de la forma, y análisis estadístico. El extenso uso que se le ha dado en los últimos años en áreas afines a las ciencias morfológicas hace necesario el conocimiento de la técnica tanto con fines formativos como para su aplicación a la solución de problemas en los que la morfología juega un rol esencial.

Morphometrics is the study of co-variation of biological form and its causes. Its development over the last decades has reached several biological sciences with a traditional descriptive approach, such as morphological sciences. The new geometric morphometric tools allow not only objective quantitative analysis, but also to assess qualitative traits due to the chance of recovering the form under study. This is possible because of the application of biometry techniques, instruments and software that allow the acquisition and analysis of shape coordinates that represent the geometry of the specimen, and that are not limited to obtaining linear data that lack of precision and amount of information of geometric data. Geometric morphometric analysis consists of three fundamental steps: obtaining primary data, obtaining shape variables and statistical analysis. The extensive use of this technique in areas related to morphological sciences over the last years makes geometric morphometrics a must-know subject for morphologists, in general knowledge, as well as for its use in solving problems where morphology plays an essential role.

Avaliação sensorial de pudins de chocolate com açúcar e dietéticos por perfil livre / Sensory evaluation of sweetened and dietetic chocolate puddings by free-choice profile

Perfil Livre, uma técnica sensorial descritiva, foi utilizada na caracterização de três amostras de pudins com açúcar e cinco de pudins dietéticos comerciais. Quatorze provadores realizaram o levantamento de atributos pelo método Rede. Foram elaborados para cada provador, as listas de definições de atributos e as fichas de avaliação, empregando escala não estruturada. Utilizou-se a Análise Procrustes Generalizada para tratamento dos dados. Foram ainda determinados o perfil de textura instrumental e a cor. Foi obtida boa discriminação e os pudins foram caracterizados com base, principalmente, nos atributos cor marrom, sinérese, aroma e sabor de chocolate, sabor residual, firmeza e cremosidade. A técnica de Perfil livre mostrou-se eficiente para discriminação sensorial das amostras estudadas considerando-se atributos de aparência, sabor, odor e textura.

Free-choice profile, a descriptive sensory technique, was applied to develop the profile for three regular and five dietetic commercial chocolate puddings. Fourteen panelists were selected using triangular tests, and the Grid method was used to obtain the descriptors. Besides, individual lists, definitions of the attributes, and score sheets, where each attribute was scored on an unstructured scale, were made for the assessors. The Generalized Procrustes Analysis was applied to data. Instrumental texture profile and color were also determined. Good discrimination was observed between the samples. Puddings were mainly characterized by brown color, sineresis, chocolate aroma and flavor, aftertaste, firmness and creaminess. Free-choice profiling was efficient to discriminate studied samples considering appearance, aroma, flavor and texture attributes.