Face Recognition Task Research Articles

Free-viewing gaze patterns reveal a mood-congruency bias in MDD during an affective fMRI/eye-tracking task

Major depressive disorder (MDD) has been related to abnormal amygdala activity during emotional face processing. However, a recent large-scale study (n = 28,638) found no such correlation, which is probably due to the low precision of fMRI measurements. To address this issue, we used simultaneous fMRI and eye-tracking measurements during a commonly employed emotional face recognition task. Eye-tracking provide high-precision data, which can be used to enrich and potentially stabilize fMRI readouts. With the behavioral response, we additionally divided the active task period into a task-related and a free-viewing phase to explore the gaze patterns of MDD patients and healthy controls (HC) and compare their respective neural correlates. Our analysis showed that a mood-congruency attentional bias could be detected in MDD compared to healthy controls during the free-viewing phase but without parallel amygdala disruption. Moreover, the neural correlates of gaze patterns reflected more prefrontal fMRI activity in the free-viewing than the task-related phase. Taken together, spontaneous emotional processing in free viewing might lead to a more pronounced mood-congruency bias in MDD, which indicates that combined fMRI with eye-tracking measurement could be beneficial for our understanding of the underlying psychopathology of MDD in different emotional processing phases.Trial Registration: The BeCOME study is registered on ClinicalTrials (gov: NCT03984084) by the Max Planck Institute of Psychiatry in Munich, Germany.

Visual Domain Adaptation through Locality Information

Traditional machine learning methods have always performed the learning tasks solely. The models of these conventional methods have to be built from scratch. The idea of domain adaptation can overcome this learning procedure by adapting the knowledge gained from one domain to another. With datasets from several domains, this research proposes a novel strategy for unsupervised visual domain adaptation. Existing transfer learning approaches try to reduce the domain shift using maximum mean discrepancy metric. We propose a combined framework termed Visual Domain Adaptation through Locality Information (DALI) that lowers the geometrical distance between the domains while probabilistically addressing the shortcomings. The proposed approach employs the input data features to develop a technique that assists in inducing the transfer of information from one domain to another. Extensive experimental findings demonstrate that the proposed method DALI excels over traditional domain adaptation algorithms as well as CNN architectures on a variety of cross-domain object, facial, and digit recognition tasks including Office-Caltech, Office-Home, PIE and COIL datasets. Our proposed model has improved the mean accuracy to 69.69%, exceeding the most recent state-of-the-art techniques when performed a comprehensive analysis on the Office-Home data.

LCAM: Low-Complexity Attention Module for Lightweight Face Recognition Networks

Inspired by the human visual system to concentrate on the important region of a scene, attention modules recalibrate the weights of either the channel features alone or along with spatial features to prioritize informative regions while suppressing unimportant information. However, the floating-point operations (FLOPs) and parameter counts are considerably high when one is incorporating these modules, especially for those with both channel and spatial attentions in a baseline model. Despite the success of attention modules in general ImageNet classification tasks, emphasis should be given to incorporating these modules in face recognition tasks. Hence, a novel attention mechanism with three parallel branches known as the Low-Complexity Attention Module (LCAM) is proposed. Note that there is only one convolution operation for each branch. Therefore, the LCAM is lightweight, yet it is still able to achieve a better performance. Experiments from face verification tasks indicate that LCAM achieves similar or even better results compared with those of previous modules that incorporate both channel and spatial attentions. Moreover, compared to the baseline model with no attention modules, LCAM achieves performance values of 0.84% on ConvFaceNeXt, 1.15% on MobileFaceNet, and 0.86% on ProxylessFaceNAS with respect to the average accuracy of seven image-based face recognition datasets.

LIAAD: Lightweight attentive angular distillation for large-scale age-invariant face recognition

Disentangled representations have been commonly adopted to Age-invariant Face Recognition (AiFR) tasks. However, these methods have reached some limitations with (1) the requirement of large-scale face recognition (FR) training data with age labels, which is limited in practice; (2) heavy deep network architectures for high performance; and (3) their evaluations are usually taken place on age-related face databases while neglecting the standard large-scale FR databases to guarantee robustness. This work presents a novel Lightweight Attentive Angular Distillation (LIAAD) approach to Large-scale Lightweight AiFR that overcomes these limitations. Given two high-performance heavy networks as teachers with different specialized knowledge, LIAAD introduces a learning paradigm to efficiently distill the age-invariant attentive and angular knowledge from those teachers to a lightweight student network making it more powerful with higher FR accuracy and robust against age factor. Consequently, LIAAD approach is able to take the advantages of both FR datasets with and without age labels to train an AiFR model. Far apart from prior distillation methods mainly focusing on accuracy and compression ratios in closed-set problems, our LIAAD aims to solve the open-set problem, i.e. large-scale face recognition. Evaluations on LFW, IJB-B and IJB-C Janus, AgeDB and MegaFace-FGNet with one million distractors have demonstrated the efficiency of the proposed approach on light-weight structure. This work also presents a new longitudinal face aging (LogiFace) database 1This database will be made available1 for further studies in age-related facial problems in future.

Non-linear effects of stress on eyewitness memory

The effect of stress on face identification is not yet clear, with recent experiments finding positive, negative and null results. Here we report the results of two experiments examining the effect of stress on eyewitness performance in line-up face recognition tasks. Both experiments use a stress manipulation and live mock crime paradigm to examine the relationship between stress at encoding and subsequent line-up performance. Experiment 1 replicated an experiment by Sauerland et al. (Behav Sci Law. 2016;34(4):580–594) which induced stress using the Maastricht Acute Stress Test. The replication found the same null result as the original experiment. Experiment 2 aimed to address a limitation of many laboratory experiments which dichotomise stress into low and high groups for comparison. As the Yerkes-Dodson law (1908) suggests that a non-linear relationship exists between stress and performance, it was hypothesised that using a low, medium and high stress manipulation might show clearer results than a dichotomous paradigm. The results of Experiment 2 show a non-linear relationship, with no difference between the low and high stress groups but better performance by the middle stress group. The results suggest that a different approach is required in experiments on stress and face recognition, as the stress–performance relationship is likely non-linear. Significance: Non-linear models are better predictors of face recognition in line-up tasks than are linear models. Two group designs provide insufficient resolution to capture the stress–performance relationship.

Kernel Reverse Neighborhood Discriminant Analysis

Currently, neighborhood linear discriminant analysis (nLDA) exploits reverse nearest neighbors (RNN) to avoid the assumption of linear discriminant analysis (LDA) that all samples from the same class should be independently and identically distributed (i.i.d.). nLDA performs well when a dataset contains multimodal classes. However, in complex pattern recognition tasks, such as visual classification, the complex appearance variations caused by deformation, illumination and visual angle often generate non-linearity. Furthermore, it is not easy to separate the multimodal classes in lower-dimensional feature space. One solution to these problems is to map the feature to a higher-dimensional feature space for discriminant learning. Hence, in this paper, we employ kernel functions to map the original data to a higher-dimensional feature space, where the nonlinear multimodal classes can be better classified. We give the details of the deduction of the proposed kernel reverse neighborhood discriminant analysis (KRNDA) with the kernel tricks. The proposed KRNDA outperforms the original nLDA on most datasets of the UCI benchmark database. In high-dimensional visual recognition tasks of handwritten digit recognition, object categorization and face recognition, our KRNDA achieves the best recognition results compared to several sophisticated LDA-based discriminators.

Robust Semisupervised Deep Generative Model Under Compound Noise.

Semisupervised learning has been widely applied to deep generative model such as variational autoencoder. However, there are still limited work in noise-robust semisupervised deep generative model where the noise exists in both of the data and the labels simultaneously, which are referred to as outliers and noisy labels or compound noise. In this article, we propose a novel noise-robust semisupervised deep generative model by jointly tackling the noisy labels and outliers in a unified robust semisupervised variational autoencoder randomized generative adversarial network (URSVAE-GAN). Typically, we consider the uncertainty of the information of the input data in order to enhance the robustness of the variational encoder toward the noisy data in our unified robust semisupervised variational autoencoder (URSVAE). Subsequently, in order to alleviate the detrimental effects of noisy labels, a denoising layer is integrated naturally into the semisupervised variational autoencoder so that the variational inference is conditioned on the corrected labels. Moreover, to enhance the robustness of the variational inference in the presence of outliers, the robust β-divergence measure is employed to derive the novel variational lower bound, which already achieves competitive performance. This further motivates the development of URSVAE-GAN that collapses the decoder of URSVAE and the generator of a robust semisupervised generative adversarial network into one unit. By applying the end-to-end denoising scheme in the joint optimization, the experimental results demonstrate the superiority of the proposed framework by the evaluating on image classification and face recognition tasks and comparing with the state-of-the-art approaches.

Pupil dilation during encoding, but not type of auditory stimulation, predicts recognition success in face memory

We encounter and process information from multiple sensory modalities in our daily lives, and research suggests that learning can be more efficient when contexts are multisensory. In this study, we were interested in whether face identity recognition memory might be improved in multisensory learning conditions, and to explore associated changes in pupil dilation during encoding and recognition. In two studies participants completed old/new face recognition tasks wherein visual face stimuli were presented in the context of sounds. Faces were learnt alongside no sound, low arousal sounds (Experiment 1), high arousal non-face relevant, or high arousal face relevant (Experiment 2) sounds. We predicted that the presence of sounds during encoding would improve later recognition accuracy, however, the results did not support this with no effect of sound condition on memory. Pupil dilation, however, was found to predict later successful recognition both at encoding and during recognition. While these results do not provide support to the notion that face learning is improved under multisensory conditions relative to unisensory conditions, they do suggest that pupillometry may be a useful tool to further explore face identity learning and recognition.

Examining the association between speed-accuracy trade-offs and psychopathy among male offenders in a facial affect recognition task.

Despite a large body of work examining the relationship between facial affect recognition and psychopathy, there is little consensus regarding the nature of emotion processing in such individuals. Although most previous studies have reported an association between psychopathy and some facial affect recognition deficits, results are mixed regarding which specific emotions are related to deficient recognition for these individuals. The current study aims to examine speed-accuracy trade-offs in psychopathy. Participants were 139 incarcerated adult male offenders assessed using the Psychopathy Checklist-Revised. Analyses demonstrated a 3-way interaction between psychopathy, response time, and level of stimulus expression for happiness in which psychopathy mitigated the relationship between response time and level of expression on accuracy. However, the overall pattern of findings does not suggest that speed-accuracy trade-offs are a central component of facial affect recognition in psychopathy. The theoretical implications of current findings are discussed. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Priority of racial and gender categorization of faces: A social task demand framework.

Social categorization of faces occurs along multiple dimensions such as race and gender and influences interpersonal interactions widely. Does social categorization of faces along different dimensions take place in parallel or give priority to some faces along a specific dimension? Across seven studies, we proposed and empirically tested a social task demand hypothesis that social categorization of a particular group of faces occurs in priority along one dimension (race or gender) in response to specific social task demands. Study 1 showed behavioral evidence that other-race faces are more densely clustered into one category, whereas same-race faces are more distantly separated into two (male and female) categories. Studies 2-4 showed electrophysiological evidence for earlier spontaneous racial categorization of other-race faces but earlier spontaneous gender categorization of same-race faces during an individual face recognition task. Studies 5-7 manipulated social task demands and provided electrophysiological evidence that early processes underlying racial or gender categorization of other-race or same-race faces were inhibited or facilitated by task demands of evaluating gender roles or group membership of perceived faces. Our findings indicate that the priority of racial or gender categorization of other-race or same-race faces varies dynamically to meet social task demands of managing intergroup and cross-sex interactions. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Human Face Emotions Recognition from Thermal Images Using DenseNet

In the current scenario face identification and recognition is an important technique in surveillance. The face is a necessary biometric in humans. Therefore face detection plays a major job in computer vision applications. Several face recognition and emotions classification approaches have been presented throughout the last few decades of research to improve the rate of face recognition for thermal pictures. However, in real-time, lighting conditions might change due to several factors, such as the different times of capture, weather, etc. Due to variations in lighting intensity, the performance of the facial expression recognition system is not good. This paper proposed a model for human thermal face detection and expression classification. Four main steps were involved in this research. Initially, the Difference of the Gaussian (DOG) filter is utilized to crop the input thermal images and then normalize the images using the median filter in pre-processing step. Then, Efficient Net is used for extracting features such as shape, location, and occurrences from thermal face images. After that, detect human faces utilized by the YOLOv4 technique to better emotions classification. Finally, classify the emotions on faces by using the DenseNet technique into seven emotions such as happy, sad, disgust, surprise, anger, fear, and neutral. The proposed method outperforms state-of-art techniques for face recognition on thermal pictures, and classifies the expressions, according to experimentations on the RGB-D-T database. The accuracy, precision, recall, and f1-score metrics will be utilized with the database to assess the efficacy of the proposed methodology. The proposed models achieve a high classification accuracy of 95.97% on the RGB-D-T database. Furthermore, the outcomes show good precision for various face recognition tasks.

Acute exercise impairs social-cognitive abilities in children with Autism Spectrum Disorder

Background
 While children with Autism Spectrum Disorder show heterogeneous cognitive profiles, they share common deficits in social-cognitive abilities and executive function. Moderate endurance exercise is known to elicit changes in the psychophysiological state, which can translate into immediate, but transient benefits for both cognitive domains. However, the cognitive response of children with Autism Spectrum Disorder to acute exercise remains unclear. We addressed this research deficit by investigating the effects of a single endurance exercise session on face recognition abilities and affective response inhibition.
 Methods
 We recruited 30 children with Autism Spectrum Disorder aged 7 to 12 years from local clinicals in Basel, Switzerland and via social media. Using a randomized cross-over design, participants completed a moderately-intense, 20-min exercise session on a cycling ergometer and control condition, which involved sitting on the ergometer. We collected heart rate and ratings of perceived exertion during these sessions. Before and after the experimental conditions, participants completed two computerized cognitive tasks. During the Mooney Face Recognition task, they had to categorize stimuli into faces and objects. In the affective Go/NoGo task, participants were instructed to inhibit their response to either sad or happy faces and press a button to all other faces.
 Results
 The experimental manipulation was successful as a main effect of condition indicated a higher heart rate and rating of perceived exertion in the exercise compared to the control condition. With regard to cognitive performance, the repeated measures ANOVA revealed an interaction of time and condition for the Mooney Face Recognition task, indicating an increase in reaction time in the exercise compared to the control condition. In contrast, no interaction of time and condition was found for accuracy on this task as well as accuracy and reaction time on the affective Go/NoGo task.
 Conclusion
 In children with Autism Spectrum Disorder, a single exercise does not seem to benefit social-cognitive abilities and executive function temporarily. A short cycling bout rather impairs the ability to recognize faces, suggesting that the maintenance of supervised exercise demands a set of joint resources. As these results are in conflict with the well-documented benefit of acute exercise in healthy populations, the atypical cognitive response in children with Autism Spectrum Disorder indicates a moderating role of this neurodevelopmental condition.

Recognizing distant faces

As an ‘early alerting’ sense, one of the primary tasks for the human visual system is to recognize distant objects. In the specific context of facial identification, this ecologically important task has received surprisingly little attention. Most studies have investigated facial recognition at short, fixed distances. Under these conditions, the photometric and configural information related to the eyes, nose and mouth are typically found to be primary determinants of facial identity. Here we characterize face recognition performance as a function of viewing distance and investigate whether the primacy of the internal features continues to hold across increasing viewing distances. We find that exploring the distance dimension reveals a qualitatively different salience distribution across a face. Observers’ recognition performance significantly exceeds that obtained with the internal facial physiognomy, and also exceeds the computed union of performances with internal and external features alone, suggesting that in addition to the mutual configuration of the eyes, nose and mouth, it is the relationships between these features and external head contours that are crucial for recognition. We have also conducted computational studies with convolutional neural networks trained on the task of face recognition to examine whether this representational bias could emerge spontaneously through exposure to faces. The results provide partial support for this possibility while also highlighting important differences between the human and artificial system. These findings have implications for the nature of facial representations useful for a visual system, whether human or machine, for recognition over large and varying distances.

Deficits in Face Recognition and Consequent Quality-of-Life Factors in Individuals with Cerebral Visual Impairment.

Individuals with cerebral visual impairment (CVI) frequently report challenges with face recognition, and subsequent difficulties with social interactions. However, there is limited empirical evidence supporting poor face recognition in individuals with CVI and the potential impact on social-emotional quality-of-life factors. Moreover, it is unclear whether any difficulties with face recognition represent a broader ventral stream dysfunction. In this web-based study, data from a face recognition task, a glass pattern detection task, and the Strengths and Difficulties Questionnaire (SDQ) were analyzed from 16 participants with CVI and 25 controls. In addition, participants completed a subset of questions from the CVI Inventory to provide a self-report of potential areas of visual perception that participants found challenging. The results demonstrate a significant impairment in the performance of a face recognition task in participants with CVI compared to controls, which was not observed for the glass pattern task. Specifically, we observed a significant increase in threshold, reduction in the proportion correct, and an increase in response time for the faces, but not for the glass pattern task. Participants with CVI also reported a significant increase in sub-scores of the SDQ for emotional problems and internalizing scores after adjusting for the potential confounding effects of age. Finally, individuals with CVI also reported a greater number of difficulties on items from the CVI Inventory, specifically the five questions and those related to face and object recognition. Together, these results indicate that individuals with CVI may demonstrate significant difficulties with face recognition, which may be linked to quality-of-life factors. This evidence suggests that targeted evaluations of face recognition are warranted in all individuals with CVI, regardless of their age.

Study of input images parameters for improving person identification information technology

This article describes the research of the algorithm based on local-textural descriptors, which is the basis of the information technology of person identification, with the aim of increasing its efficiency and reducing the probability of identification errors. An analysis of previous studies has shown that the performance of a face recognition and identification algorithm can vary significantly after applying it to images from different datasets. In addition, the issue of using local texture descriptors in the tasks of face recognition and identification on images of different quality has not been sufficiently studied in modern research. Therefore, the main goal of this work is to study the parameters of face images that will be used for identification process with the appliance of the algorithm based on such local-textural methods as histograms of oriented gradients and local binary patterns in one-dimensional space. During the experiments, it was established that the algorithm identification accuracy rate can be decreased by 35-70% when it is applied to image samples taken in unconstrained conditions. In order to investigate the possibility of reducing the variation of the results, experiments were conducted on the same images with the transformation of their properties, such as format and resolution. During the research, experiments were carried out with several of the most common databases of face images, as a result of which the efficiency of the algorithm reached the level of accuracy of identification of 95% on images taken under controlled conditions and unified within one database. As a result, it was found that in some cases, the conversion of the image format submitted to the input of the algorithm can increase the accuracy of the algorithm’s identification by 5%. Also, the efficiency of the algorithm is affected by the resolution of the input images, in particular, due to the application of the algorithm to transformed images from the Database of Faces database, the accuracy of identification increased by 5-10%, FERET — decreased by 5-35%, SCface — decreased by 10-30%, CFP — increased by 5-10%, Tinyface — increased by 25-35%, LFW — increased by 15%, AgeDB — increased by 5%. The highest identification accuracy rate of the algorithm, which is 95%, was obtained after its application to images taken under constrained conditions and unified within the same database.

The Effect of Aging on Face-Name Recognition: An fMRI Study.

The aim of this study is to detect functional changes in the brain during the memory task with aging and the association between functional changes and memory performance. The study consisted of Young Adult Group (YAG, n=20) aged 20 to 25 and Late Adult Group (LAG, n=18) aged 60 to 70. Individuals with Montreal Cognitive Assessment (MoCA) scores above 21 and no family history of Alzheimer's Disease were included in the study. Functional Magnetic Resonance Imaging (fMRI) scanning was performed on all participants during a memory task including encoding (face and name), face and name recognition sub-tasks. Results indicated that LAG showed increased activity during face recognition task in left posterior cingulate cortex, left superior frontal cortex, left fusiform face area and another increased activity was found out during name recognition task in left superior frontal cortex, right prefrontal cortex, left anterior + posterior cingulate cortex. The accuracy of face recognition and name recognition memory tests were significantly lower in LAG (respectively, p=0.026; p=0.001). These results indicated that advanced age were associated with more widespread activation in brain during memory task. Thus with aging, individuals require more neuronal and cognitive resources during memory processing.

ОСОБЕННОСТИ ВОСПРИЯТИЯ ПРИВЛЕКАТЕЛЬНОСТИ ЛИЦА ПРИ РАЗНОМ УРОВНЕ ЭМОЦИОНАЛЬНОГО ИНТЕЛЛЕКТА

Background. Many studies have been devoted to the relationship between the perception of facial expression and emotional intelligence (EI). As a rule, the study of this issue involved the tasks of expression recognition with stimuli expressing strong basic emotions. However, diX erences in the success of expression recognition in groups with diX erent levels of EI are not always signiY cantly pronounced. Z is may be connected to the fact, that the recognition of basic expressions is too easy for subjects with both high and low EI scores. In order to conduct a more effective study of this issue, it is necessary to complicate the task of facial expressions recognition. To this end, one of the conY gurational eX ects of E. Brunswick stating that neutral faces with diX erent conY gurations of facial features are perceived as emotionally colored, can be used. Under these conditions, a more adequate task is not to assess the success of expression recognition, but to assess the attractiveness of the face, which is closely related to the perceived facial expression. Z us, faces with positive expressions are more o[ en rated as attractive, and vice versa. It is assumed that the study of the perception of neutral faces will reveal the features of the relationship between the perception of facial expression and EI. Objective. Z e study is meant to explore the relationship between the perception of facial attractiveness and EI. It was assumed that subjects with diX erent levels of EI would evaluate the attractiveness of neutral faces diX erently. Sample. Z e experiment involved 49 people (20 men and 29 women, in the age range of 18–35 years). Methods. Images of faces were presented on a monitor screen. Eye movements were recorded during stimuli presentation. Z e level of EI was assessed based on the results of the “Test of emotional intelligence” (TEI) (Sergienko et al., 2019). We calculated indicators of TEI for all branches of EI (the abilities to recognize, understand, manage, and use emotions), as well as the total score of EI. An analysis of face attractiveness scores as well as eye movement parametres (duration and number of Y xations, duration of saccades) was carried out for two groups of subjects with high and low EI scores.

Face ShapeNets for 3D Face Recognition

In this paper, we present a deep learning-based method for 3D face recognition. Unlike some previous works, our process does not rely on face representation methods as a proxy step to be accepted by Convolutional Neural Networks (CNNs). Applying 2D CNNs to irregular domains such as 3D meshes is challenging. Therefore, we employed 3D ShapeNets to recognize faces covering the full 3D shape since 3D face datasets are available and 3D data augmentation techniques to enlarge 3D datasets are widespread. The reduced size of 3D datasets is overcome by an appropriate 3D data augmentation to train our model. 3D ShapeNets are commonly used to recognize and analyse objects. To the best of the authors’ knowledge, this is the first time they are used for face recognition. This research work focuses on the preprocessing step. Whatever the nature of the face image is (either 2D or 3D representation), and whatever the acquisition conditions are, a 3D regular mesh of each input face image is first generated. Furthermore, meshes are converted to voxels to get the occupancy grid across all possible views. Finally, 3D ShapeNets are trained and recognition tests are performed. Indeed, 3D ShapeNets prove the efficiency and efficacy of 3D shape analysis task in 3D face recognition. The experimental results show that 3D face recognition using deep 3D CNNs such as 3D ShapeNets leads to significant improvement over the state-of-the-art performance on LFPW, BU3DFE, and FRAV3D datasets, with competitive recognition rates of 94.25%, 97.9% and 98.31% respectively.

Privacy-Preserving Split Learning for Large-Scaled Vision Pre-Training

The growing concerns about data privacy in society lead to restrictions on the computer vision research gradually. Several collaboration-based vision learning methods have recently emerged, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">e.g</i> ., federated learning and split learning. These methods protect user data from leaving local devices, and make training performed only by uploading gradients, parameters, or activations, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">etc</i> . However, there is little research on collaborative learning based on state-of-the-art and large-scaled models, mainly due to the high computation or communication overheads of the latest models. Training these models may be still unrealized for users’ terminals. In this paper, we make a first attempt at the sensitive image pre-training with large-scaled models in the collaborative learning scenario, and propose a new lightweight framework for split learning based on mask, Masked Split Learning (MaskSL). We further ensure its security by differential privacy. Besides, we model the computation and communication overheads of several collaborative learning approaches by deduction to illustrate advantages of our scheme. Finally, we design and conduct a series of experiments on real-world datasets, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">e.g</i> ., in face recognition and medical image classification tasks, to demonstrate the performance of MaskSL.

Effects of early exposure to phthalates on cognitive development and visual behavior at 24 months

Background and objectivesStudies focusing on the neurodevelopmental effects of phthalates seldom consider exposure during infancy, a critical period for brain development. Most rely on parent-completed questionnaires to assess child neurodevelopment, which may be subject to reporting error. We studied the associations between prenatal and infancy exposure to phthalates and objective measures of neurodevelopment at the age of two. MethodsWe relied on 151 mother-child pairs from the SEPAGES mother-child cohort. Women were asked to collect three spot urine samples per day over seven consecutive days during the second (median: 18.0 gestational weeks) and third (median: 34.2 gestational weeks) trimesters of pregnancy. They then collected one urine sample per day over seven consecutive days from their infants around the age of 12 months. Metabolites of phthalates and non-phthalate plasticizers were measured in within-subject and within-period pools of repeated urine samples. Eye tracking tasks were performed at two years allowing to compute four indicators linked with cognitive development and visual behavior: mean fixation duration, novelty preference, percent time spent looking at the eyes and mean reaction time. ResultsPre-natal exposure to monobenzyl phthalate at the second and third trimesters was associated with shorter fixation durations. In models allowing for interaction with child sex, these associations were only observed among girls. Exposure to di(2-ethylhexyl) phthalate at the third but not the second trimester was associated with increased time spent looking at a novel face and eyes. We observed faster reaction times and decreased time spent looking at the eyes in a face recognition task, with increased post-natal exposure to monoethyl, mono-iso-butyl and mono-n-butyl phthalates. DiscussionRelying on improved exposure assessment, we highlighted associations of pre- and post-natal exposure to phthalates with indicators derived from eye tracking tasks, mainly in girls. Some of these indicators have been affected in individuals with neurodevelopmental disorders.