Unconstrained Infant Development Assessment Based On 3D Eye Movement Characteristics

Vision is one of the main channels for humans to obtain external information,and plays an important role in integrating other perceptions.As the language and movement abilities of infants are not yet fully mature,vision can be one of the important ways to understand the brain and cognitive development of infants before they can express themselves clearly and completely.Studies have shown that abnormal eye movement characteristics are related to the brain’s information processing,so they can reflect a potential relationship with infant developmental disorders,which have been used in traditional clinical evaluations and diagnosis.Although the visual development plasticity mechanism and eye movement control mechanism have always been the focus of infant ophthalmology research,the development of this field still has the following problems: Firstly,eye movement diagnosis methods for infants and children often rely heavily on the experience of doctors.However,at present,clinical professional assessors are not only lacking,but also distributed unevenly.Secondly,due to the poor compliance of infants,there is still a large lack of knowledge about related eye movement research.Therefore,the World urgently needs to promote the popularization of intelligent eye movement assessment methods and equipment.Finally,the current eye movement assessment tools are expensive and hinder their promotion in primary hospitals.At the same time,wearing equipment will restrict head movements and make infants uncomfortable,which will affect the authenticity of users’ feedback data.In order to solve the above problems,in this paper,a non-constrained evaluation method for infant development based on 3D eye movement features is studied.At the same time,in order to ensure the authenticity and accuracy of the experimental results,a clinical experiment was conducted in conjunction with Nanjing Maternal and Child Health Hospital.First,based on the physical and psychological characteristics of infants and young children,this thesis relies on stereo imaging technology to design and build a non-contact eye movement data acquisition platform based on binocular stereo vision.Second,the double fitting method based on the farthest point finite asymptotic and least squares algorithm realizes the direct extraction of the pupil information parameter,which is the key target.Finally,different eye movement task paradigms are designed under different disease diagnosis requirements to stimulate infant vision,and the data captured from different experiments are analyzed individually.In the "Induced horizontal motion mapping" experiment,by comparing and analyzing the mean and variance in the regional time,the pupil coordinate dispersion,and the binocular pupil synergy extracted from the pupil information parameters of different types of experimental subjects,it is found that there is a significant difference in eye movement data between normal and abnormal children,and it also successfully verifies the correlation between eye movement characteristics and direction.A " Facial expression vertical mapping " experiment is then conducted on the basis of this experiment,where K-nearest neighbors algorithm is used to perform intelligent three-region classification on the data,and results indicate that children with Autism Spectrum Disorder have abnormalities in face recognition and processing of emotional expressions.Through the analysis of clinical trial data,this thesis demonstrates the clinical applicability of non-contact eye movement data acquisition platforms and the correlation between eye movement characteristics and developmental disorders,thereby verifying the feasibility of non-constrained infant development evaluation.And It also lays the foundation for the establishment of the relationship between the pathological database and corresponding features in the later stages,and the accurate mapping of 3D eye movement features and developmental disability categories.