Artificial Intelligence and Genetic Predictors as the Basis of Personalized Prediction of Perinatal Pathology

Artificial intelligence integration into the medicine for prognosing perinatal pathology is one of the priority research areas, and offers more accurate, effective and innovative diagnosis and screening. The objective: to develop a perinatal pathology forecasting model using an artificial neural network (ANN) with the study of intergenic associations of the investigated and to analyze the working information technology to prevent the development of pathology. Materials and methods. Correlation analysis was conducted between multifactorial indicators along with genetic parameters of both the mother and the child, which influence the development of perinatal pathology. Calculations were performed using the package MedCalc® Statistical Software (v.22.009). A multi-sphere perceptron was chosen for mathematical modeling of the influence of genetic factors on the development of perinatal pathology. To calculate the weight coefficients of neurons, the method of inverse error propagation was chosen, the predictive accuracy of which is 90%. Results. During the study the structure and algorithms of the ANN were analyzed, the significant influence of genetic portraits of the mother and the child on the development of perinatal pathology compared to other factors was determined. It has been shown that the learning of ANN directly depends on the initial weighting coefficients and thresholds. It is proven that the method can estimate the degree of prognostic significance and the interaction of new factors with the available set. This distinguishes it from traditional statistical methods. The results of the research confirmed that the preparation of input and output data should be carried out by the method of scaling to bring the data into the acceptable range. It is emphasized that the use of ANN in medicine contributes to the automation of analysis, the processing of large clinical data volumes, and the minimization of the risk of errors. Conclusions. The main feature of ANN, which are actively used in genomic research, is its fast learning ability, convergence to the optimal solution when increasing the number of objects, and their efficiency when working with variable data. Prognosing perinatal pathology with the help of ANN with the study of intergenic associations improves the approach to providing medical care to newborns and ensures the prevention of development of long-term outcomes in the future life of the child. The proposed method of prognosis can be used in the development of methodological approaches to the creation of the State program of genetic screening among the population in Ukraine.