Analysis Of Brain Development In Late Preterm And Term Infants Based On Deep Learning

Objective:To compare the differences in brain volume and brain age between full-term and late-term infants when corrected gestational age to full-term.And to analyze the risk factors affecting neonatal brain development so that we can take comprehensive measures to prevent its occurrence and reduce the risk of neurodevelopment.In order to,improve the prognosis and improve the quality of life of newborns.Methods:We enrolled late preterm infants(34 weeks ≤gestational age<37 weeks)and term infants(37 weeks ≤ gestational age<42 weeks)who were admitted to the second hospital of Jilin University from January 2019 to December 2019.All the infants enrolled had performed cranial ultrasound within 2 days of birth and the enrolled late preterm infants were all established with cranial MRI when corrected gestational age to term.We collected 21 terms of clinical data including,gender,singleton or multiple births,birth weight,weight at the time of MRI,maternal condition of pregnancy(mainly gestational diabetes,gestational hypertensive disorders),mode of delivery(antepartum or cesarean),presence of premature rupture of membranes,presence of intrauterine distress,contamination with amniotic fluid and the amount of amniotic fluid,placenta or not,umbilical cord or not,land 5-minute Apgar scores,affected neonatal conditions(mainly neonatal respiratory distress syndrome,neonatal infection,patent ductus arteriosus,wet lung,neonatal anemia),postnatal cranial ultrasound whether or not they have subependymal hemorrhage.The dl4fetalba model was used to construct a brain age prediction model using deep learning from MRI images of normal term infants and validated using its own dataset.Whole brain tissue volumes were measured using MIPAV software and the collected clinical data were statistically analyzed using spss26.0.Results:1.The whole brain tissue volume in term infants(398.91±40.80ml)was larger than that in late preterm infants(359.66±40.89ml)corrected for gestational age to term and the difference was statistically significant(t=9.639,P<0.001).2.For the univariate analysis of factors affecting whole brain volume,gender,single or multiple birth,birth weight,maternal disease status during pregnancy,1-minute Apgar score and 5-minute Apgar score of the term group could affect the whole brain tissue volume with statistical differences(P<0.05).Whole brain tissue volume could be influenced by gender and birth weight in the late preterm group with statistical differences(P<0.05).3.The results of multiple linear correlation analysis on the factors affecting the volume of whole brain tissue showed that in the full-term infant group,sex,birth weight,maternal gestational diabetes and hypertension during pregnancy were independent factors affecting the volume of whole brain tissue(F=10.390,P<0.001),construct a regression equation,Y=289.932+13.803 × sex+0.029× birth weight-39.961×GDM combined with HDCP.Birth weight and gender in the late premature group were independent factors affecting the total brain tissue volume(F=6.474,P<0.05).A regression equation was constructed,Y=319.367+12.973×gender+0.0014×birth weight.4.Five slices(two higher than central,two lower than central)prediction accuracy is superior with lower MAE(0.83 weeks)and lower computational cost.5.For the univariate analysis of factors influencing the outcome of brain age prediction,only the subgroup of gestational age was related to the outcome of brain age prediction.Conclusions:1.Measurement of whole brain volume in newborns is feasible and reproducible with mipav software.2.Gender and birth weight are independent risk factors influencing differences in neonatal whole brain volume.3.The model has high accuracy for brain age prediction and low computational cost.4.In late preterm infants corrected for gestational age to term with less brain maturation than term infants,early neurodevelopmental monitoring of preterm infants may reduce neurodevelopmental risk.