Analysis Of Spatiotemporal Differences In Antenatal Cocaine Exposure In Adolescents Based On Resting-state Functional Magnetic Resonance Imaging

Prenatal exposure to cocaine(PCE)refers to the exposure of pregnant women to cocaine or drugs during pregnancy,which affects the development of embryos indirectly.Many studies have shown that prenatal cocaine exposure can not only improve the probability of premature delivery and abortion,but also affect children’s long-term neurocognitive development.Moreover,it is more likely to lead to social problems.Therefore,researches on exposure of pregnant are needed.When in the state of emotional activation,cognitive function will be regulated to maintain balance of internal and external brain.Adolescents prenatally exposed to cocaine have higher demand for cognitive function with high emotional arousal.According to this,we hypothesized that the frequency of transition between emotional and cognitive activation was higher in adolescents prenatally exposed to cocaine than the controls.Based on the above hypothesis,we applied traditional topological analysis of brain networks,unsupervised clustering algorithm and Hidden Markov Model(HMM)to the restingstate functional magnetic resonance imaging data of prenatal cocaine exposed adolescents,for the sake of spatial and temporal differences between the exposed and controls.In the study of topological analysis of resting-state functional brain networks,we found out there were decrease of clustering coefficient,small world and local efficiency in the exposed.These results mean that the functional brain networks of the exposed youths are more disordered than the controls,and may be relating to the defects in cognitive function and arousal regulation in exposed adolescents.We clustered the resting-state functional magnetic resonance imaging data with K-means clustering algorithm.The elbow method,contour coefficient method and Partitioning Around Medoid algorithm were applied to determine the optimal cluster number of K-means.The outcomes showed that in different resting-state functional brain networks,there were different cluster numbers between the exposed and controls.The clustering results are as followed: the emotional brain network owns 5 and 6 categories in the controls and exposed respectively,while the executive control brain network owns 9 and 10 categories in the controls and exposed respectively.Both in emotional brain network and executive control brain network,the number of categories in exposed is more than the controls.In other words,there are more states in the exposed.This experimental finding shows the spatial differences between two groups.In the experiment of Hidden Markov Model with resting-state functional brain networks,the clustering results of K-means was applied to define the observations.Then the stochastic inference was carried out with defined observations and preprocessed data,and the Hidden Markov Models were finally obtained.The transition probability matrix and the fractional occupancy of each state can be calculated with Hidden Markov Model.Our results showed that the transfer frequency of states of the exposed was higher than the controls in both emotional brain network and executive control brain network.This experimental finding shows the temporal differences between two groups,and provides support for our hypothesis.Above all,we discovered the spatial and temporal differences between the exposed and controls in emotional brain network and executive control brain network in this study,.This finding will provide significant information for understanding the inherent pathological mechanism of prenatal cocaine exposed.