The invention proposes a
pregnancy risk assessment early warning method and
system based on multi-
source data fusion, and the method comprises the steps: collecting multi-
modal data, carrying out the preprocessing of the multi-
modal data, and obtaining first data, and the preprocessing comprises the
time sequence alignment, window reconstruction and
standardization; extracting physiological features and behavior features from the first data, wherein the physiological features are segmented
convolution features of uterine electrical activity on a multi-frequency structure; the behavior feature is a
fetus image frame behavior
complexity index, is coded by an ultrasonic
image sequence and is calculated based on inter-frame similarity; constructing a maternal-fetal interaction
state diagram based on the physiological features and the behavior features, and constructing a diagram embedding vector of the maternal-fetal interaction
state diagram; the
graph embedding vector is obtained by performing propagation type embedding calculation on all node features on a graph, and represents an interactive propagation state of a physiological-behavior
signal in a current maternal-fetal
system; and performing
risk level evaluation on the current
pregnancy state based on the maternal-fetal interaction
state diagram and the diagram embedding vector, and outputting a risk cause chain.