The invention discloses a premature delivery
risk assessment method and
system based on multi-
modal physiological signals. The method comprises the following steps: S1, acquiring
uterus electromyographic signals of a pregnant woman, identifying explosive
waves, and obtaining a
uterus electromyographic
characteristic matrix X1; s2, collecting basic information BMI, age AGE, premature delivery history and
pregnancy complications of the pregnant woman to obtain a classification
feature matrix X2; s3, acquiring an iconography identification length sequence VL of the
cervix uteri of the pregnant woman to obtain an image
feature matrix X3; s4, performing model training on the X1, the X2 and the X3 to obtain a model M1, a model M2 and a model M3; s5, training the models M1, M2 and M3 by adopting adaptive weighted fusion to obtain weights W1, W2 and W3 of the models M1, M2 and M3 in the premature delivery prediction task; and S6, sorting the weights calculated in the step S5 from high to low, splicing the weights to obtain fused feature vectors, and retraining the fused feature vectors to obtain a final premature delivery
risk assessment model. According to the method,
early prediction of the premature delivery risk is realized, the
risk probability and the
confidence interval are output, and objective decision support is provided for clinicians.