Intelligent interpretation method for prenatal fetal monitoring based on deep forest

Abstract

The invention discloses an intelligent interpretation method for prenatal fetal monitoring based on a deep forest. The method comprises the following steps: scanning a preprocessed p-dimensional CTG clinical feature vector with known classification through three multi-granularity sliding windows; combining the two forest models to obtain three characterization vectors of 2m (p-d1 + 1) dimension, 2m (p-d2 + 1) dimension and 2m (p-d3 + 1) dimension, and inputting the characterization vectors into four forest models in the cascade forest stage. The cascade forest uses a feature vector after multi-granularity scanning processing as the input of a first layer, obtains a 4m-dimensional category vector through four forest models, and then splices the 4m-dimensional category vector with an original input feature vector to obtain a (4m + d1)-dimensional input feature which is used as the input feature of a next layer. According to the method, the misjudgment problem of suspicious and normal samples in the prenatal fetal monitoring intelligent interpretation model is effectively solved, the workload of medical staff is reduced, and assistance is provided for prenatal examination work of primary hospitals.

Description

technical field [0001] The present invention relates to a deep forest machine learning method, in particular to a deep forest-based intelligent interpretation method for prenatal fetal monitoring, which is used for intelligent classification and judgment of prenatal fetal status assessment. Background technique [0002] In recent years, with the full opening of my country's second-child policy and the development of urbanization, elderly women (≥35 years old) have increased significantly. The National Health and Medical Commission released the "National Medical Service and Quality Safety Report 2018" in October 2019, which shows that: The proportion of elderly women in the 2018 survey was 13.57%, showing a growing trend, and there is a huge demand for prenatal fetal monitoring. [0003] However, the overall level of medical care in my country's rural areas is not high, and there is a serious shortage of fetal monitoring medical personnel in poverty-stricken areas. Most pregna...

AI Technical Summary

Problems solved by technology

[0006] There are defects of high sensitivity and low specificity in the clinical application of the fetal heart contraction monitoring scoring method, and false positives are prone to occur when the prenatal inspection time is less than 40 minutes

However, most of the existing studies on fetal monitoring models based on machine learning are designed based on the balance of sample distribution and the maximization of classification accuracy. The unbalanced distribution is often classified as a normal distribution, and the problem of CTG data imbalance is ignored. The obtained suspicious class accuracy rate is only 45-82%, and the abnormal class accuracy rate is only 66-94%, which cannot be applied in prenatal fetal monitoring

Images