基于节点输入策略贝叶斯网络的骨盆骨折分型研究

基于历史数据的统计和收集,选取骨盆骨折患者存在的18个体表特征,采用基于K2算法的贝叶斯网络方法挖掘各体表特征之间和骨盆骨折类型与体表特征间的相互关系;设计不同的节点输入策略,分析不同输入策略对算法性能的影响;基于骨盆稳定性将骨盆骨折分成A、B、C三种类型,分别找到与其直接相关的体表特征,作为判断骨盆骨折类型的依据.基于体表特征和骨盆骨折类型的分析结果,借助早期的观察及简单检查,对患者进行初步分型.

Pelvic Fracture Classificaiton Based on the Bayesian Network of Node Ordering Strategy

Early diagnosis and pre-hospital care are the key to the treatment of pelvic fractures. Traditional diagnostic methods rely on instruments and electronic devices, which can provide clear images while also increasing time and possibly missing the prime time of rescue. Based on the statistics and collection of historical data, 18 surface features of patients with pelvic fractures were selected. Bayesian network structure learning algorithm based on K2 algorithm was used to mine the causal relationship between the 18 surface features, also between the surface features and the pelvic fracture types. Different node ordering strategies are designed to analyze the influence on algorithm performance. Based on the stability of the pelvis, pelvic fracture is divided into A, B and C 3 types according to the Tile type. Then found the symptoms directly associated with A, B and C three types of pelvic fracture, which was the basis of judgment. Based on the analysis of surface features and pelvic fracture types, preliminary diagnosis and classification were designed by means of early observation and simple examination.