Application of machine learning risk prediction mathematical model in the diagnosis of Escherichia coli infection in patients with septic shock by cardiovascular color doppler ultrasound

Abstract

this study was to explore the diagnosis of septic shock patients with Escherichia coli (E. coli) infection based on cardiovascular color Doppler ultrasound (CCDUS) images under the machine learning risk prediction mathematical model (risk prediction model). 120 septic shock patients with Escherichia coli (E. coli) infection, admitted to xxx hospital were selected as research subjects, and they were randomly divided into experimental group and control group, including 76 males and 44 females, with an average age of (45.47 ± 11.35) years old. The prediction model, random forest mathematical model (RF model), and feature combination were trained and applied in the CCDUS. The error rate, F1-score, and area under the curve (AUC) were compared. It was found that the prediction effect of the risk prediction model was better (P < 0.05). The receiver operating characteristic curve (ROC) was drawn based on the risk prediction model, and it was found that the AUC was 0.924, and the best cutoff value was 0.247. The consistency test between the predicted death result and the actual result showed that Kappa = 0.824, which was higher than 0.75. The pathogenic microorganisms of the patients were mainly Gram-positive bacteria (GPB) in 32 cases (53.33%). There were 19 cases whose pathogenic bacteria was E. coli, and 11 cases (57.9%) of which were acquired in the intensive care unit (ICU). The patient mortality rate was 41.67%. Finally, the acute physiology and chronic health II (APACH II) score and D-dimer of the patients were substituted into the Logistic regression model. The effect of the risk prediction model was better than the RF model and feature combination; the measurement results based on the risk prediction model had good consistency; the D-dimer and APACH II score were independent factors for death of the septic shock.

Keywords

• Machine learning
• Rick prediction
• Mathematical model
• Cardiovascular color Doppler ultrasound
• Septic shock
• E. coli