Machine learning model of imipenem‐resistant <i>Klebsiella pneumoniae</i> based on MALDI‐TOF‐MS platform: An observational study

Background and Aim Machine learning is an important branch supporting technology of artificial intelligence, we established four machine model for the drug sensitivity Klebsiella pneumoniae to imipenem based on matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF-MS) compared their diagnostic effect. Methods The data MALDI-TOF-MS 174 cases K. isolated from clinical specimens in laboratory microbiology department Tianjin Haihe Hospital 2019 January 2020 December were collected. 70 imipenem-sensitive resistant randomly selected establish training set model, 17 sensitive test model. Mass spectral peak subjected orthogonal partial least squares discriminant analysis (OPLS-DA), was by absolute shrinkage selection operator (LASSO) algorithm, logistic regression (LR) support vector machines (SVM) neural network (NN) area under curve (AUC) confusion matrix calculated Grid search 3-fold Cross-validation respectively, accuracy prediction verified matrix. Results R²Y Q² OPLS-DA 0.546 0.0178. AUC best models 0.9726 0.9100, 1.0000 0.8581, 0.8462 0.6263, 0.7180 evaluated LASSO, LR, SVM NN respectively. 87%, 79%, 62%, 68% set, Conclusion LASSO this study has a high rate potential decision ability.