常见伤口感染细菌高场不对称波形离子迁移谱识别算法

目的:使用模式识别算法对常见伤口感染细菌（大肠杆菌、金黄色葡萄球菌、铜绿假单胞菌）TH肉汤培养液及纯TH培养液的高场不对称波形离子迁移谱（FAIMS图谱）进行分类识别。 方法:使用FAIMS分析仪收集了4种样品的训练及测试样本。预处理后，用主成分分析和线性判别分析对样本进行降维和特征提取，得到了训练集和测试集的空间三维分布模型，再用最近邻规则算法进行样本识别。 结果:特征提取后，正负模式样本均具有良好的分离效果，并且正模式可分性明显优于负模式。当K取合适的值时，正负模式识别率分别可达90%和70%以上。对于本文的数据模型，K取值等于或最接近每类样本数的奇数为最佳。 结论:该种算法可用于常见伤口感染细菌肉汤培养液FAIMS图谱的分类及识别。

Algorithmic identification for high-field asymmetric ion mobility spectra of common wound-infecting bacteria

Objective To classify and recognize the high-field asymmetric ion mobility spectra (FAIMS) of common wound-infecting bacteria (escherichia coli, staphylococcus aureus and pseudomonas aeruginosa) cultured in thioglycolate (TH) broth and bare TH broth by pattern recognition algorithm. Methods The spectra of training set and testing set of four types of samples were obtained by FAIMS analyzer. After pretreatment, principal component analysis and linear discriminant analysis were used to reduce the dimension and extract features for obtaining the three-dimensional spatial didtribution patterns of training set and testing set. Finally, k-nearest neighbors algorithm was used to identify the four types of samples. Results After feature extraction, the samples showed good separation effects under both positive and negative modes, and the separability of positive mode was superior to that of negative mode. When K was appropriately assigned, the recognition rate under positive and negative modes could be up to 90% and 70%, respectively, or even higher. For the data model in this study, the optimal K was equal to the odd number closest to sample number of each type. Conclusion The proposed algorithm can be used to classify and recognize the FAIMS of common wound-infecting bacteria cultured in TH broth.