Lvova Larisa, Martinelli Eugenio, Dini Francesca, Bergamini Alberto, Paolesse Roberto, Di Natale Corrado, D'Amico Arnaldo
Department of Chemical Science and Technologies, University Tor Vergata, Rome, Italy.
Talanta. 2009 Jan 15;77(3):1097-104. doi: 10.1016/j.talanta.2008.08.021. Epub 2008 Sep 2.
The Electronic tongue (ET) composed of different kind of potentiometric chemical sensors has been applied for the detection of urinary system dysfunctions and creatinine levels. The creatinine contents evaluated by ET were compared with those obtained by automated Jaffe's method and GC-MS, obtaining a satisfying agreement for both methods. Partial least square regression discriminate analysis (PLS-DA) and feed forward back-propagation neural network (FFBP NN) classified 51 urine specimens from healthy volunteers in four classes, according to the creatinine content, showing that both techniques can satisfactorily differentiate urines according to this parameter. The best accuracy result of 92.2% correct classification of unknown samples was achieved with FFBP NN. Moreover, the possibility of ET system to distinguish between urine samples of healthy patients, and those with malignant and non-malignant tumor diagnosis of bladder has been shown.
由不同类型的电位化学传感器组成的电子舌（ET）已被应用于泌尿系统功能障碍和肌酐水平的检测。将电子舌评估的肌酐含量与通过自动杰氏法和气相色谱-质谱法获得的含量进行比较，两种方法取得了令人满意的一致性。偏最小二乘回归判别分析（PLS-DA）和前馈反向传播神经网络（FFBP NN）根据肌酐含量将51份健康志愿者的尿液样本分为四类，表明这两种技术都能根据该参数令人满意地区分尿液。前馈反向传播神经网络对未知样本的正确分类准确率最高，达到了92.2%。此外，还展示了电子舌系统区分健康患者尿液样本以及膀胱癌恶性和非恶性肿瘤诊断患者尿液样本的可能性。
