Hummel A D, Maciel R F, Sousa F S, Cohrs F M, Falcão A E J, Teixeira F, Baptista R, Mancini F, da Costa T M, Alves D, Rodrigues R G D S, Miranda R, Pisa I T
Programa de Pós-graduação em Informática em Saúde, Universidade Federal de São Paulo, São Paulo, Brazil.
Transplant Proc. 2011 May;43(4):1343-4. doi: 10.1016/j.transproceed.2011.02.029.
The gold standard for nephrotoxicity and acute cellular rejection (ACR) is a biopsy, an invasive and expensive procedure. More efficient strategies to screen patients for biopsy are important from the clinical and financial points of view. The aim of this study was to evaluate various artificial intelligence techniques to screen for the need for a biopsy among patients suspected of nephrotoxicity or ACR during the first year after renal transplantation. We used classifiers like artificial neural networks (ANN), support vector machines (SVM), and Bayesian inference (BI) to indicate if the clinical course of the event suggestive of the need for a biopsy. Each classifier was evaluated by values of sensitivity and area under the ROC curve (AUC) for each of the classifiers. The technique that showed the best sensitivity value as an indicator for biopsy was SVM with an AUC of 0.79 and an accuracy rate of 79.86%. The results were better than those described in previous works. The accuracy for an indication of biopsy screening was efficient enough to become useful in clinical practice.
肾毒性和急性细胞排斥反应（ACR）的金标准是活检，这是一种侵入性且昂贵的程序。从临床和经济角度来看，采用更有效的策略对患者进行活检筛查非常重要。本研究的目的是评估各种人工智能技术，以筛查肾移植术后第一年疑似肾毒性或ACR的患者是否需要进行活检。我们使用了诸如人工神经网络（ANN）、支持向量机（SVM）和贝叶斯推理（BI）等分类器来判断提示活检必要性的事件的临床进程。每个分类器通过其各自的灵敏度值和ROC曲线下面积（AUC）进行评估。作为活检指标显示出最佳灵敏度值的技术是SVM，其AUC为0.79，准确率为79.86%。结果优于先前研究中描述的结果。活检筛查指征的准确性足以在临床实践中发挥作用。
