Emu Mahzabeen, Kamal Farjana Bintay, Choudhury Salimur, Alves de Oliveira Thiago E
Annu Int Conf IEEE Eng Med Biol Soc. 2020 Jul;2020:5382-5387. doi: 10.1109/EMBC44109.2020.9176542.
Fibrosis is a significant indication of chronic liver diseases often due to hepatitis C Virus. It is becoming a global concern as a result of the rapid increase in the number of HCV infected patients, the high cost and flaws associated with the assessment process of liver fibrosis. This study aims to determine the features that significantly contribute to the identification of the stages of liver fibrosis and to generate rules to assist physicians during the treatment of the patients as a clinically non-invasive approach. Also, the performance of different Multi-layered Perceptron (MLP), Random Forest, and Logistic Regression classifiers are estimated and compared for the full and reduced feature sets. Decision Tree produced 28 rules in contrast with previous research work where 98002 rules had been generated from the same dataset with an accuracy rate of approximately 99.97%. The resulting rules of this study achieved a prediction accuracy for the histological staging of liver fibrosis of 97.45%. Among all the machine learning methods, MLP achieved the highest accuracy rate.
纤维化是慢性肝病的一个重要指征，通常由丙型肝炎病毒引起。由于丙型肝炎病毒感染患者数量的迅速增加，以及肝纤维化评估过程的高成本和缺陷，它正成为一个全球关注的问题。本研究旨在确定对肝纤维化阶段识别有显著贡献的特征，并生成规则，作为一种临床无创方法，在患者治疗过程中协助医生。此外，还针对完整和简化特征集估计并比较了不同多层感知器（MLP）、随机森林和逻辑回归分类器的性能。与之前的研究工作相比，决策树产生了28条规则，之前的研究工作从同一数据集中生成了98002条规则，准确率约为99.97%。本研究得出的规则对肝纤维化组织学分期的预测准确率达到了97.45%。在所有机器学习方法中，MLP的准确率最高。
