An improvement in integrating clustering method and neural network to extract rules and application in diagnosis support

Most of chronic liver diseases without suitable treatment will lead to cirrhosis of the liver, eventually progressing to liver cancer. Thus, early diagnosis is very important in detecting the liver diseases and suggesting the treatment at the right time. A useful model that effectively predicts the patient's liver fibrosis has great importance in reducing the load on doctors, especially in lower-level hospitals. In this paper, a new model combining semi-supervised learning method and fuzzy min max neural network with selective fuzzy rule set rendering is proposed. Cirrhosis level is evaluated by APRI and FIB-4. The proposed method is experimented on data sets from machine learning databases, including UCI and CS. Apart from that, our method is also implemented on the liver data set collected from the hospitals of Thai Nguyen province. The comparison among our proposed method and other related ones is also given. The obtained results show that our proposed model has better performance than compared methods in terms of execution time and the number of rules.

An improvement in integrating clustering method and neural network to extract rules and application in diagnosis support Keywords:

Artificial neural network , semi-supervised clustering , chronic liver diseases , liver disease diagnosis , cirrhosis