A new tool for engineering education: Hepatitis diagnosis using associative memories

Classification is one of the key issues in medical diagnosis. In this paper, a new tool for engineering education is presented: it is an automatic hepatitis diagnosis system based on associative memories. The characteristic of this approach is twofold: first, learning the fundamental set of associations in order to get an associative memory; second, computing a differential associative memory in order to get a threshold value for each unknown input pattern to be classified. Hepatitis disease dataset, taken from UCI machine learning repository, was used as medical dataset. Classification accuracy of the proposed approach is 82.67% and it was assessed using stratified 10 fold cross-validation. The correct diagnosis performance of the proposed approach is validated not only using classification accuracy, but also performing sensitivity and specificity analysis. The results presented in this paper demonstrate associative memories potential for automatic medical diagnosis systems.