Liver disorder detection based on artificial immune systems

Liver is one of the most important internal organs in human body which performs many vital functions such as filtering out toxins and harmful chemicals. However, liver disorder is usually difficult to diagnose because its symptoms can be vague and sometimes easily confused with other health problems. In this research, algorithms based on artificial immune systems (AIS) are employed for liver disorder detection by analyzing blood test results. Two modified real-valued negative selection algorithms based on AIS are proposed; their performances are compared with artificial neural networks (ANN) and other typical classification methods via computer simulations. It is shown that both the modified variable-radius algorithm (MVD) based on AIS and the ANN model yields better results than traditional algorithms such as C4.5 detection trees etc. Though ANN outperforms the MVD algorithm by a small margin, it fails to perform when only “normal” (or “self”) data are available for training. By contrast, the new proposed MVD algorithm can be trained by “normal” (or “self”) samples only and still yields a competitive detection rate of 81.18%. Note this feature of AIS is especially important for the field of biomedical research, where “normal” samples are usually much easier to obtain.