ANN-based computer-aided diagnosis of toxocariasis

The problem of finding information processing means to support the detection of Toxocariasis parasitic disease is considered. It is pointed out that due to the polymorphism of symptoms this disease is often overlooked. Its presence can be confirmed or rejected based on expensive serological tests. It is postulated that artificial neural networks can be used to help doctors decide whether there is an indication to carry out such tests or not. Data from clinical examination and standard laboratory tests are used to make the decision. Experimental data collected for 39 female and 49 male hospital patients were used to train a multilayer perceptron neural network. A comparison was made to classical 1-NN and 3-NN nearest-neighbour classifiers applied to the same data, showing substantially lower classification error rate for the artificial neural network.