Recognition and prediction of leukemia with Artificial Neural

  Abstract   Background : Leukemia is one of the mostcommon cancers in children, comprising more than a third of all   childhood cancers. Newly affected patients in USA are estimated as 10100cases, and if these cases are diagnosed late or proper treatment is not applied, then it can be mortal. Because rapid and proper diagnosis of leukemia based on clinical or medicinal findings (without biopsy) is impossible, we decided to apply artificial neural network for rapid leukemia diagnosis. For this aim we used clinical and medical parameters taken from 131 patients of Sina hospital of Hamadan.   Methods : We carried out independent sample T-test with SPSS software for 38 parameters. With regard to   the results of this analysis we selected 8 parameters that had lowest sig for ANN analysis (among parameters, whose sig were less than 0.05). Selected parameters of 131 patients were applied for training network with Levenberg-Marquardt learning algorithm, with learning rate of 0.1.   Results : Performance of learning was 0.094. The Relationship between the output of trained network for test data and real results of test data was high and the area under ROC curve was 0.967.   Conclusions : With these results we can conclude that training process was done successfully and accurately. Therefore we can use artificial neural network for rapid and reliable leukemia recognition.