[A decision-support system for hematology]

Thrombocytopenia and following bleeding at the treatment of hematological malignancies is a dangerous complication. The indication of thrombocyte transfusion is the key point for the therapy and proylaxy of bleeding. The all problem is divided into two parts. Evaluation of the risk of bleeding (80% of decision), estimation of the risk of aloimunization and risk of connected with the transfusion (20% decision). For now we are concentrated to the evaluation of the risk of bleeding. In the first stage we are concentrated to statistical evaluation of values to define factors possibly highering the risk of bleeding. Factors were determined with help of two test, GUHA method and using literature. For recognized factors were trained 3 layer neuron nets with a non-linear method pack propagation. After that an application was developed to determine the risk of bleeding for a routine use in clinical practice.