NEURAL-NETWORK PREDICTION OF RELAPSE IN BREAST-CANCER PATIENTS

When a woman diagnosed as having breast cancer has a tumour removed, i t is important to try and predict whether she is likely to relapse wit hin, say, the next three years. In this paper, the performance of a ne ural network classifier trained on a number of prognostic indicators i s shown to be better than that of the clinical experts working with th e same information. To obtain meaningful statistics with the relativel y small dataset available, the network is trained using a modified for m of the leave-one-out method. A procedure is also introduced for inve stigating how much independent information each input parameter contri butes. This shows that, in this type of retrospective study, the type of therapy given to the woman does not significantly affect the networ k's prediction of whether or not she will relapse within three years. Finally, since this problem, in common with many other medical problem s, is plagued by a shortage of data, the final section of the paper re ports on an investigation of whether or not multi-centre databases mig ht be feasible.