Telling someone he or she has cancer, is serious. To tell someone who is concerned that he or she has cancer, that he or she DOES NOT have cancer is even more serious. Sometimes even difficult. I wouldn't imagine relegating such a responsibility to an unemotional unemphatic robot!

Thanks dear abu, but i talk about diagnosis, not telling the patient

That's a very interesting question Mohammad. I really don’t know if artificial methods are better than typical methods in diagnosis of cancer. However, I think the more accurate findings in all medical specialties, and that includes clinical oncology, are possible when we associate both artificial (exams, questionnaires) and typical methods (interview and physical exams). It’s interesting you pointing out this topic because recently I was trying to find a professor to guide me in preparing a questionnaire to evaluate the need to use certain equipment for the diagnosis of prostate cancer, based in modern statistics techniques (fuzzy, item response theory, etc.)

dear augusto, i think you should read this paper: http://www.sciencedirect.com/science/article/pii/S0957417410012054

and fuzzy logic is my special field in artificial intelligence, most of my publication are in this matter