This paper proposed an intelligent medical diagnostic supporting model for an e-healthcare infrastructure, which automatically acquires practical and useful knowledge and regulations from massive and historical medical data, to assist in making diagnostic and treatment decisions. We propose to explore the hidden usefulness of false irrelevant attributes, and take their supportive correlation into pre-processing. Moreover, we suggest to mimic learning in real world, which is dynamic, incremental and from multiple dimensions. Thus, incremental learning should be dynamic enough to deal with new attributes other than new instances. The empirical results reveal that our model with our novel methodologies is indeed a valuable tool in supporting diagnostic and treatment decision-making for the e-healthcare infrastructure.