Disease diagnosis often involves acquiring medical images using devices such as MRI, CT scan, x-ray, or mammograms of patients' organs. Though many medical diagnostic applications have been proposed; finding subtle cancerous cells is still an issue because they are very difficult to be identified. This paper presents an architecture that utilizes a learning algorithm, and uses soft computing to build a medical knowledge base and an inference engine for classifying new images. This system is built on the strength of artificial neural networks, fuzzy logic, and genetic algorithms. These machine intelligence are combined in a complementary approach to overcome the weakness of each other. Moreover, the system also uses Wavelet Transform and Principal Component Analysis for pre-processing and feature to produce features to be used as input to the learning algorithm. © 2011 Springer-Verlag.