A FPGA-based embedded system for automatic classification of microcalcifications

This paper describes the development of an embedded system that automatically classifies microcalcifications detected on digital mammograms into one of the five types proposed by Michele Le Gal, a classification scheme that allows radiologists to decide whether a breast cancer is malignant or not without the need for surgeries. The hardware part of the developed system is based on an Altera Nios II software processor and the embedded software is based on wavelets and artificial neural networks. We have used an Altera DE2-US development kit in order to create a custom System-on-Chip (SoC) that has many advantages over common desktop computers. In our tests the system correctly classified 94.90% of test images, proving it can be used as a second opinion by radiologists in breast cancer early diagnosis.