Digital trigger algorithms performed over large sequences of data can be an efficient way to shift processing time from baseline samples or singles to coincident photo multiplier tube pulses. This paper describes a coincidence processing system intended to implement fully digital trigger and pre-processing algorithms for a positron emission tomography scanner dedicated to breast imaging. In order to efficiently address all the emerging issues of high resolution PET detectors, a high performance DSP processor has been embedded into the backbone of the system. Signals from 12 channels of a dual-head PET camera are acquired in free-running sampling while a first stage of FIFO memories, implemented using a Virtex-II Pro FPGA, translate data from a sequential sample-by-sample processing basis to a more efficient block-by-block one. This approach enables us to carry out trigger and pre-processing tasks in parallel. Moreover, the scheme offers additional benefits: intrinsic temporal coherence, zero data acquisition dead time, and a more flexible software approach to pre- and post-processing issues (from pile-ups and scatter correction to pre-reconstruction processing). It also heavily reduces the bandwidth required for the link to the host computer, enabling the use of a high speed USB port. The DAQ system is capable of handling count rates up to 10 Mevents/s, pre-processing the samples on the fly and lastly delivering data to the host computer for image reconstruction.