Simulation-Based Performance Analysis of a Medical Image-Processing Architecture

摘要

In this paper, we show a simulation method for performance prediction of medical image processing chains, based on the real-time calculus of [9]. In particular, we focus on estimating the latency and throughput of a given image processing chain and its distribution over hardware resources. The architectural level of abstraction of the real-time calculus approach makes our method flexible, so that different design decisions can be studied using similar models. The choice for simulation rather than the usual algebraic analysis of real-time calculus equations, gives us the possibility to include a number of performance relevant implementation details in our models for which analytical estimates are not available.