Studying And Implementation Fixed-point Algorithm On Cone-Beam CT

The first time in2002Katsevich proposed a spiral cone-beam filtered backprojection （Filtered Back-projection, FBP） and precise reconstruction algorithm. Thisalgorithm is more accurate. And the main object of this article is such algorithms.Second, the fixed-point arithmetic in hardware implementation, compared to thefloating-point hardware implementation can be obtained faster speed, lower powerconsumption and smaller area, so the fixed-point arithmetic has important applicationsspace. For the full custom ASIC chip, designers will use the fixed-point hardwaremechanism. The main purpose of this article is dedicated ASIC implementation of theCT area optimization.In the beginning of the article, Katsevich FBP algorithm is studied and derivated.For the simplification, the whole algorithm is divided into five parts, simultaneousextraction of each part of the data flow diagram for the lower part of the fixed-pointanalysis. In the fixed-point algorithm, we choose algorithms based on affine arithmetic（AA） in range analysis and quantization error propagation model algorithm （QOE） andanalog adaptive annealing algorithm （ASA） in the precision analysis. For the resultsobtained, the article lists the main signal the integer bits wide and fractional bits wide,precision analysis, different a precision requirements to eradicate this paper theaccuracy of analysis results get2^-4，2^-8，2^-13three groups for the different results. Thispaper uses MiniBit area function estimates of the various parts of the area consumption.The results compared with the one form Matlab tool. In the range of analysis andprecision analysis part of this article is a more optimized while better than the other interms of area consumption. We select CORDIC algorithm linear processing algorithmtranscendental functions. With the requirements of the medical use of images, finallydetermined in the hardware implementation, CORDIC algorithm required pipelinestages.