Proceeding of the 2011 IEEE Students' Technoloy Symposium
14�16 January, 2011, IIT Kharagpur
ASIC Implementation of a 512-point FFT/IFFT
Processor for 2D CT Image Reconstruction Algorithm
Indranil Hatai\ Rakesh Biswas2, Swapna Banerjee3
Indian Institute of Technoloy Kharagpur,
Department ofElectronics & Electrical Communication Engineering
Kharagpur,lndia-72I302
"indranilh@cse.iitkgp.ernet.in
'rakeshbiswas2003@gmail.com
3
swapna@ece.iitkgp.ernet.in
Abstract- The CT scan medical imaging requires huge amount
of computations for reconstructing the images. Modiied Fast
Radon Transform (MFRT) uses FFT based parallel algorithm
for reconstruction of 2D/3D CT images from its sonogram data
using convolution operation by concurrent ID FFT/IFFT and
matrix multiplications. To achieve optimum hardware utilization
with low power consumption an FFT module, based on iterative
radix-2 decimation in frequency (DIF) algorithm, has been
designed and implemented. The module has been designed in
such way so that it can also be used for IFFT computation only
by changing a single parameter. To compute the FFT, the
twiddle factor has been calculated using Coordinate Rotation
Digital Computer (CORDIC), by steering the data properly in
the butterly structure. The synthesized frequency of FFT/IFFT
module is 220 MHz and gate count is 1,040,136 using 130 nm
faraday digital libraries. The power has been analyzed using
prime power and the value of the power consumption is 15mW.
The designed FFT/IFFT ASIC chip is very much suitable for
low-area and low power biomedical applications like CT image
reconstruction, Doppler wave spectrogram etc.
Keywords- Modiied Fast Radon Transform (MFRT), CT images,
FFTIIFFT, Coordinate Rotation Digital Computer (CODIC).
I. INTRODUCTION
Recent advances of CT scanner technology (1-3] have
effectively yielded real-time scanning performance, making
the monitoring of dynamic events such as cardiac motion
possible. However, the reconstruction of images in
interventional and minimally invasive surgery requires high
performance computing solutions that meet operational room
demands [4-8]. To achieve a higher spatial, contrast, or
temporal resolution, the dimensions of input and output
datasets used by the reconstruction methods are increasing
rapidly [9]. The application of CT imaging is effectively
limited by the performance of a single CPU computer as the
reconstruction speed is concened [9]. The 'FRT' algorithm
(1-2] could solve the shortcomings of the Beylkin's 'DRT'
algorithm [3]; however, the'Aliasing effect' still remained as
the main drawback of this method [1-3]. The antialiasing
solution provided in inal form of the FRT algorithm gave rise
to other subsequent diiculties such as the varying
computational complexity with the worst case complexity as
high as O(N
4
). The 'zero-padding' technique effectively
increases the aspect ratio r of an image and so as the
maximum allowable projection angle, where Max(8)=tan' \r)
(1]. However, it is practically impossible to achieve the
required minimum angular coverage (±900) for complete
reconstruction of an image rom its projection. This is because
at the limiting condition (±90
o
) the aspect ratio r approaches to
c. Therefore in spite of its several advantages the FRT is
considered as an incomplete algorithm with respect to
complete reconstruction. Also, the 'zero-padding' of an image
signiicantly increases the computational complexity of this
algorithm [1]. The Modiied-FRT (FRT) algoritm splits an
image into two angular parts viz. angular region below
tan-
l
(±r) and angular region above tan-
l
(±r). For a square
image, the image is divided for the angular region below and
above ±45°, respectively. The coordinate axes X and Y are
considered to be opposite for these two images. FRT and
IFRT [1] are now applied on both of these images. For a same
input image the column-wise operations (e.g. FFT, IFFT) of
the FRT algorithm are just changed to row-wise operations
with the change in axes and vice-versa. In either of the cases
angular sampling only rom 45° to +45° is considered. In this
method a complete reconstruction is possible without zero
padding the image. This results in a very simple
reconstruction algorithm with a signiicant reduction in
computational complexity of the FRT algorithm. The
proposed algorithm only contains the operations like
FFT/IFFT and Vector-Matrix multiplications, which can be
computed either by eficient sotware programming or by
using some dedicated processors to urther speed-up the
computation process. However, both the methods has the
same computational complexity, O(N
3
).
The block diagram of FRT processor to compute 2D radon
transform at high speed is shown in Fig. 1. It consists of two
parallel processing elements (PE), each constructed using FFT,
vector matrix multiplier and IFFT unit.
. iotm
Fig. 1 Block diagram of MFRT processor
TS11VLSIOl173 978-1-4244-8943-5/11/$26.00 ©2011 IEEE 220