| Budget Amount *help |
¥2,500,000 (Direct Cost: ¥2,500,000)
Fiscal Year 2005: ¥500,000 (Direct Cost: ¥500,000)
Fiscal Year 2004: ¥1,000,000 (Direct Cost: ¥1,000,000)
Fiscal Year 2003: ¥1,000,000 (Direct Cost: ¥1,000,000)
|
|---|
| Research Abstract |
For the purpose of developing a high speed and high precision X-ray CT image reconstruction system for soft tissues, we investigated the previously proposed 2D-image reconstruction method in computed tomography based on the non-uniform sampling theorem, from the viewpoints of both software and hardware implementations, in order to enhance its performance, i.e., sensitivity reduction to additive noises in projection data acquisition and highly speedy image reconstruction, and the following main results are obtained :
(1)By automatically extracting the spatial frequency spectrum superior to the estimated mean level of the noises at high frequency, sensitivity of the system to them is reduced to an amount lower than that of the 2D-FT method, which is the most precise one of the conventional. As a result, precision of the system is enhanced by 6dB to 10dB, compared with that of the previously proposed one under the detected SNR of 40dB to 20dB, while its reduction being suppressed only by a
… More
n amount less than or equal to 0.5dB when no noise exists.
(2)The newly developed method mentioned above is effective for reducing computation time. It can save 74% of the time by the previously proposed, but needs an extra time by the amount of 30% more than that by the conventional 2D-FT one. Since the developed method works further effectively with increase of the noise level, it is expected to reconstruct the image in a computation time comparable with the latter in actual applications.
(3)With a high speed DSP board of floating point type, computation time is reduced to 64% of that by the previous method without any reduction of precision, while it is to 38% with reduced precision of about 10dB when the DSP of fixed point type being used. In this research, to detour storing the vast amount of sine and cosine table required for 2D-IDFT of non-uniformly sampled spatial frequency spectrum, the previously proposed speedy iterative algorithm based on the additive law of triangular functions is used, but contrary, the algorithm is considered to inhibit the high speed calculation ability by the pipeline processing equipped with the DSPs.
(4)By computer simulation based on the developed algorithm of iterative parallel processing, it is confirmed that iteratively reconstructed images can be observed every a pre-designated amount of projection data is acquired, and the final image can be reconstructed almost in a computation time inversely proportional to the number of used personal computers, just after all the projection data have been acquired. Less
|
