The helical reconstruction algorithm was first introduced by Mori. In this method, the data of interest is generated by interpolation of neighboring helical projection data along z-axis, which was followed by convolution and back projection for image reconstruction. The major advantage of this "z-axis interpolation" algorithm is the use of conventional reconstruction technique such as convolution and back projection so that the standard CT image processor can be applied.
This z-axis interpolation theory has been brought into the reconstruction system of current multislice CT with some modification. The actual ways of modification differs from system to system.
It should be noted that the total elimination of cone-angle artifact is impossible as long as z-axis interpolation is used. However, it can be reduced by the geometrical design of the scanner, and by the specific design of the algorithm.
Our scanner -Aquilion- is equipped with new reconstruction algorithm (MUSCOT) which was proposed by Taguchi et al. MUSCOT is designed to reduce the effect of cone angle. In addition, it is suitable to solve various problems associated with multislice CT.
MUSCOT consist of three parts, optimized sampling scan, multipoint weighted filter interpolation, and convolution / backprojection.
One of the problems in multislice helical reconstruction is the sampling density along z-axis. If the conventional helical pith (=pitch 4) is simply extended, the sampling density is not intense enough because of the overlapping of direct data and complementary data.
This shortcoming can be overcomed by sifting the data trail. As is shown by the figure, sifted sampling such as pitch 3.5 successfully increase the sampling density along z-axis. This pitch is not limited to 3.5, but it can be selected freely as long as there is no significant overlap of the data.
After the process of data acquisition has completed, the process of image generation has take place. In this method, a range of certain width is set in the z-axis, and multiple data within this range obtained by optimized sampling are used for filter interpolation. By changing the width and the shape of this z-axis filter, it is possible to freely adjust the effective slice thickness, the slice profile configuration, and the image noise characteristics.
Another merit of multipoint interpolation is the cancellation of cone angle as a result of filtered interpolation of multiple data with a variety of cone angle (figure). This characteristic of MUSCOT allows us to use free-pitch selection. In another system in which the two-point interpolation is employed, the pitch should be fixed to 3 and 6 because the cancellation of cone angle can only be achieved by optimal setting of the scanning pitch.
Furthermore, since the use of multiple data for interpolation brings the averaging effect, signal-to-noise ratio is improved in 20%. This improvement is unique to multipoint weighted filter interpolation which can not be achieved by other multislice CT systems in which two point interpolation are employed.
The selection of scanning pitch is important since it is closely related to the cancellation of the effect of cone angle. There are several "special" pitches which is advantageous in canceling cone angle. Typical examples are pitch-3 and pitch 6. These pitches contain a certain amount of overlapped data in which the cone angle are canceled each other .
In their document, General Electric stated that pitch-3 is superior in image quality, however, no scientific explanation has been given so far. Our explanations are as follows.
1. With pitch 3 scanning, the data sampling density along z-axis is increased by a factor of two in comparison with classical pitch 4. Moreover, the artifacts due to interpolation error are reduced for the benefit of the homogeneous distribution of the data along z-axis.
2. As shown in the figure, pitch-3 contains overlapped data for cone-angle cancellation (arrows). These two data - one is the data from right outermost detector and the other is the data from left outermost detector - poses opposite cone angle. By averaging these two data with reversed cone angle, it is possible to cancel out the effect of cone angle.
3. Another advantage of pitch-3 is the improvement of signal-to-noise ratio. This improvement is mainly due to the overlap scan by pitch-3. In addition, averaging of data from both side of outermost detectors improves data quality. Since X-ray attenuates in proportion to the distance, the outermost detector(s), which has slightly longer FDD than inner-sided detector(s), has smaller number of X-ray photon. For this reason, averaging of two data from outermost detector gives the advantage of improved signal-to-noise ratio.
4. There are some disadvantages in pitch 3 scanning. Most significant shortcomings of pitch 3 is the shortage in scanning range since four rows of detector is used as three. To compensate this shortcoming, the scanner is expected to equip with faster scan capability such as half-second rotation.
5. Increase in patient dose due to overlapped scan will be another disadvantage of pitch 3 scanning.
6. Pitch 3 scanning inevitably require four rows of detectors. The trick of pitch 3 would not work in two-rows sampling such as 0.5-mm x 2. This means that the scanner which equipped with wider selectability of 4-rows of detector is suitable to extract full advantages of pitch 3 scanning.
In conclusion, we agree to General Electric in this particular concept of pitch 3. Our system is considered to be more advantageous in order to extract full advantages of pitch 3 scanning since it is equipped with faster rotation speed and wider range of slice-thickness selectability in 4-rows scanning.
1. Mori, I.: Computerized tomographic apparatus utilizing a radiation source. U.S. Patent No4630202, 1986. -back
2. Taguchi, K., Aradate, H.: Algorithm for image reconstruction in multi-slice helical CT. Med.Phys. 25(4): 550-561, 1998. -back
