1887
banner image
No data available.
Please log in to see this content.
You have no subscription access to this content.
No metrics data to plot.
The attempt to load metrics for this article has failed.
The attempt to plot a graph for these metrics has failed.
Minimizing image noise in on-board CT reconstruction using both kilovoltage and megavoltage beam projections
Rent:
Rent this article for
USD
10.1118/1.2768862
/content/aapm/journal/medphys/34/9/10.1118/1.2768862
http://aip.metastore.ingenta.com/content/aapm/journal/medphys/34/9/10.1118/1.2768862

Figures

Image of FIG. 1.
FIG. 1.

Illustration of kV and MV angular coverage. With 105-deg gantry rotation, kV and MV beams have a 15-deg overlap, which is used for converting contrast information between kV and MV projections.

Image of FIG. 2.
FIG. 2.

Conversion and composition of sinograms. (a) and (b) represent the kV and MV sinograms, respectively. (c) shows the converted MV sinograms. (d) shows the composed sinogram.

Image of FIG. 3.
FIG. 3.

CT study on a head phantom. (a) and (b) are the MV and kV CT images with BBs inserted, respectively. (c) The aggregated MV/kV CT image, showing less streak artifact than the kV CT image and better contrast than MV CT image. (d) The kV CT image after removing the BBs. The kV bowtie filter was not used.

Image of FIG. 4.
FIG. 4.

Comparison of the two kV/MV conversion models for the contrast phantom. For the model of using 360-deg projections, the kV and MV CT images are reconstructed by the FBP method as shown in (a) and (b). Then by comparing (a) and (b) in the five highlighted ROIs, a kV/MV conversion model with is obtained in (c). For the model of using 15-deg projections, limited-angle kV and MV CT images are reconstructed by the FBP. As shown in (d) and (e). By comparing (d) with (e), a kV/MV conversion model with is obtained in (f). Limited-angle kV and MV CT images are also reconstructed by the BP method as shown in (g) and (h). The consequent kV/MV conversion model has higher correlation coefficient but the slope of the model was lower than the slope in (c) and (f). The kV bowtie filter was not used in the scan.

Image of FIG. 5.
FIG. 5.

Aggregated CT images for phantom A which includes a contrast phantom and two pieces of solid-water phantoms. (a) shows the planning CT image which has five ROIs highlighted by circles. (b)–(i) are the eight aggregated CT images. The image quality of these images shows a periodic change with the start angle for the gantry rotation. Note the solid-water phantoms are not displayed inside the fields of views. The kV bowtie filter was not used in the scan.

Image of FIG. 6.
FIG. 6.

Comparison of the profiles of the kV and MV projections along the PA and -lateral directions for phantom A. The profiles of the digital radiograph (DRR) calculated from planning CT scan are plotted as ground true values. Both kV and MV projections are scaled to match the DRR ones. The logarithmic transform has been performed for all the projections.

Image of FIG. 7.
FIG. 7.

Comparison of the profiles of the kV and MV projections along PA and -lateral directions for phantom B. The profiles of the DRR calculated from planning CT scan are plotted as ground true values. Both kV and MV projections are scaled to match the DRR ones. The logarithmic transform has been performed for all the projections.

Image of FIG. 8.
FIG. 8.

Linearity of calibrating aggregated CT values for the five ROIs. The CT values inside five ROIs were calibrated for both phantoms A and B with two beam orientations, and 135. All calibration points are fitted closely by the line except two points from phantom A with .

Image of FIG. 9.
FIG. 9.

The standard deviations of CT values for the five ROIs. are plotted against the gantry start angle. Phantom A shows a periodic change for all five ROIs. The peak-to-valley ratios for the five ROIs are 1.8, 2.9, 2.9, 1.8, and 1.5, respectively. Phantom B has much lower and the changes of are also small. The peak-to-valley ratios for the five ROIs are 1.2, 1.3, 1.4, 1.6, and 1.7, respectively.

Image of FIG. 10.
FIG. 10.

The CNR for the ROI in phantoms A and B are plotted with the gantry start angle . The CNR of phantom A has a periodic change with the peak-to-valley ratio of 2.89. Two maximums at and and two minimums at and match the corresponding angles in the image quality study in Fig. 8. Phantom B has much higher CNR but with smaller peak-to-valley ratio of 1.36.

Image of FIG. 11.
FIG. 11.

Aggregated CT images for phantom C which is a pelvic phantom laterally placed. (a) is the planning CT image. (b)–(i) are the eight aggregated CT images. The image qualities are highly correlated with the beam orientation. The kV bowtie filter was used in the scan.

Image of FIG. 12.
FIG. 12.

NMI between the planning CT and aggregated CT. The NMI exhibits a sinusoidal oscillation with the gantry start angle and the amplitude of oscillation for phantom C is bigger than that of phantom D. The NMIs were also calculated for the images reconstructed with kV-only and MV-only projections. They are 1.26 (kV) and 0.67 (MV) for phantom C and 0.86 (kV) and 0.52 (MV) for phantom D.

Image of FIG. 13.
FIG. 13.

Comparison of aggregated CT images reconstructed by the FBP and SART methods. The gantry start angel is . The iteration number is 10 and the relaxation factor is 0.2 for the SART method. The standard deviations of the CT values for the highlighted region of interest were calculated to be 14.2, 17.0 18.9, 11.9, 21.4, and 56.6 for (a)–(f). The arrow in (e) indicated some artifacts caused by the discontinuities between the kV and MV projections with same angles. The SART method was less sensitive to the discontinuities because the iterative method can smooth these artifacts. Aliasing artifacts were also stronger in (f) than in (c), indicating the advantage of the ART method when the number of projections is low.

Image of FIG. 14.
FIG. 14.

The effect of MV beam truncations on the image quality. (a) shows an aggregated CT image reconstructed using full kV beams with the beam angles from to and truncated MV beams with the beam angles from to . The field size of MV beams was only one quarter of the filed size of kV beams. (b) shows the image without truncating the MV beams. The kV bowtie filter was used. The effect of the bowtie filter was not corrected completely, which can still be observed at the edges of the field of view.

Image of FIG. 15.
FIG. 15.

Effects of the kV/MV conversion on image quality. The three models included model A which matched kV and MV CT values, model B which underestimated MV CT values by , and model C which overestimated MV CT values by . The planning CT is shown in (a) with the five highlighted regions of interest. The gantry state angle is the same for all the three models.

Image of FIG. 16.
FIG. 16.

Effect of the conversion model on the CT values. The three models use the same calibration factor obtained in Fig. 4.

Tables

Generic image for table
TABLE I.

Comparison of the MV/kV slope for three phantoms by two methods which used 360-deg and 15-deg projections, respectively. The slopes vary between different phantoms but the ones given by the 15-deg method are very close to the ones by the 360-deg method.

Loading

Article metrics loading...

/content/aapm/journal/medphys/34/9/10.1118/1.2768862
2007-08-27
2014-04-20
Loading

Full text loading...

This is a required field
Please enter a valid email address
752b84549af89a08dbdd7fdb8b9568b5 journal.articlezxybnytfddd
Scitation: Minimizing image noise in on-board CT reconstruction using both kilovoltage and megavoltage beam projections
http://aip.metastore.ingenta.com/content/aapm/journal/medphys/34/9/10.1118/1.2768862
10.1118/1.2768862
SEARCH_EXPAND_ITEM