The prototype optical CT scanner: (a) schematic diagram and (b) full setup. Diagram in (a) is not to scale.
The scanner features two collimator options: (a) a single-slot collimator, and (b) a precision machined multihole collimator with 0.7 × 0.8 mm2 holes cut for each detector element [note: photodiode arrays and daughter boards are attached in (b)].
A scattering phantom (left) and an irradiated polymer gel dosimeter (right) are shown. Insets show examples of simple fiducial marks used for “rough” registration purposes.
The workflow diagram for artefact removal and image reconstruction. Beginning with a 360° absorbance sinogram (upper left), this diagram illustrates all actions involved in obtaining a reconstruction (lower right). Tasks in the middle with dashed outlines are those related to artefact removal techniques. These tasks are skipped if artefact removal techniques are not to be implemented.
Simple treatment plans for the gel dosimeters: (a) single beam, (b) cross beam, and (c) sagittal plane illustrating the placements of each distribution. Dashed lines in (c) indicate the locations of the slices sampled and the approximate dimensions of the gel.
Detector signal versus collimator depth for a single photodiode detector measuring light transmitted through a highly opaque sample. The benefit of increased collimator depth is shown to diminish around 1.5 cm.
The multihole (MH) and single-slot (SS) collimators were used to obtain (a) measured absorbance values versus theoretical absorbance values for a range of blue dye concentrations. In (b), the same plots are shown for a range of scattering agent concentrations. In both plots, a dashed line is used to indicate ideal values (i.e., A m = A th ) and error bars (±1σ) are shown. Opaque regions shown in each plot indicate absorbances beyond the current system's fully extended dynamic range.
Reconstructions of a uniform scattering solution. All three images use the same scan data, I, but different reference data, I ○. Reference data used were (a) a light profile through an empty bath, (b) an average profile of all profiles from a light sinogram acquired through a water-filled flask, and (c) an angular-dependent profile from a light sinogram acquired through a water-filled flask. Ring artefacts that remain in (c) are attributed to data corruption that occurs in the detector array. Note that the outermost ring in each image corresponds to the wall of the flask. The radial span of a disk ROI used for noise analysis is indicated in (c). The same window and levelling values are used in all three images.
Variations in surface quality of a water-filled flask. In sinogram space: (a) relative (%) light fluctuations per detector element over 360°, (b) map of rays deviating >10% from their average, (c) relative difference per ray position between two identical scans, and (d) map of rays with >10% deviation in the second scan. In a 50 mm vertical scan: (e) relative light fluctuations per detector element through the scan, (f) map of rays deviating >10% from their average, (g) relative difference per ray between two identical scans, and (h) map of rays with >10% deviation in the second scan.
Images of a scattering solution reconstructed from (a) an unaltered sinogram, and (b) a sinogram with seams and rings removed and interpolated-for, as described in Sec. II C . Two image profiles [location indicated by dashed line in (b)] are shown in (c) with and without artefact removal. Regions of interest ROI1 and ROI2 are indicated in the upper and lower sections of (a), respectively.
A light survey of angular and vertical positions near a sought after “zero” position allows for precise registration through noise analysis. The image in (a) shows relative standard deviation values ( ) for each position in the survey with the zero position being the least noisy pixel in the survey (here, the central pixel). Profiles in (b) and (c) show how these relative standard deviation values vary with respect to angular and vertical mismatches, respectively.
Percent difference images illustrating the performance of the flask registration technique: (a) flask left unmoved between I ○ and I scans, (b) data deliberately mismatched 1 mm vertically, (c) data deliberately mismatched 1° angularly, and (d) flask fully removed and zeroed using the registration technique. The ROI used for quantification is shown in (a).
Initial reconstructions of a cold dosimeter obtained by: (a) comparing scan data (I) of the irradiated slice to reference data (I ○) obtained through an unirradiated slice of the dosimeter, or (b) comparing scan data of the same slice before and after being irradiated (I ○ and I, respectively). Identical window and level values are used in both images.
Demonstration of the dramatic effect gel temperature and quick cooling can have on results obtained with gel dosimeters. The first protocol, shown in (a) and (b), placed liquid gels into a cold bath in a refrigerator in order to speed cooling. Both preirradiation and postirradiation scans were taken of the dosimeter while cold using a room temperature matching bath. The second protocol, shown in (c) and (d), allowed gels to cool overnight in a room temperature bath. Here, preirradiation and postirradiation scans were taken of the dosimeter at room temperature in a room temperature bath. Identical window and level values are used in all four images.
Preliminary noise reduction results. Filtered images were obtained using adaptive mean filtering in sinogram space. Binned images take the average of 64-pixel groups to obtain lower resolution images. Profiles in (g–j) compare raw data, filtered data, binned data, and data that has been filtered and binned against expected values calculated by treatment planning software. Positions of these profiles are indicated in (a) and (d). Images in (a–f) are all identically windowed and levelled.
Article metrics loading...
Full text loading...