Keywords:
Performed at one institution, Experimental, Prospective, Tissue characterisation, Image registration, Segmentation, Physics, Computer Applications-Detection, diagnosis, CT-Quantitative, CT, Radiation physics, Computer applications, Physics in Medical Imaging
Authors:
A. Yoneyama1, R. Baba2, M. Kawamoto3, T.-T. Lwin4; 1Tosu, Saga/JP, 2Tokyo/JP, 3Tosu/JP, 4Sagamihara/JP
DOI:
10.26044/ecr2020/C-08185
Results
1. Feasibility test of phantom
Fig. 6 shows a sectional (CT) image of the phantom obtained with 20-keV SR (a) and an illustration of the phantom (b). CT observations were performed by rotating the phantom vertically in the air over 360 degrees. The exposure time for each projection image was 1.0 s, and the projection number was 1000.
Developed segmentation was tested by using CT spectrum data obtained with 20-, 25-, 30-, and 35-keV SR. Fig. 7 (a) shows the calculated SOM map, which shows three big main clusters corresponding to iodine, aluminum, and acrylic resin. Fig. 7 (b) and (c) show an SOM image segmented with a k-means of 10 and a segmented image of the phantom, respectively. All components of the phantom were clearly segmented without training data, as shown in Fig. 7 (c). The distance between the clusters in the SOM map indicates the similarity of the clusters, and the red and blue regions show long and short distances, respectively. Therefore, the spectrum of he iodine was completely different from that of the acrylic resin and background (air), and that of the aluminum was similar to that of the calcium powder. The result makes sense because the absorption edge of iodine was 33 keV and the spectrum changed rapidly from 30 to 35 keV while the other spectrums simply decreased.
2. Feasibility test of biological sample
Fig. 8 shows a sectional (CT) image of a rat tail obtained with 15-keV SR. CT observations were performed by rotating the sample horizontally in formalin solution over 360 degrees. The exposure time for each projection image was 1.5 s, and the projection number was 1000. The structure in the bone region was clearly visualized because of its large absorption. The sample preparation was approved by the President of Kitasato University through the judgment of the Animal Care and Use Committee of Kitasato University (approval no. 14-02).
Fig. 9 (a) shows a SOM map obtained by using the CT spectrum data acquired with 15, 20, 25, 30, and 35-keV SR. Three main clusters corresponding to the muscle, bone, and background region appeared. Fig. 9 (b) and (c) show a SOM image segmented with a k-means of 12 and a segmented image, respectively. Not only bone, muscle, hair, and the background region but also detailed structures in the bone were segmented very clearly. In addition, the SOM image indicates that the spectrum of the bone was different from that of the muscle and hair.