This article first introduces the generation principles of synchrotron radiation (SR), its main performance characteristics, and major synchrotron facilities in Japan. It then provides an overview of various SR-based imaging techniques utilizing synchrotron radiation, discussed from the three perspectives of spatial, temporal, and density resolution. Furthermore, the article presents imaging studies conducted by the author, including an outline and measurement examples of the cryo-micro X-ray CT system operating at beamline BL07 of the Kyushu Synchrotron Light Research Center (SAGA-LS), and the principle, instrumentation, and biomedical imaging examples obtained using a two-crystal X-ray interferometric imaging system installed at beamline BL-14C of the Photon Factory (KEK PF).

X-ray Talbot interferometry is an approach for X-ray phase CT, and because it does not use crystals, its applications are expanding to various research purposes, such as high-resolution measurements, quantitative measurements, and even dynamic measurements. This article introduces examples of X-ray phase CT measurements obtained using an X-ray Talbot interferometer at SPring-8, focusing on applications to biomedical samples, and looks ahead to the future.

Recent technological advances have made it possible to analyze the elements in the body precisely using an attogram scale. These elements are derived from those in the space around us, suggesting that our body is part of that space. Thus, it is natural that elements (e.g., minerals and metals) are essential for a healthy body. On the other hand, the intracellular distribution of elements and its function are not well understood, although many studies related to biomolecules, e.g., proteins and nucleic acids have been conducted at the molecular level. In this study, we describe the development of a scanning X-ray fluorescence microscope system (SXFM) that can reliably determine the cellular distribution of multiple elements with resolutions for viewing intracellular organelles. Visualizing intracellular elements and understanding their kinetics may provide great insight into the behaviors of elements at the molecular level in biology and medicine.

Superimposed Wavefront Imaging of Diffraction-enhanced X-rays (SWIDeX) is a novel refraction-contrast X-ray CT imaging technique in which a thin Si single-crystal analyzer plate is placed in direct contact with the X-ray camera to acquire projections corresponding to the second derivative of the refractive index distribution 𝛿. This configuration enables close proximity between the specimen and the detector, which has been difficult to achieve in conventional phase-contrast imaging, thereby significantly reducing image blurring caused by the finite X-ray focal spot size. In this study, we present the imaging principle and reconstruction theory of SWIDeX, compare its image quality with that of XDFI using human biological tissue specimens, and demonstrate that the algebraic reconstruction method incorporating Total Variation regularization effectively reduces the required number of CT projections by exploiting the sparsity of SWIDeX-CT.

Synchrotron radiation (SR)-based micro X-ray CT enables high-resolution three-dimensional observations with the micrometer scale by leveraging the high intensity, parallelism, and monochromaticity of SR. However, image degradation due to motion artifacts caused by sample displacement during long acquisition times, which typically range from several minutes to several hours, remains a significant challenge. In this study, we developed a software-based method to suppress such artifacts by detecting sample displacement through matching projection images acquired at rotation angles of 0°, 180°, and 360°, followed by displacement correction. The underlying principle of this approach, an overview of SR-based micro X-ray CT, and its performance, demonstrated through both numerical simulations and experimental data, are presented.

In SPECT, off-set acquisition method can reduce the total number of projections. However, no clinical systems have been developed to support this, because a special mechanism is required to shift the detector slightly from even intervals. In this study, we proposed a novel acquisition method for triple-detector, and investigated its effectiveness. The proposed method changes the step angle only once without acquiring opposite data, and projection data are arranged at even intervals in consideration of the opposite data. Therefore, the need for a special mechanism is eliminated. The projection data were generated using a Monte Carlo simulation, considering factors such as scatter, attenuation, and collimator aperture. Images were reconstructed using ML-EM. Although there were no significant differences in spatial resolution, the NRMSE of the proposed method with an even number of projections (less than 24) was better than that of the conventional method. The proposed method obtained accurate reconstructed images because the projection data of the proposed method were balanced at 360° to avoid distortion.

In magnetic resonance imaging (MRI), motion artifacts vary in intensity and direction, which prevents accurate image diagnosis. However, through use of transfer learning and employing gradient-weighted class activation mapping (Grad-CAM) to create a classification model, the characteristics of these classifications can be clarified. Previous studies have indicated that the model exhibited a greater response to more prominent artifacts, and this response distribution may depend on the difference in the phase-encoding direction of the image. The study aim was to create motion-artifact classification models for different phase-encoding direction and use Grad-CAM to evaluate their effects. Using T2-weighted images from 25 head MRI cases, we performed computational simulation to create 9,000 images with artifacts. The data were divided into 70% training, 30% validation, and GoogLeNet was used to perform transfer learning. The test accuracy was 91.5% for the combined phase-encoding direction of the vertical and horizontal models, 96.4% for the vertical model, and 96.7% for the horizontal model. The Grad-CAM results reflected the characteristics of the models, and the combined model did not show any phase-encoding direction effects.