Refraction-based computed tomography using x-ray dark-field imaging(XDFI-CT), proposed by M. Ando in 2002, can obtain high-definition three-dimensional structures of biomedical samples. Since the advent of XDFI-CT, many methods including hardware and software techniques have been developed to improve the image quality and the imaging efficiency. In the paper, we first introduce the imaging principle of XDFI-CT, and then outline the reconstruction techniques to suppress artifacts and to reduce the views, which have been recently developed for XDFI-CT. Finally, we demonstrate the XDFI-CT images acquired from several pathological samples, such as human iliac artery, human liver, and human breast nipple, to show the effectiveness.

View full abstract

X-ray phase tomography that maps the three-dimensional distribution of refractive index was invented in early 1990s by myself, and various X-ray phase-contrast techniques, such as the method for detecting X-ray phase shift or refraction (phase-shift differentiation) and the propagation-based phase retrieval, have been applied to X-ray phase tomography. Recently, the degradation in phase contrast caused by unresolvable scatterers is used as a signal for tomography to visualize three-dimensional scatterers distribution. Although synchrotron radiation was considered to be indispensable for X-ray phase-contrast generation for many years, X-ray phase tomography compatible with a laboratory X-ray generator has been feasible by X-ray Talbot interferometry or X-ray Talbot-Lau interferometry, which detect X-ray refraction and scattering by using X-ray transmission gratings. This technology has been attracting attention from industry, and I will introduce recent progress in this trend.

View full abstract

Optical coherence tomography, which we call as OCT, is non-destructive imaging technique of inside structure of biological tissue with um resolution, using broadband light source and optical interferometer. Thanks to the advance of related technologies and improvement of imaging speed, OCT have been used in several medical fields, especially ophthalmology. We have been investigating ultrahigh resolution OCT using broadband supercontinuum pumped by ultrashort pulses. The performance of OCT imaging depends on the optical wavelength range. The wavelength range of 0.8 μm is suitable for high resolution OCT imaging. Recently, since the magnitude of scattering coefficient is low and there is local minimum of water absorption, the wavelength range of 1.7 μm absorbs a lot of attentions in terms of the large penetration imaging. In this article, based on the author's works, the high resolution OCT technologies will be reviewed from the fundamental to the recent results.

View full abstract

This paper shows examples of micro anatomical structure imaging using a desktop micro-focus X-ray CT scanner. Currently clinical X-ray CT scanners can take CT images in 0.5mm to 1mm per voxel in resolution. Such scanner can take anatomical structures close to image resolution of scanners. Micro-focus X-ray CT (micro CT) scanner can take CT images of 1μm per voxel to 50μm per voxel in resolution. Such scanner enables us to explore micro anatomical structures. This paper shows several examples of micro CT images of lung and heart specimens with discussion on future direction.

View full abstract

In this paper, we propose a new airway segmentation algorithm from 3D chest CT volumes based on the volume of interest (VOI). The algorithm segments each bronchial branch by recognizing the airway regions from the trachea using the VOIs to segment each branch. A VOI is placed to envelop the branch currently being processed. Then a cavity enhancement filter is performed only inside the current VOI so that each branch is extracted. At the same time, we perform a leakage detection scheme to avoid any leakage regions inside the VOI. Next the gradient vector flow magnitude map and a tubular-likeness function are computed in each VOI. This assists the predictions of both the position and direction of the next child VOIs to detect the next child branches to continue the tracking algorithm. Finally, we unify all of the extracted airway regions to form a complete airway tree. We used a dataset that includes 50 standard-dose human chest CT volumes to evaluate our proposed algorithm. The average extraction rate was approximately 78.1% with a significantly decreased false positive rate compared to the previous method.

View full abstract

Statistical shape models (SSM) of organs have played important roles in medical image analysis applications, including organ segmentation and assessment of the degree of abnormality. The SSM construction methods can be categorized by the shape representation methods, among which the level set distribution model (LSDM) and the point distribution model (PDM) are two of the most popular SSMs. However, there have been little discussion on the difference between LSDM and PDM. In this paper, after the brief introduction to the basics of the level set based shape representation and the LSDM construction method, difference between LSDM and PDM is discussed from the several viewpoints.

View full abstract