Magnetic resonance imaging (MRI) provides excellent soft-tissue visualization, thereby increasing the sensitivity for detecting tumor and normal tissue boundary. Recently, hybrid MR scanners and linear accelerators (MR-Linac) have been developed, including the MRIdian (ViewRay, Oakwood, USA) and Unity (Elekta, Stockholm, Sweden). This article explained the MR-linac system and related medical imaging research.

Radiogenomics for lung cancer can be considered a “dry biopsy” and may support conventional wet biopsy. It may provide gene-related information of patients with lung cancer, who cannot undergo conventional biopsy or have spatially heterogeneous tumors. In this review paper, we share the basic flows and results of our research on radiogenomics, which focused on discovery of associations between image features and genes in lung cancer, developments of prediction models for HOPX, EGFR mutation, and its subtypes.

Advances in post-genome research have revealed the molecular and genetic characteristics of cancer. By using these knowledge, molecular targeted drugs have also been developed. Precision medicine, in which treatment is com bined with molecular classification based on the cancer genotype, is being performed in clinical practice. Radiogenomics is research to estimate genotype from cancer phenotype, and its purpose is to create innovations to maintain the competitive advantage of diagnostic imaging in cancer genomic medicine and to expand the role of imag ing test. This article describes the current state of radiogenomics for breast cancer. They are (1) subtype classification of breast cancer, (2) difference from CAD for differential diagnosis, (3) relationship with blood test, (4) prediction of the efficacy of preoperative drug therapy, and (5) application to preventive imaging medicine.

There has been enormous research interest in using radiomics or deep learning to predict genetic mutations within gliomas. These technologies were expected to solve the qualitative nature of radiographical images and allow us to analyze them quantitatively. Furthermore, these technologies were promised to make a direct connection between the biological characteristics of the tumor and radiographical information. On the other hand, it has now become clear that radiomics and deep learning have particular problems, such as over-fitting and domain shift, which must be solved to render them clinically applicable. The ideal diagnostic accuracy is still debatable, and it is still being determined whether radiomics or deep learning can achieve a correct diagnosis, even when non-tumor lesions are included. This review paper discusses the current state, issues, and future perspective of radiomics and deep learning in the glioma imaging research area.

Parallel imaging (PI), a fast-scanning technique, has been developed to shorten the scan time of magnetic resonance imaging. However, this technique generates a wide range of noise level and distributes them non-uniformly in the reconstructed image, which requires appropriate noise reduction according to the region and noise level of the image. In this study, we propose a method to adaptively improve the image quality according to the noise level of the image using multiple trained convolutional neural networks (CNNs) even when the training data are scarce. In our method, the training data are separated into multiple datasets based on the g-factor map, which indicates the features of the noise distribution in the imaging region, and a CNN is trained on each dataset. We defined Blur as an index of the low definition of a fine structure and evaluated the performance of the proposed method using PI images captured at three-times high speed as an input. As a result, we have confirmed that Blur is lower than that of a single CNN, and the signal-to-noise ratio exceeds +70 ％, which is equivalent to that of a full-sampled image.

Dynamic X-ray imaging and analysis system has received JAMIT Technological Achievement Award. The imaging system is consisted of a dynamic flat-panel detector with a large field of view and X-ray generator/tube capable of pulsed irradiation. For the evaluation of pulmonary function, the movement of the diaphragm, rib cage, and structures in the lungs, as well as density changes in the lung field due to respiration and pulmonary blood flow, are quantified and visualized on the obtained dynamic chest radiographs. The product was commercialized in November 2018 through animal testing and initial clinical research through industry-academia-government collaboration. The system is in operation at approximately 70 facilities in Japan (as of February 2023), and its usefulness in various functional diagnoses has been reported in many cases. This paper describes an overview of a dynamic X-ray imaging and analysis system that realizes lung function imaging in the general radiography room, and perspectives on this field.