In recent years, along with the progress of minimally invasive surgery that reduces the burden on the patient, there is a growing need for robotic surgery. Japan is one of the world’s leading robot powers and has achieved more than half of the world’s market share of industrial robots, however, no made-in-Japan surgical robot had been launched on the market.

To meet such expectations for medical robots, Medicaroid was jointly established in 2013 by Kawasaki Heavy Industries, Ltd., a leading company of industrial robots, and Sysmex Corporation, an experienced business player which has expertise in inspection and diagnosis and an extensive network in the medical field. Medicaroid has been developing a medical robot based on the concept of “co-existence of humans and robots” since 2015, and finally achieved Japanese regulatory approval for the hinotori^TM Surgical Robot System, as the first made-in-Japan robotic assisted surgery system. This paper describes the details of the hinotori^TM Surgical Robot System.

Recent years, the utilization of deep learning formedical images has spread rapidly in areas such as organ segmentation, computerdiagnostic detection/diagnosis (CAD) and image noise reduction. Especially, application of deep learningreconstruction (DLR) for imagequality improvement has attracted attention. We have developed artificialintelligence (AI) solution invarious fields, such as Advanced intelligent Clear-IQ Engine: AiCE (DLR), Precise IQ Engine: PIQE (super resolution-DLR), SpectralImaging System of a new dual energy technology and Abierto Reading SupportSolution：Abierto RSS (CAD) for workflow improvement of radiologic interpretationefficiency. AiCE is a process that distinguishes between noise and signalcomponents, and achieves a significant noise reduction effect while maintainingspatial resolution, contributing to both high image quality and radiation dosereduction. In order to improve spatial resolution, PIQE uses target data fromSHR mode (0.25mm, 1792ch) of Aquilion Precision, and super resolution processing to improve spatialresolution of data scanned by conventional CT (AquilionONE). Additionally, significant noise reduction andgraininess maintenance effect can be obtained, therefore high resolution imagecan be obtained with lower radiation exposure. This paper describesreconstruction principles of AiCE and PIQE, and introduces their physicalproperties and clinical benefit.

Bolus tracking method is used in 3D-computedtomography angiography (CTA) ofthe head for retrieval of aneurysms, in which a region of interest (ROI)is placed over the cervical vessels to monitor the arrival of contrast medium.However, it is difficult to recognize the cervical vessels on the pre contrastimage, and experience and knowledge are required to place the ROI. Here, recentadvances in AI technology have made it possible to convert images withdifferent characteristics. In this study, we focused on Cycle-GAN, one of theimage conversion technologies, and investigated the technology for generatingcontrast-enhanced CT images from non- contrast-enhanced cervical CT images andits potential application to support for 3D-CTA examination. First, CT imageswere collected for cases in which both non-contrast-enhanced cervical CT andcontrast-enhanced CT examinations were performed, and Cycle-GAN was studiedusing 8-bit images converted from DICOM images. We evaluated the performance ofCycle-GAN in converting non-contrast-enhanced CT images to contrast-enhanced CTimages using the cases that were not used in the training. As a result of theevaluation, pseudo images translated by Cycle-GAN using CT images showedcontrast of only blood vessels and thyroid gland while maintaining the pixel valuesof normal tissues such as bone and thyroid cartilage. Next, thenon-contrast-enhanced CT images obtained during the 3D-CTA examination werethen transformed using Cycle-GAN. Ten radiological technologists were tested todetermine the region of interest of the carotid artery usingnon-contrast-enhanced CT images and pseudo-contrast-enhanced CT images, and thepseudo-contrast-enhanced CT images had a higher rate of correct regiondetermination. These results suggest that the proposed method can convertnon-contrast-enhanced CT images into contrast-enhanced CT images and may beeffective as an assistive technology for radiological technologists who performemergency examination.