In response to the COVID-19 pandemic in Japan, the Japan Radiological Society made recommendations and issued early stage chest CT examinations guidelines to maintain an appropriate clinical practice. Surveillance on diagnostic imaging of viral pneumonia was conducted using computed tomography to detect the occurrence of COVID-19 pneumonia in Japan. The surveillance was proven to be essential, especially when the PCR testing system was inadequate in the early stages. In April of 2020, during the spread of COVID-19 infections in Japan, we collected chest CT images and created teaching data of COVID-19 pneumonia using the Japan Medical Image Database. Using the data, we then developed AI in diagnostic radiology in collaboration with the National Institute of Informatics.

The National Institute of Informatics has built a cloud platform that can be accessed by the academic network SINET5, and has been developing AI image analysis technology by continuously accepting image data from six medical societies on the cloud platform. Regarding radiological images, we have accepted CT images from Japanese Medical Imaging Database（J-MID）run by the Japan Radiological Society. By October 2020, we have received more than 180 million images. On the other hand, among the cases of coronavirus infection（COVID-19）that occurred around the end of 2019, there are CT images showing pneumonia peculiar to COVID-19. Therefore, we have begun AI image analysis with additional incidental information from the Japan Radiological Society. Multiple universities are participating in this analysis, and it is important to use common data. Therefore, we extracted the data to be analyzed from the large amount of data and created a mechanism to share the information generated by the analysis researchers.

We introduce AIs for diagnosis assistance of COVID-19 patients from CT volumes that were developed in Nagoya University. Novel coronavirus disease 2019（COVID-19）spreads over the world causing the large number of infected patients and deaths. Diagnosis assistance by AI is effective for diagnosing the large number of patients that are caused by infective diseases. We developed diagnosis assistant AIs for COVID-19 cases that evaluate the typical-ness of COVID-19 case based on image appearances from a CT volume. We developed three essential methods for the AIs including the lung region segmentation method, the clustering method of lung region, and the COVID-19 typical-ness evaluation method. To develop the AIs, we utilized huge number of medical images stored in the cloud platform of medical bigdata. We confirmed our AI has high performance for diagnosis assistance in the evaluations using CT volumes of real COVID-19 patients.

We propose a method for classifying the pneumonia by COVID-19 from CT volumes using metric learning. Although PCR method is currently used for COVID-19, the accuracy is low. On the other hand, researches on automatic diagnosis from medical images using deep learning have been done actively, and we considered that it is effective for classification of COVID-19. In this paper, we aim to establish an automatic COVID-19 classification method based on deep learning from CT volumes. The proposed method is classified between COVID-19 pneumonia and other diseases using metric learning, which has been attracting attention in the field of image recognition. We achieved 80％ accuracy in five cross-validation experiments.

We report on the construction of a machine to discriminate positive/negative for COVID-19 based on chest X-ray CT images. All slices of the input CT images are input to the same encoder and are identified as positive/negative for COVID-19 based on their maximum values. In this paper, we report the results of evaluating the impact of（1）different encoders,（2）the presence or absence of segmentation of the lung field region, and（3）the presence or absence of a large amount of additional negative data on the discrimination performance. In particular, although the negative data are collected in overwhelmingly greater numbers than the positive data, simply adding the negative data to the training data causes a significant imbalance in the number of positive and negative data, which hampers the learning of the discriminator. Therefore, we introduce a method to correct the imbalance in the number of data based on the margin and utilize the negative data to improve the discriminator's performance. In this paper, we explain the outline of how to deal with the imbalance in the number of discriminant data and report the experimental results.

In the diagnosis of white matter lesions by MRI, lesions are recognized depending on whether the signal intensity is higher or lower than that of normal white matter. The images used for diagnosis are displayed in gray scale using window processing. The gray level of each tissue is affected by the window processing. Therefore, it is difficult to quantitatively recognize and evaluate white matter lesions on gray scale image. In order to deal with this problem, an image display method and a quantitative evaluation method based on some reference value which is not affected by window processing are required. Therefore, we introduced a mean signal value of the normal white matter as a reference value by which the white matter signal value is normalized. Next, the normalized signal value of the image was obtained by normalizing the signal value with the reference value. Using the normalized signal value, an image display method for quantitative display and quantitative evaluation of the signal value distribution of the white matter high signal region of FLAIR images were investigated. As a result, it was possible to quantitatively display and evaluate the high signal region of the white matter.