Japanese journal of radiology
-
Knowledge of CT characteristics of COVID-19 pneumonia might be helpful to the early diagnosis and treatment of patients, and to control the spread of infection. ⋯ Our case of COVID-19 pneumonia showed multiple subpleural GGOs in bilateral lung, rapid progression, and it also accompanied nodular GGOs on chest CT. These findings were consistent with the previous reports, and they might be useful for early detection and evaluation of severity of COVID-19 pneumonia.
-
This study aimed to characterize the computed tomography (CT) findings of pseudolithiasis and investigate the outcomes and natural history in adult patients receiving CTRX therapy. ⋯ The high-density sludge pattern is the most common typical CT finding of CTRX-associated pseudolithiasis in adults.
-
Review
CT-based mediastinal compartment classifications and differential diagnosis of mediastinal tumors.
Division of the mediastinum into compartments is used to help narrow down the differential diagnosis of mediastinal tumors, assess tumor growth, and plan biopsies and surgical procedures. There are several traditional mediastinal compartment classification systems based upon anatomical landmarks and lateral chest radiograph. ⋯ These CT-based classification systems are useful for more consistent and exact diagnosis of mediastinal tumors. In this article, we review these CT-based mediastinal compartment classifications in relation to the differential diagnosis of mediastinal tumors.
-
Deep learning has caused a third boom of artificial intelligence and great changes of diagnostic medical imaging systems such as radiology, pathology, retinal imaging, dermatology inspection, and endoscopic diagnosis will be expected in the near future. However, various attempts and new methods of deep learning have been proposed in recent years, and their progress is extremely fast. Therefore, at the initial stage when medical artificial intelligence papers were published, the artificial intelligence technology itself may be old technology or well-known general-purpose common technology. Therefore, the author has reviewed state-of-the-art computer vision papers and presentations of 2018 using deep learning technologies, which will have future clinical potentials selected from the point of view of a radiologist such as generative adversarial network, knowledge distillation, and general image data sets for supervised learning.
-
The "democratization of AI" is progressing, and it is becoming an era when anyone can utilize AI. What kind of radiologists are new generation radiologists suitable for the AI era? The first is maintaining a broad perspective regarding healthcare in its entirety. Next, it is necessary to study the basic knowledge and latest information concerning AI and possess the latest knowledge concerning modalities such as CT/MRI and imaging information systems. ⋯ It is an urgent task to nurture human resources by realizing such a healthcare AI education program to educate radiologists at an early stage. If we can evolve to become radiologists suitable for the AI era, AI will likely be our ally more than ever and healthcare will progress dramatically. As we approach the "democratization of AI," it is becoming an era in which all radiologists must learn AI as they learn statistics.