Fujitsu develops AI to analyze CT studies

2017 06 27 10 04 15 631 Fujitsu 400

Fujitsu Laboratories has unveiled artificial intelligence (AI)-based technology that can find and retrieve CT studies of patients who have similar diseases.

Jointly developed with the Fujitsu R&D Center, the software uses machine learning to find similar cases in which abnormal shadows have spread in three dimensions on the image, according to the vendor. In testing conducted on real-world data at Hiroshima University in Japan, the AI-based application yielded 85% accuracy compared with answers from physicians, Fujitsu said.

Fujitsu noted that technology already exists to retrieve similar cases based on CT images for diseases in which abnormal shadows are concentrated in one place on the image, such as in early-stage lung cancer. Abnormal shadows on the image are spread throughout the organ in all directions in diffuse lung diseases such as pneumonia, however. As a result, it has been necessary for doctors to reconfirm 3D similarities on these cases, which increases the time required to reach a conclusion, Fujitsu said.

Fujitsu's technology first applies machine learning to recognize candidates for abnormal shadows from CT images. Next, it automatically estimates the boundaries of the core and the periphery of the lung based on the relatively clear parts of the CT images and, in succession, divides the lungs into core and peripheral areas. From there, the technology creates histograms of the number of abnormal shadow candidates located in those areas and then looks at the 3D spread of abnormal shadows to retrieve similar cases, according to the company.

Fujitsu's AI-based technology is designed to find previous CT studies that represent similar disease cases. Image courtesy of Fujitsu.Fujitsu's AI-based technology is designed to find previous CT studies that represent similar disease cases. Image courtesy of Fujitsu.

By dividing up the organ spatially into areas and focusing on the spread of the abnormal shadows in each area, it becomes possible to assess the images in the same way as doctors do when determining similarities for diagnosis, Fujitsu said. The firm believes that its technology has the potential to shorten the diagnostic time required to evaluate these cases to one-sixth of previous levels.

Researchers shared details about the technology at the Pattern Recognition and Media Understanding conference held June 22-23 at Tohoku University. Fujitsu now plans to conduct field trials using CT images for a variety of cases.

The goal is to deploy this technology with related offerings from Fujitsu Limited, according to the firm. In addition, the technology is not just applicable for the diagnosis of diffuse lung diseases; it could also be used for other imaging studies such as head CT, stomach CT, MRI, and ultrasound exams, Fujitsu said.

Page 1 of 653
Next Page