AI-based software for in vitro diagnosis authorized to diagnose prostate cancer

Lim Chang-won Reporter() | Posted : April 3, 2020, 12:28 | Updated : April 3, 2020, 12:38

[Gettyimages Bank]

SEOUL -- An artificial intelligence-based software for in vitro diagnosis, developed by South Korea's medical device company Deep Bio, won state approval, allowing pathologists to diagnose prostate cancer with precision readings. It is South Korea's first authorized AI-based cancer diagnosis software.

The Ministry of Food and Drug Safety said in a statement on Friday that it has approved DeepDx-Prostate, which analyzes prostate cancer tissue images with software learned with artificial intelligence and plays an auxiliary role in diagnosing prostate cancer.

Usually, pathologists diagnose prostate cancer with microscopes after dyeing prostate tissues. DeepDx-Prostate will help clinicians analyze dyed tissue images and detect the presence of prostate cancer tissues, the ministry said, adding that biopsy based on hematoxylin and eosin (H&E) stained digital images was used for AI-based diagnosis.

H&E staining is widely used in medical diagnosis. The stain shows the general layout and distribution of cells and provides a general overview of a tissue sample's structure, helping pathologists to easily differentiate between the nuclear and cytoplasmic parts of a cell.

DeepDx-Prostate uses deep learning for slide image analysis and recognizes cancer in slides, allowing pathologists to diagnose prostate cancer with selective precision readings and reduces the risk of misdiagnosis.

The ministry said that DeepDx-Prostate showed 98.5 percent sensitivity and 92.9 percent specificity compared to the readings of pathology specialists with more than five years of experience in clinical trials.

Sensitivity refers to the proportion of patients testing positive in a target group of people with diseases, while specificity refers to the proportion of those testing negative in a target group of people who have no disease.
© Aju Business Daily & Copyright: All materials on this site may not be reproduced, distributed, transmitted, displayed, published or broadcast without the authorization from the Aju News Corporation.