VinDr-CXR: An open dataset of chest X-rays with radiologist’s annotations

• DOI: 10.1038/s41597-022-01498-w
• Publication Date: 2022-07-20T16:03:18Z
• Authors: Ha Q. Nguyen; Khanh Lam; Linh T. Le; Hieu H. Pham; Dat Q. Tran; Dung B. Nguyen; Dung D. Le; Chi M. Pham; Hang T. T. Tong; Diep H. Dinh; Cuong D. Do; Luu T. Doan; Cuong N. Nguyen; Binh T. Nguyen; Que V. Nguyen; Au D. Hoang; Hien N. Phan; Anh T. Nguyen; Phuong H. Ho; Dat T. Ngo; Nghia T. Nguyen; Nhan T. Nguyen; Minh Dao; Van Vu
• Keywords: DICOM; Training set; Data set

Most of the existing chest X-ray datasets include labels from a list findings without specifying their locations on radiographs. This limits development machine learning algorithms for detection and localization abnormalities. In this work, we describe dataset more than 100,000 scans that were retrospectively collected two major hospitals in Vietnam. Out raw data, release 18,000 images manually annotated by total 17 experienced radiologists with 22 local rectangles surrounding abnormalities 6 global suspected diseases. The released is divided into training set 15,000 test 3,000. Each scan was independently labeled 3 radiologists, while each consensus 5 radiologists. We designed built labeling platform DICOM to facilitate these annotation procedures. All are made publicly available format along both set.