Deep learning versus human graders for classifying diabetic retinopathy severity in a nationwide screening program
| Autor: | Bilson J. L. Campana, Pipat Kongsap, Rajiv Raman, Peranut Chotcomwongse, Chaiyasit Thepchatri, Korntip Mitvongsa, Greg S. Corrado, Surapong Orprayoon, Srirut Kawinpanitan, Sukhum Silpa-archa, Jitumporn Fuangkaew, Kasumi Widner, Chetan Rao, Jirawut Limwattanayingyong, Jeffrey Tan, Siriporn Lawanasakol, Lalita Wongpichedchai, Oscar Kuruvilla, Ramase Sukumalpaiboon, Jesse J. Jung, Chawawat Kangwanwongpaisan, Sonia Phene, Kornwipa Hemarat, Jonathan Krause, Lily Peng, Mongkol Tadarati, Paisan Ruamviboonsuk, Lamyong Chualinpha, Chainarong Luengchaichawang, Sarawuth Saree, Rory Sayres, Dale R. Webster |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2019 |
| Předmět: |
Developing world
medicine.medical_specialty education.field_of_study business.industry Diabetic macular edema Population Medicine (miscellaneous) Health Informatics Diabetic retinopathy medicine.disease lcsh:Computer applications to medicine. Medical informatics Article Computer Science Applications Health Information Management Diabetes complications Internal medicine Diabetes mellitus medicine lcsh:R858-859.7 education business Reference standards Kappa |
| Zdroj: | NPJ Digital Medicine npj Digital Medicine, Vol 2, Iss 1, Pp 1-9 (2019) |
| ISSN: | 2398-6352 |
| Popis: | Deep learning algorithms have been used to detect diabetic retinopathy (DR) with specialist-level accuracy. This study aims to validate one such algorithm on a large-scale clinical population, and compare the algorithm performance with that of human graders. A total of 25,326 gradable retinal images of patients with diabetes from the community-based, nationwide screening program of DR in Thailand were analyzed for DR severity and referable diabetic macular edema (DME). Grades adjudicated by a panel of international retinal specialists served as the reference standard. Relative to human graders, for detecting referable DR (moderate NPDR or worse), the deep learning algorithm had significantly higher sensitivity (0.97 vs. 0.74, p p p p |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|