The article is simplified (a retrospective metastudy) and might not be indicative of what real-life performance. Even reader studies (which would be more rigorous) skip so much that would be crucial to actual deployment (integration into the clinical workflow being one such critical factor).