腹部CT诊断的专家级通用人工智能

Qi Zhang†, Jianpeng Zhang†, Weiwei Cao†, Zilin Lu†, Wanxing Chang†, Haonan Ding† et al. INTRODUCTION: Radiology is critical for modern clinical diagnosis, enabling the detection and characterization of diverse diseases. A long-standing vision in the field is to realize a generalist artificial intelligence (AI) capable of expert-level diagnostic support for varied real-world scenarios. Contrast-enhanced abdominal computed tomography (CT) poses one of the most complex challenges, requiring assessment of dozens of organs and hundreds of conditions. Despite milestones in AI for specific domains, current AI solutions remain limited in disease coverage and accuracy. They largely depend on supervised learning with expensive manual annotation, restricting scalability and applicability in open-ended clinical contexts. …