E2928. A Comprehensive Overview of the Current State of Artificial Intelligence in Radiology
  1. Robert Hubley; Nassau University Medical Center
  2. Rahul Anand; Nassau University Medical Center
  3. Saurabh Patel; Nassau University Medical Center
  4. Steven Lev; Nassau University Medical Center
In the presence of an overwhelming amount of literature regarding artificial intelligence (AI) in radiology, we seek to elucidate the state of the art of AI in radiology from different perspectives – that of the radiologist, the clinician and the patient. We discuss medico-legal, economic, safety, and workflow concerns. We aim to alleviate many of the fears and worries that radiologists and medical students may have about the future role of AI and how it may affect their livelihoods.

Educational Goals / Teaching Points
Medico-legal (Ethics): Investigate, based on different levels of implementation of AI, who will ultimately be responsible for diagnosis? Investigate how much the creators of AI algorithms (programmers, etc.) will be held accountable if something goes wrong? Economic: Investigate how much it costs, on average, to implement different levels of AI into radiology programs at various hospitals Investigate return-on-investment (ROI) of implementation of AI in radiology programs. Patient Safety: Investigate patient perceptions regarding how they would feel if their doctor recommends a treatment plan based on AI imaging findings. Investigate the effectiveness of AI alone vs radiologist vs both in the diagnosis of various conditions, and how that affects patient outcomes. Workflow: Investigate how AI will integrate in the radiologist’s workflow: Will AI partially or completely replace the radiologist? Will AI be used for detection purposes? Will it be integrated into the workflow to improve efficiency?

Key Anatomic/Physiologic Issues and Imaging Findings/Techniques
Imaging Findings. Compare the effectiveness of AI alone vs Radiologist at detecting imaging findings in various conditions. Imaging Techniques, Machine Learning, Unsupervised, Reinforced, Supervised, Semi-supervised, Deep Learning, Artificial Neural Networks.

While AI has many promising applications in the field of radiology, there is a vast amount of AI research that is still ongoing in order to determine its final role. Currently, there is no danger of AI replacing the radiologist in the near future because there are a lot of ongoing issues, ranging from medico-legal to economic, that limit its application. Regarding the near future, AI will assist the radiologist in detection and/or workflow because data continues to show that AI is effective at pinpointing imaging findings that can help the radiologist with diagnosis and/or treatment.