The latest developments in artificial intelligence and machine learning with insights into medical imaging were covered at a July 23 workshop, sponsored by the National Institutes of Health (NIH).
NIH Director Francis Collins, MD, PhD, opened the session stressing the importance of artificial intelligence (AI), machine learning (ML), and deep learning (DL) in the institutes’ research portfolio of approximately 667 active research projects, totaling $337 million.
Ronald Summers, MD, PhD, a radiologist at the NIH’s Clinical Center, discussed the latest developments in imaging AI/ML/DL. He highlighted the NIH Clinical Center’s efforts to help address the ongoing need for access to large-scale annotated datasets, including its recent public release of the Deep Lesion Dataset, a compilation of 32,000 annotated lesions identified on computed tomography (CT) images.
Summers also reported an exponential increase of research interest in the application of AI/ML/DL in radiology, encompassing nine percent of all research papers related to health care AI topics. He described how AI has led to large opportunistic population screening and sophisticated queries resulting in developments, such as tumor growth modeling and lesion measuring systems. He predicted new discoveries benefitting individual patients and patient populations through the routine integration of radiology data with other clinical data, triage and critical result monitoring, more automation and quantitation, and changes in radiology practice.
Collins concluded the workshop by outlining NIH goals which included prioritizing data sets, harmonizing and cleaning up data sets to be machine learnable, and developing standards for and enforcing access to NIH-funded data. Collins identified the mammoth “All of Us” research study, cancer genomics and therapeutics, and the BRAIN 2.0 Initiative as projects requiring AI integration.
He suggested building a community for biomedical researchers utilizing AI/ML by creating a better reward system for software that is not just limited to research publications, nurturing more effective interactions with statisticians, physicians, and creating a working group of the NIH Advisory Committee that will report to the NIH director to further outline AI technological developments and identify next steps.
Visit the ACR Data Science Institute website for more information on medical imaging artificial intelligence.