Artificial Intelligence Task Force


MISSION STATEMENT FOR AI TASK FORCE

The mission of the AI TASK FORCE of the ASNR is to both foster the development of training, educational programs and lifelong learning and promote research initiatives in artificial intelligence (machine learning and deep learning) in neuroimaging for neuroradiologists, trainees, scientists, and associated professionals. For the ASNR, the AI TASK FORCE will develop the leadership needed to successfully translate and implement AI into routine clinical workflow and practice while maintaining the highest standards in patient care.

Vision Statement for AI at ASNR

The ASNR is the nationally and internationally recognized primary leader and information source in all aspects of neuroimaging AI for all neuroradiologists and our professional collaborators in neurology, neurosurgery, and neuropsychiatry and related fields as well as researchers, developers, industry leaders, and regulatory officials. In collaboration with other societies such as RSNA and ACR, the ASNR identifies value propositions for neuroimaging AI that drive improvements in diagnosis, treatment, and management to enhance patient care and promote precision health and wellness.

STRATEGIC AIMS:

  1. TRAINING AND EDUCATION: (AI WORKING GROUP)

    1. RESIDENTS and FELLOWSHIP TRAINING
      1. Curriculum dedicated to machine learning/AI
      2. Access to workshops in AI (ASNR/ASFNR)
    2. TRAINING FOR NEURORADIOLOGY FACULTY
      1. Workshops co-run by ASNR/ASFNR
      2. ISMRM workshops
  2. RESEARCH: (AI WORKING GROUP)

    1. ASNR AI Grant
    2. R and E Fund of RSNA (another funding source for AI grants)
    3. Seed grants in AI from subspecialty societies
    4. NIH grant mechanism
    5. Development of a consortium like IPSS (international pediatric study group) to enhance data sharing and collaboration to spur development, testing, validation and implementation of neuroimaging AI.
      • Data discovery
      • Anonymization
      • Curation
      • Dataset design to minimize all forms of bias
      • Minimizing Dataset shift
      • Standards for markup/annotation
      • Translation of research to vendor space
  1. CLINICAL IMPLEMENTATION:

    1. Goal: Identification of value propositions for workflow and clinical neuroimaging problems with AI
    2. QA Issues
      1. Standardization, Reliability and Reproducibility (work with NIH MIST)
      2. Seamless interoperability across all PACS and EMR/electronic workflow systems
      3. Establishment of QA assessment practice (how to monitor efficacy over time? ACR standard? NIH?)
      4. Compliance with regulatory environment and legal frameworks
      5. Ethical considerations on AI and neuroimages including bias (conscious and unconscious)
      6. Patient safety
      7. Testing of Algorithms (see below on Consortium of Neuroimaging AI Experts)
    3. Consortium of Neuroradiology AI Experts (ASNR AI TASK FORCE & AI WORKING GROUP)
      1. Access to databases with large repositories of neuroradiology images for algorithm development which needs ease of access, efficiency of access, and scalability. Do these data bases serve as training, testing and/or validation sets?
      2. Need for a standard approach for vetting AI technology prior to implementation into routine clinical process? A validation function for groups developing AI technology.
      3. Requirements for publication (making algorithms available to other researchers)
  1. GROWTH:

    1. Establishment of Consortium as a pathway for AI developers in Neuroradiology
      1. Creation of this collaboration working group to expedite creation and use of AI neuroimaging solutions/products to solve clinical work flow problems using neuroimaging
      2. AI with input from key stakeholders (i.e. neuroradiologists who will actually use it if useful)
      3. Develop solutions/products with neuroimaging AI to improve diagnosis, treatment, management, and patient outcomes (give us tools so that we can succeed in delivering better health care, improving access, and reducing costs)
      4. Partner with industry by leveraging multidisciplinary and multispecialty experts in the development of neuroimaging AI solutions
  1. CHALLENGES:

    1. Scalability
    2. Cost efficiency
    3. Automation is needed prior to widespread acceptance by stakeholders (Neuroradiologists, clinical colleagues, insurers, and hospital administrators) into routine clinical practice

Password Request

Enter the Username for your user account, then click Submit. We will email you a temporary password.

Find Username

Enter email address for your user account, then click Submit.