New white paper provides framework for AI/ML in cardiovascular imaging
Friday, August 30, 2024
(0 Comments)
Posted by: Jessica Frizen
ARLINGTON, VA (August 30, 2024) — A new white paper from the Society of Cardiovascular Computed Tomography (SCCT) offers current state-of-the-art applications of artificial intelligence (AI) and machine learning (ML) in cardiovascular computed tomography (CT), highlighting challenges and considerations for implementation in both research and clinical settings, and exploring the future of this technology. The white paper – “Artificial Intelligence and Machine Learning for Cardiovascular Computed Tomography (CCT)” – outlines eight areas within the cardiovascular CT field in which AI/ML could aid, including: patient selection and screening, referrals and scheduling, image acquisition and reconstruction, image analysis and diagnosis, report generation, prognostication and risk stratification, management recommendations and follow-up. “There is great potential for AI/ML to improve patient care, and for cardiovascular CT this area is evolving rapidly,” said Michelle Williams, MBChB, PhD, FSCCT, writing group co-chair. “However, there are important challenges that must be considered during the development and use of AI/ML tools, and this white paper highlights both the potential benefits and challenges ahead.” Co-led by Williams and Jonathan R. Weir-McCall, MD, FSCCT, the writing group said that in order for its full potential to be realized, “challenges related to validation, trust, paucity of prospective and diverse setting evaluations, and generalizability must be addressed.” Published in the Journal of Cardiovascular Computed Tomography (JCCT), the scientific document first outlines the potential for AI/ML to streamline scheduling, workflow and patient selection processes for cardiac imaging teams. For example, the writing group explains autonomous decision-making based on previously collected patient data can aid the process of patient selection, which has grown more complex as multi-modality cardiac imaging becomes more technologically advanced. “With the integration of cardiac imaging guidelines into electronic medical records, AI/ML can move the field towards more robust imaging pathways and provide suggestions for test choice and ordering, which can otherwise be challenging for the busy physician,” the authors wrote. When it comes to image acquisition, reconstruction and optimization, the authors share that AI/ML can lead to further reductions in radiation dose by improving image reconstruction algorithms and “de-noising” low radiation dose images. “[Deep learning] DL can also facilitate motion correction by preserving imaging features and estimating the direction and magnitude of coronary motion,” the authors said. The group also outlines the potential of AI/ML in disease-specific areas and focal points, including coronary artery disease, structural heart disease, vascular disease, thoracic fat and non-cardiovascular findings. They further outline emerging biomarkers including radiomics, as well as the use of AI/ML to combine data and improve prognostication. The authors stress that, despite this “enormous potential for AI/ML to improve patient care,” important challenges remain and must be considered for the development, assessment and implementation of AI/ML to ensure that it is safe, reliable, cost effective and improves outcomes for patients. According to the authors, challenges for cardiovascular CT in AI/ML research include dataset size, diversity, creation, storage and access, as well as reference standards and model development, evaluation and interpretation. The authors also share specific challenges and considerations for reporting and reviewing AI/ML papers. In terms of current challenges for cardiovascular CT AI/ML in clinical practice, the scientific document offers guidelines, standards, frameworks and regulations currently in place, as well as considerations when both appraising for deployment and implementing new AI/ML technologies in clinical practice. The paper further describes challenges with reimbursement, and legal and ethical aspects to consider in clinical settings. The authors identified the following priority areas for SCCT, to optimize the development and use of AI/ML in cardiovascular CT: - Keep patient care at the center of all activities regarding AI/ML
- Articulate the most important clinical needs that may be addressed through AI/ML solutions
- Create unbiased education opportunities that complement AI/ML advances
- Encourage interoperability and standardization to facilitate the development and deployment of AI/ML tools
- Encourage diversity and representativeness in the development and assessment of AI/ML to ensure generalizability and minimize bias in clinical practice
- Provide guidance to regulatory bodies regarding the most pertinent metrics to assess the safety and efficacy of new AI/ML technologies in clinical practice
- Pursue high quality evidence that demonstrates whether AI/ML technologies provide added value, including assessment of cost effectiveness
- Develop strategies for fair and equitable access to appropriate AI/ML technologies
- Support a framework for adjudicating the technical development of AI/ML technologies
- Be adaptive to rapid technological change in AI/ML to support clinicians, patients and other society members
- Assess suitability and relevancy of AI/ML technologies for inclusion in future guidelines
### About the Society of Cardiovascular Computed Tomography Founded in 2005, the Society of Cardiovascular Computed Tomography (SCCT) is the international professional society devoted to improving health outcomes through effective use of cardiovascular computed tomography (CCT). SCCT is a community of physicians, scientists and technologists from over 85 countries advocating for access, research, education and clinical excellence in the use of CCT. For more information, please visit www.SCCT.org. About the Journal of Cardiovascular Computed Tomography
The Journal of Cardiovascular Computed Tomography (JCCT) is a peer-review journal of the SCCT that integrates the international cardiovascular CT community and addresses a broad range of topics affecting cardiovascular CT imaging. The journal’s major focus is on original research and on the clinical and technical aspects of cardiovascular CT. For more information, please visit www.journalofcardiovascularct.com. ###
|
Print Page
Email to a Friend
Newest FSCCTs
Newest FSCCTs