Purpose or Learning Objective: To synthesize the existing literature on how the environmental sustainability of artificial intelligence (AI) in medical imaging is being addressed and to identify specific strategies that have been used.

Methods or Background: A scoping review was conducted following the Joanna Briggs Institute methodology. Comprehensive literature search was performed in MEDLINE, Embase, CINAHL, and Web of Science, targeting publications from 2014 to 2024 in English or French. The search used a combination of keywords and MeSH terms related to environmental sustainability, AI, and medical imaging modalities. Three independent reviewers screened abstracts, titles and full texts for eligibility. Results or Findings: The search identified 2812 results, of which 11 met the inclusion criteria. The selected papers comprised 8 research articles, 3 reviews. Three key themes emerged: energy consumption (n=10), carbon footprint (n=4), and computational resources (n=4). The metrics CO2 equivalent, carbon intensity, training time, power use effectiveness, equivalent distance travelled by car were proposed to assess potential AI impact on the environment. Most energy-efficient techniques involved data, AI modelling and training such as data augmentation, data quantisation, lightweight model development, reduction of parameters. Identified strategies to enhance efficiency and reduce environmental impact include (i)integrating energy and carbon metrics in AI evaluation in addition to accuracy assessments, (ii)developing an ecolabel for AI tools, (iii)transitioning to cloud computing and (iv)developing lightweight AI models.

Conclusion: This review identified critical metrics and actionable strategies used to assess and improve sustainable practices in AI for medical imaging which include the integration of specific sustainability-related metrics, cloud computing adoption and development of efficient AI models.